{ "exported_at": "2026-11-15 20:44:15", "count": 36, "records": [ { "id": 134, "title": "A Support System for Diacritic-aware Classical Arabic Language Processing: Integration of Speech, Text, and Vision Modalities", "title_ar": "", "title_en": "A Support System for Diacritic-aware Classical Arabic Language Processing: Integration of Speech, Text, and Vision Modalities", "author": "Adel AbdelSabour Ahmad Aly", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "University of Washington", "university_source": "repository_host", "college": "", "department": "", "country": "United States", "governorate": "", "publication_year": 2024, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "machine-learning; deep-learning; computer-vision; nlp; ai-applications; multimodal; image-segmentation; llm; machine-translation; speech-recognition; arabic-nlp; ai-education", "keywords": "", "abstract": "Artificial intelligence (AI) impresses us daily, outperforming humans in complex games and tasks. Yet, AI and Large Language Models (LLMs) stumble to grasp a language that is thousands of years old. It's Arabic, where subtle diacritical marks can completely alter a word's meaning. Top language models like ChatGPT and Google's Gemini face challenges with Arabic's unique features, potentially leading to critical misunderstandings. The main obstacle is adapting successful NLP systems from other languages to Arabic without understanding its distinct nuances. This dissertation presents a support system—a Multimodal Integration System—for diacritic-aware Classical Arabic language processing. The system integrates speech, text, and vision modalities to address the unique challenges of Arabic's rich linguistic features, such as diacritics and linguistic styles. Arabic has multiple correct linguistic styles, each preserving the same text but with different diacritics. These variations reflect regional dialects, adding meaning, and altering grammar and rhetoric. The Holy Quran, with its 20 linguistic styles based on a single core text, serves as our ideal dataset. We developed innovative databases and models that push the boundaries of Arabic language processing. Our scalable databases store texts in 7 different Arabic linguistic styles. QR-Vision excels at recognizing precise diacritics in images. QRDiaRec adds diacritics to un-diacritical text in various styles. QRSR and DASAM specialize in speech processing and alignment for Arabic diacritical text and voice. SemSim stands out as a dual-space similarity explorer, analyzing numeric and semantic data. Our methodology involves advanced techniques in data modeling, data quality validation, deep learning, computer vision, signal processing, and interactive data visualization.Our findings demonstrate improvements in addressing key challenges in Arabic Natural Language Processing (NLP). For text processing, we created an Automated Diacritization Deep Learning model. The model supports multiple Arabic diacritical styles, a unique feature in the field. Our best-performing model achieved a 94.2% accuracy rate in adding correct diacritics to Arabic text. In image processing, we built a specialized Optical Character Recognition (OCR) model for diacritic-aware Arabic text. Our OCR model reached an accuracy rate of 91.67% showing an improvement over the existing models. For speech processing, we developed two key systems. The first system, operating at the sentence level, combines our novel FuzTPI algorithm with machine learning models. This hybrid approach achieved up to 96% accuracy in audio segmentation and text-audio classification. Our second system focuses on word-level segmentation and alignment for Arabic diacritic-based speech. It achieved R² values of 0.959 for word start times and 0.957 for end times. These results show an improvement over existing Arabic speech recognition technologies. The dissertation is structured around these three modalities, with each section detailing the challenges, methodologies, and results achieved in processing Arabic with diacritics. They have far-reaching implications for applications such as machine translation, information retrieval, speech recognition, natural language understanding, educational technology, and the preservation of linguistic heritage. By addressing the unique challenges of Arabic diacritics across multiple modalities, this research paves the way for more nuanced and culturally sensitive AI applications in Arabic-speaking contexts.", "source_url": "https:\/\/openalex.org\/W7130276073", "pdf_url": "", "doi": "", "handle": "", "repository_name": "ResearchWorks at the University of Washington (University of Washington)", "source_record_id": "W7130276073", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:47", "updated_at": "2026-11-15 16:52:47" }, { "id": 279, "title": "Gender bias in natural language processing", "title_ar": "", "title_en": "Gender bias in natural language processing", "author": "Christine Basta", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Universitat Politècnica de Catalunya", "university_source": "repository_host", "college": "", "department": "", "country": "Spain", "governorate": "", "publication_year": 2022, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; deep-learning; nlp; responsible-ai; predictive-analytics; machine-translation; speech-recognition; fairness", "keywords": "", "abstract": "(English) Gender bias is a dangerous form of social bias impacting an essential group of people. The effect of gender bias is propagated to our data, causing the accuracy of the predictions in models to be different depending on gender. In the deep learning era, our models are highly impacted by the training data transferring the negative biases in the data to the models. Natural Language Processing models encounter this amplification of bias in the data. Our thesis is devoted to studying the issue of gender bias in NLP applications from different points of view. To understand and manage the effect of bias amplification, evaluation and mitigation approaches have to be explored. The scientific society has exerted significant efforts in these two directions to enable proposing solutions to the problem. Our thesis is devoted to these two main directions; proposing evaluation schemes, whether as datasets or mechanisms, besides suggesting mitigation techniques. For evaluation, we proposed techniques for evaluating bias in contextualized embeddings and multilingual translation models. Besides, we presented benchmarks for evaluating bias for speech translation and multilingual machine translation models. For mitigation direction, we proposed different approaches in machine translation models by adding contextual text, contextual embeddings, or relaxing the architecture’s constraints. Our evaluation studies concluded that gender bias is encoded strongly in contextual embeddings representing professions and stereotypical nouns. We also unveiled that algorithms amplify the bias and that the system’s architecture impacts the behavior. For the evaluation purposes, we contributed to creating several benchmarks for the evaluation purpose; we introduced a benchmark that evaluates gender bias in speech translation systems. This research suggests that the current state of speech translation systems does not enable us to evaluate gender bias accurately because of the low quality of speech translation systems. Additionally, we proposed a toolkit for building multilingual balanced datasets for training and evaluating NMT models. These datasets are balanced within the gender occupation-wise. We found out that high-resource languages usually tend to predict more precise male translations. Our mitigation studies in NMT suggest that the nature of datasets and languages needs to be considered to apply the right approach. Mitigating bias can rely on adding contextual information. However, in other cases, we need to rethink the model and relax some influencing conditions to the bias that do not affect the general performance but reduce the effect of bias amplification. (Español) El prejuicio de género es una forma peligrosa de sesgo social que afecta a un grupo esencial de personas. El efecto del prejuicio de género se propaga a nuestros datos, lo que hace quela precisión de las predicciones en los modelos sea diferente según el género. En la era del aprendizaje profundo, nuestros modelos se ven afectados por los datos de entrenamiento que transfieren los prejuicios de los datos a los modelos. Los modelos de procesamiento del lenguaje natural pueden además amplificar este sesgo en los datos. Para comprender el efecto de la amplificación del prejuicio de género, se deben explorar enfoques de evaluación y mitigación. La sociedad científica ha visto la importancía de estas dos direcciones para posibilitar la propuesta de soluciones al problema. Nuestra tesis está dedicada a estas dos direcciones principales; proponiendo esquemas de evaluación, ya sea como conjuntos de datos y mecanismos de evaluación, además de sugerir técnicas de mitigación. Para la evaluación, propusimos técnicas para evaluar el prejuicio en representaciones vectoriales contextualizadas y modelos de traducción multilingüe. Además, presentamos puntos de referencia para evaluar el prejuicio de la traducción de voz y los modelos de traducción automática multilingüe. Para la dirección de mitigación, propusimos diferentes enfoques en los modelos de traducción automática agregando texto contextual, incrustaciones contextuales o relajando las restricciones de la arquitectura. Nuestros estudios de evaluación concluyeron que el prejuicio de género está fuertemente codificado en representaciones vectoriales contextuales que representan profesiones y sustantivos estereotipados. También revelamos que los algoritmos amplifican el sesgo y que la arquitectura del sistema afecta el comportamiento. Para efectos de evaluación, contribuimos a la creación de varios datos de referencia para fines de evaluación; presentamos un conjunto de datos que evalúa el sesgo de género en los sistemas de traducción de voz. Esta investigación sugiere que el estado actual de los sistemas de traducción del habla no nos permite evaluar con precisión el sesgo de género debido a la baja calidad de los sistemas de traducción del habla.", "source_url": "https:\/\/openalex.org\/W4387558641", "pdf_url": "https:\/\/upcommons.upc.edu\/bitstream\/2117\/379361\/1\/TCRSB1de1.pdf", "doi": "10.5821\/dissertation-2117-379361", "handle": "", "repository_name": "UPCommons institutional repository (Universitat Politècnica de Catalunya)", "source_record_id": "W4387558641", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:48", "updated_at": "2026-11-15 16:52:48" }, { "id": 393, "title": "MAWT: Multi-Attention-Weight Transformers", "title_ar": "", "title_en": "MAWT: Multi-Attention-Weight Transformers", "author": "Enver Askin", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Canada", "governorate": "", "publication_year": 2023, "language": "en", "research_domain": "deep-learning", "research_subdomain": "transformers", "categories": "machine-learning; deep-learning; nlp; predictive-analytics; transformers; machine-translation", "keywords": "", "abstract": "Transformers are machine learning models designed to learn and predict sequential and structured data, which are crucial to tasks such as neural machine translation and semantic parsing.They have become state-of-the-art engines for both of these tasks and much research in natural language processing is devoted to increasing their performance by introducing modifications to their architectures.In light of this trend, this thesis introduces a new Transformer architecture called MAWT: Multi-Attention-Weight Transformers in an attempt to increase the accuracy and variety of the acceptable predictions of a Transformer.It attempts to achieve this by training multiple weights per each Transformer attention head, which then are used to test the accuracy of the engine.This creates a new architecture under which the system produces a candidate set of outputs (instead of a singly output), along with a method for selecting from the candidate set.My proposal rests on the assumption --motivated by statistical considerations --that having a candidate set increases the probability of finding an exact match within the set.Upon testing, I observed that my system outperforms the regular transformer on 5\/6 benchmark neural machine translation and semantic parsing datasets, where engine performance is measured by exact match accuracy.Exact match accuracy demands syntactic identity between the output and the target.In order to investigate how well my new architecture generalizes to measures of semantic equivalence that don't also demand syntactic identity, I also recorded the BLEU scores on these datasets.The BLEU score is a measure of performance based on n-grams rather than exact symbolic match (i.e., how many contiguous sequence of n-many strings from the predicted output match the desired output).The results I report on the BLEU scores are more iii mixed, raising important questions that I highlight about the role of syntax in measures of semantic equivalence.", "source_url": "https:\/\/openalex.org\/W4362577036", "pdf_url": "http:\/\/repository.library.carleton.ca\/downloads\/g732d990n", "doi": "10.22215\/etd\/2023-15442", "handle": "", "repository_name": "Carleton University's Institutional Repository (MacOdrum Library, Carleton University)", "source_record_id": "W4362577036", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:49", "updated_at": "2026-11-15 16:52:49" }, { "id": 395, "title": "Evaluation of the Transformer Model for Abstractive Text Summarization", "title_ar": "", "title_en": "Evaluation of the Transformer Model for Abstractive Text Summarization", "author": "Fredrik Jönsson", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "KTH Royal Institute of Technology", "university_source": "repository_host", "college": "", "department": "", "country": "Sweden", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "deep-learning", "research_subdomain": "transformers", "categories": "deep-learning; nlp; rnn; transformers; machine-translation", "keywords": "", "abstract": "Being able to generate summaries automatically could speed up the spread and retention of information and potentially increase productivity in several fields. Using RNN-based encoder-decoder models with attention have been successful on a variety of language-related tasks such as automatic summarization but also in the field of machine translation. Lately, the Transformer model has been shown to outperform RNN-based models with attention in the relatedfield of machine translation. This study compares the Transformer model to a LSTM-based encoderdecoder model with attention on the task of abstractive summarization. Evaluation is done both automatically, using ROUGE score, as well as using human evaluators to estimate the grammar and readability of the generated summaries. The results show that the Transformer model produces better summaries both in terms of ROUGE score and when evaluated with human evaluators.", "source_url": "https:\/\/openalex.org\/W2991849516", "pdf_url": "", "doi": "", "handle": "", "repository_name": "KTH Publication Database DiVA (KTH Royal Institute of Technology)", "source_record_id": "W2991849516", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:49", "updated_at": "2026-11-15 16:52:49" }, { "id": 431, "title": "Natural Language Processing in de menselijke hersenen", "title_ar": "", "title_en": "Natural Language Processing in de menselijke hersenen", "author": "Helena Balabin", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Belgium", "governorate": "", "publication_year": 2026, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "deep-learning; nlp; medical-ai; transformers; llm; sentiment-analysis; machine-translation; speech-recognition", "keywords": "", "abstract": "Natural Language Processing (NLP) is a subfield of artificial intelligence focused on processing and interpreting human language. With the recent success of NLP approaches across a diverse range of application areas, such as machine translation and sentiment analysis, the question arises as to how far the seemingly human-like performances indeed resemble the processing of language in the human brain. Neurological studies of connected speech are central to providing deeper insights into the possible connection between language representations in the human brain and language models used in NLP. The collection of connected speech data in multiple cohorts of older adults will provide the basis for this work. By employing state-of-the-art NLP methods such as the Transformer, numerical feature representations will be generated for sentences and multi-sentence units extracted from textual transcriptions of the collected speech data. These NLP features will be compared to functional brain connectivity patterns to study their link to human language processing. Another aim of this work is to investigate whether NLP features based on connected speech from older adults can be used to differentiate between healthy aging, preclinical, and the early dementia stage of Alzheimer's disease. As a result, this work will potentially provide a deeper understanding of the connection between the derived NLP features and the aging brain and the possible existence of human brain counterparts of given NLP representations.", "source_url": "https:\/\/openalex.org\/W7162916085", "pdf_url": "", "doi": "", "handle": "", "repository_name": "Lirias (KU Leuven)", "source_record_id": "W7162916085", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:49", "updated_at": "2026-11-15 16:52:49" }, { "id": 557, "title": "Natural language processing as autoregressive generation", "title_ar": "", "title_en": "Natural language processing as autoregressive generation", "author": "Xiang Lin", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Nanyang Technological University", "university_source": "repository_host", "college": "", "department": "", "country": "Singapore", "governorate": "", "publication_year": 2023, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "machine-learning; deep-learning; computer-vision; nlp; generative-ai; responsible-ai; predictive-analytics; llm; machine-translation; fairness", "keywords": "", "abstract": "The advances in deep learning have led to great achievements in many Natural Language Processing (NLP) tasks. With the nature of language, i.e., sequential data, most NLP tasks can be framed into the sequence learning framework, such as text generation. As one of the most important foundations for modern NLP techniques, autoregressive generation models have achieved dominant performance in a great deal of NLP tasks. Therefore, this thesis emphasizes improving the autoregressive generation model for different NLP tasks. While many tasks can naturally fit into the sequence learning framework, some of them, e.g., building discourse parsing tree, require sophisticated designs to fit into neural models. Therefore, this thesis firstly emphasizes a novel unified framework for discourse parsing, which builds a discourse tree in a top-down depth-first manner, and it frames the task as an autoregressive generation task with the goal of each step being the prediction of the node position given a piece of text. The proposed approach is proven effective with extensive empirical experiments. In addition, I extend the above framework by proposing a hierarchical decoder, which leverages the information from parents and siblings of the nodes that are currently processed. The proposed decoder utilizes the nature of the tree structure and further improves the experiment performance on both discourse parsing and dependency parsing tasks. On the other hand, the de facto strategies, i.e., cross entropy loss and teacher forcing, for training the autoregressive generation models have been shown problematic in certain aspects. For example, cross entropy loss, which is one of the widely leveraged training objective functions, often leads to text degeneration in text generation, and teacher forcing suffers from the exposure bias problem, where there exists a mismatch between the training and testing setup. For text degeneration, I introduce a class of diminishing attentions, which enforces the submodularity of the coverage calculated by cross attention in the sequence-to-sequence model. The proposed diminishing attentions achieve notable improvement on several neural text generation tasks, including text summarization, machine translation, and image paragraph generation. Further, I propose a novel training objective, ScaleGrad, to replace cross entropy, which significantly reduces the degeneration problem in different text generation tasks. In fact, ScaleGrad can be extended to problems beyond text degeneration. It provides wide flexibility to inject different inductive biases into the text generation model by directly modifying the gradient information in the output layer. Next, for the exposure bias problem, this thesis introduces a novel type of scheduled sampling based on training accuracy, which requires only minimal hyper-parameter tuning compared to existing scheduled sampling methods. Additionally, a novel imitation loss is proposed to further enforce the model’s generative behavior to match the teacher-forced behavior. Moreover, this thesis demonstrates that reducing exposure bias can improve the robustness of language models against repetition and toxic errors.", "source_url": "https:\/\/openalex.org\/W4382929605", "pdf_url": "", "doi": "10.32657\/10356\/168487", "handle": "", "repository_name": "DR-NTU (Nanyang Technological University)", "source_record_id": "W4382929605", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:50", "updated_at": "2026-11-15 16:52:50" }, { "id": 749, "title": "Adaptive Fusion Techniques for Effective Multimodal Deep Learning", "title_ar": "", "title_en": "Adaptive Fusion Techniques for Effective Multimodal Deep Learning", "author": "Gaurav Sahu", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "University of Waterloo", "university_source": "repository_host", "college": "", "department": "", "country": "Canada", "governorate": "", "publication_year": 2020, "language": "en", "research_domain": "deep-learning", "research_subdomain": "transformers", "categories": "deep-learning; computer-vision; nlp; generative-ai; transformers; multimodal; video-analysis; sentiment-analysis; machine-translation; speech-recognition; gan", "keywords": "", "abstract": "Effective fusion of data from multiple modalities, such as video, speech, and text, is a challenging task due to the heterogeneous nature of multimodal data. In this work, we propose fusion techniques that aim to model context from different modalities effectively. Instead of defining a deterministic fusion operation, such as concatenation, for the network, we let the network decide “how” to combine given multimodal features more effectively. We propose two networks: 1) Auto-Fusion network, which aims to compress information from different modalities while preserving the context, and 2) GAN-Fusion, which regularizes the learned latent space given context from complementing modalities. A quantitative evaluation on the tasks of multimodal machine translation and emotion recognition suggests that our adaptive networks can better model context from other modalities than all existing methods, many of which employ massive transformer-based networks.", "source_url": "https:\/\/openalex.org\/W3084985854", "pdf_url": "https:\/\/uwspace.uwaterloo.ca\/bitstreams\/d8ed4daf-ffb2-4594-8566-a7a3085f4b99\/download", "doi": "", "handle": "", "repository_name": "UWSpace (University of Waterloo)", "source_record_id": "W3084985854", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:52", "updated_at": "2026-11-15 16:52:52" }, { "id": 897, "title": "Transfer learning for language model adaptation", "title_ar": "", "title_en": "Transfer learning for language model adaptation", "author": "Bari M. Saiful", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Nanyang Technological University", "university_source": "repository_host", "college": "", "department": "", "country": "Singapore", "governorate": "", "publication_year": 2023, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "machine-learning; deep-learning; computer-vision; nlp; transfer-learning; self-semi-supervised; predictive-analytics; llm; sentiment-analysis; machine-translation", "keywords": "", "abstract": "Language is the pathway to democratize the boundary of land and culture. Bridging the gap between languages is one of the biggest challenges of Artificial Intelligent (AI) systems. The current success of AI systems is dominated by the supervised learning paradigm where gradient-based learning algorithms (i.e., SGD, Adam) are designed to optimize complex high-dimensional planes. These algorithms learn from statistical observations that are typically collected with the intention of a specific task (i.e., product review, sentiment analysis). The use of task-dependent samples makes the learning procedure tedious as it requires manually annotated data. On the contrary, without having a sufficient amount of samples to represent the distribution, deep learning models tend to suffer from the lack of robustness. Due to the natural puzzle of randomness, not all sets of observations are observed in the data collection procedure, thus creating Out-of-Distribution (OOD) problem in learning algorithms. In the search of a generic task agnostic distribution, a large collection of text over various domains can be considered as - Standard Natural Text Distribution (SNTD). The generic idea of Transfer Learning for Natural Language Processing (NLP) is to utilize the SNTD knowledge for any other task-dependent training. Learning SNTD, followed by a task-adaptation method with a smaller volume of annotated data has yielded state-of-the-art (SOTA) results in various supervised NLP tasks. However, annotated data for every task in every language is rare. In language models, there are many kinds of distributional variance. One of the most common ways that distributional variance is encoded into a language model is when it is trained with mono-lingual text and learned dis-jointly. Word embeddings produced from these language models are then used as pre-trained embedding vectors to adapt to a downstream task. We propose adversarial training to project two monolingual distributions into the same space and then improve the robustness of the model by augmented fine-tuning with parameter sharing. By projecting mono-lingual language distribution into the same cross-lingual space, the distribution of the language become aware of each other. These projected distribution are semantically aware of each other in the latent space. Thus when we train one distribution the other distribution automatically adapt to the training data, making it easier to transfer (exchange) knowledge. On top of that, the proposed novel self-training architecture improves cross-lingual transfer by a large margin. Then, we focus on the jointly trained multi-lingual language model, where there is no pre-dominant distributional variance. In the multi-lingual language model, we put more focus on downstream task adaptation. We found that semi-supervised learning with pseudo-augmented data from a pre-trained language model can greatly improve the performance of tasks further down the line. In the end, we introduce a new novel data augmentation framework that uses the neighboring (vicinal) samples of the original training data without explicitly using any parallel text corpora or machine translation system. Our proposed method performs simultaneous self-training with data augmentation and unsupervised sample selection. It also proposes curriculum strategies for different domain samples. With extensive experiments on three different cross-lingual tasks spanning many language pairs, we have demonstrated the effectiveness of our proposed method. While all of the work done above focused on improving task adaptation in multiple languages without supervision, we further investigate how adding a few samples affects multi-lingual task adaptation. To this end, we leverage a small number of support samples from each language and propose an inference-time transductive nearest neighbor-based approach which utilizes the entropy of the query samples for predictions. We show that our proposed method outperforms full model\/full head fine-tuning as well as cross-task fine-tuning. We also show impressive performance gain (~37x) in computational costs for the full inference prediction. However, as the language models grow bigger, it becomes increasingly difficult to perform efficient inferences, especially for multiple tasks. The jointly optimized multi-lingual distribution helps transfer knowledge from a high-resource language to a low-resource language. While working on transductive nearest neighbor inference, we observe that language models are significantly prone to task distributions. Unless we use an extremely large language model (>100B), a model used for a specific task adaptation cannot be used for another task. In this dissertation, our final proposed method addresses this issue by multitask prompted learning.", "source_url": "https:\/\/openalex.org\/W4386409617", "pdf_url": "https:\/\/dr.ntu.edu.sg\/bitstream\/10356\/169892\/2\/sbmaruf_thesis.pdf", "doi": "10.32657\/10356\/169892", "handle": "", "repository_name": "DR-NTU (Nanyang Technological University)", "source_record_id": "W4386409617", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:54", "updated_at": "2026-11-15 16:52:54" }, { "id": 1034, "title": "Deep Learning for Amharic Text-Image Recognition: Algorithm, Dataset and Application", "title_ar": "", "title_en": "Deep Learning for Amharic Text-Image Recognition: Algorithm, Dataset and Application", "author": "Birhanu Hailu Belay", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Kaiserslautern University of Technology", "university_source": "repository_host", "college": "", "department": "", "country": "Germany", "governorate": "", "publication_year": 2021, "language": "en", "research_domain": "deep-learning", "research_subdomain": "transformers", "categories": "machine-learning; deep-learning; computer-vision; nlp; ai-applications; cnn; rnn; transformers; image-classification; machine-translation; ai-industry", "keywords": "", "abstract": "Optical Character Recognition (OCR) is one of the central problems in pattern recognition. Its applications have played a great role in the digitization of document images collected from het- erogeneous sources. Many of the well-known scripts have OCR systems with sufficiently high performance that enables OCR applications in industrial\/commercial settings. However, OCR sys- tems yield very-good results only on a narrow domain and very-specific use cases. Thus, it is still a challenging task, and there are other exotic languages with indigenous scripts, such as Amharic, for which no well-developed OCR systems exist. As many as 100 million people speak Amharic, and it is an official working language of Ethiopia. Amharic script contains about 317 different alphabets derived from 34 consonants with small changes. The change involves shortening or elongating one of its main legs or adding small diacritics to the right, left, top, or bottom of the consonant character. Such modifications lead the characters to have similar shapes and make the recognition task complex, but this is particularly interesting for charac- ter recognition research. So far, Amharic script recognition models are developed based on classical machine learning techniques, and they are very limited in addressing the issues for Amharic OCR. The motivation of this thesis is, therefore, to explore and tailor contemporary deep learning tech- niques for the OCR of Amharic. This thesis addresses the challenges in Amharic OCR through two main contributions. The first contribution is an algorithmic contribution in which we investigate deep learning approaches that suit the demand for Amharic OCR. The second is a technical contribution that comprises several works towards the OCR model development; thus, it introduces a new Amharic database consisting of collections of images annotated at a character and text-line level. It also presents a novel CNN- based framework designed by leveraging the grapheme of characters in Fidel-Gebeta (where Fidel- Gebeta consists of the full set of Amharic characters in matrix structure) and achieves 94.97% overall character recognition accuracy. In addition to character level methods, text-line level methods are also investigated and devel- oped based on sequence-to-sequence learning. These models avoid several of the pre-processing stages used in prior works by eliminating the need to segment individual characters. In this design, we use a stack of CNNs, before the Bi-LSTM layers and train from end-to-end. This model out- performs the LSTM-CTC based network, on average, by a CER of 3.75% with the ADOCR test set. Motivated by the success of attention, in addressing the problems’ of long sequences in Neural Machine Translation (NMT), we proposed a novel attention-based methodology by blending the attention mechanism into CTC objective function. This model performs far better than the existing techniques with a CER of 1.04% and 0.93% on printed and synthetic text-line images respectively. Finally, this thesis provides details on various tasks that have been performed for the development of Amharic OCR. As per our empirical analysis, the majority of the errors are due to poor annotation of the dataset. As future work, the methods proposed in this thesis should be further investigated and extended to deal with handwritten and historical Amharic documents.", "source_url": "https:\/\/openalex.org\/W3163416829", "pdf_url": "https:\/\/kluedo.ub.uni-kl.de\/frontdoor\/index\/index\/docId\/6291", "doi": "10.26204\/kluedo\/6291", "handle": "", "repository_name": "Publication Server of Kaiserslautern University of Technology (Kaiserslautern University of Technology)", "source_record_id": "W3163416829", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:57", "updated_at": "2026-11-15 16:52:57" }, { "id": 1081, "title": "Automatic speech recognition and machine translation with deep neural networks for open educational resources, parliamentary contents and broadcast media", "title_ar": "", "title_en": "", "author": "Gonzalo Vicente Garcés Díaz-Munío", "supervisor": "", "degree_type": "phd", "degree_evidence": "نص السجل: «tesis doctoral»", "university": "Universitat Politècnica de València", "university_source": "repository_host", "college": "", "department": "", "country": "Spain", "governorate": "", "publication_year": 2024, "language": "es", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "deep-learning; computer-vision; nlp; ai-applications; machine-translation; speech-recognition; ai-education", "keywords": "", "abstract": "[ES] En la última década, el reconocimiento automático del habla (RAH) y la traducción automática (TA) han mejorado enormemente mediante el uso de modelos de redes neuronales profundas (RNP) en constante evolución. Si a principios de los 2010 los sistemas de RAH y TA previos a las RNP llegaron a afrontar con éxito algunas aplicaciones reales como la transcripción y traducción de vídeos docentes pregrabados, ahora en los 2020 son abordables aplicaciones que suponen un reto mucho mayor, como la subtitulación de retransmisiones audiovisuales en directo. En este mismo período, se están invirtiendo cada vez mayores esfuerzos en la accesibilidad a los medios audiovisuales para todos, incluidas las personas sordas. El RAH y la TA, en su estado actual, son grandes herramientas para aumentar la disponibilidad de medidas de accesibilidad como subtítulos, transcripciones y traducciones, y también para proporcionar acceso multilingüe a todo tipo de contenidos. En esta tesis doctoral presentamos resultados de investigación sobre RAH y TA basadas en RNP en tres campos muy activos: los recursos educativos abiertos, los contenidos parlamentarios y los medios audiovisuales. En el área de los recursos educativos abiertos (REA), presentamos primeramente trabajos sobre la evaluación y postedición de RAH y TA con métodos de interacción inteligente, en el marco del proyecto de investigación europeo \"transLectures: Transcripción y Traducción de Vídeos Docentes\". Los resultados obtenidos confirman que la interacción inteligente puede reducir aún más el esfuerzo de postedición de transcripciones y traducciones automáticas. Seguidamente, en el contexto del posterior proyecto europeo X5gon, presentamos una investigación sobre el desarrollo de sistemas de TA neuronal basados en RNP, y sobre sacar el máximo partido de corpus de TA masivos mediante filtrado automático de datos. Este trabajo dio como resultado sistemas de TA neuronal clasificados entre los mejores en una competición internacional de TA, y mostramos cómo estos nuevos sistemas mejoraron la calidad de los subtítulos multilingües en casos reales de REA. En el ámbito también en crecimiento de las tecnologías del lenguaje para contenidos parlamentarios, describimos una investigación sobre técnicas de filtrado de datos de habla para el RAH en tiempo real en el contexto de debates del Parlamento Europeo. Esta investigación permitió la publicación de Europarl-ASR, un nuevo y amplio corpus de habla para entrenamiento y evaluación de sistemas de RAH en continuo, así como para la evaluación comparativa de técnicas de filtrado de datos de habla. Finalmente, presentamos un trabajo en un ámbito en la vanguardia tecnológica del RAH y de la TA: la subtitulación de retransmisiones audiovisuales en directo, en el marco del Convenio de colaboración I+D+i 2020-2023 entre la radiotelevisión pública valenciana À Punt y la Universitat Politècnica de València para la subtitulación asistida por ordenador de contenidos audiovisuales en tiempo real. Esta investigación ha resultado en la implantación de sistemas de RAH en tiempo real, de alta precisión y baja latencia, para una lengua no mayoritaria en el mundo (el catalán) y una de las lenguas más habladas del mundo (el castellano) en un medio audiovisual real.", "source_url": "https:\/\/openalex.org\/W4404863865", "pdf_url": "", "doi": "10.4995\/thesis\/10251\/212454", "handle": "", "repository_name": "RiuNet (Universitat Politècnica de València)", "source_record_id": "W4404863865", "other_sources": "", "verification_status": "verified", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1114, "title": "Discourse-aware neural machine translation", "title_ar": "", "title_en": "Discourse-aware neural machine translation", "author": "Wang Longyue", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Dublin City University", "university_source": "repository_host", "college": "", "department": "", "country": "Ireland", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; nlp; predictive-analytics; machine-translation", "keywords": "", "abstract": "Machine translation (MT) models usually translate a text by considering isolated sentences based on a strict assumption that the sentences in a text are independent of one another. However, it is a truism that texts have properties of connectedness that go beyond those of their individual sentences. Disregarding dependencies across sentences will harm translation quality especially in terms of coherence, cohesion, and consistency. Previously, some discourse-aware approaches have been investigated for conventional statistical machine translation (SMT). However, this is a serious obstacle for the state-of-the-art neural machine translation (NMT), which recently has surpassed the performance of SMT. In this thesis, we try to incorporate useful discourse information for enhancing NMT models. More specifically, we conduct research on two main parts: 1) exploiting novel document-level NMT architecture; and 2) dealing with a specific discourse phenomenon for translation models. Firstly, we investigate the influence of historical contextual information on the perfor- mance of NMT models. A cross-sentence context-aware NMT model is proposed to consider the influence of previous sentences in the same document. Specifically, this history is summarized using an additional hierarchical encoder. The historical representations are then integrated into the standard NMT model in different strategies. Experimental results on a Chinese–English document-level translation task show that the approach significantly improves upon a strong attention-based NMT system by up to +2.1 BLEU points. In addition, analysis and comparison also give insightful discussions and conclusions for this research direction. Secondly, we explore the impact of discourse phenomena on the performance of MT. In this thesis, we focus on the phenomenon of pronoun-dropping (pro-drop), where, in pro-drop languages, pronouns can be omitted when it is possible to infer the referent from the context. As the data for training a dropped pronoun (DP) generator is scarce, we propose to automatically annotate DPs using alignment information from a large parallel corpus. We then introduce a hybrid approach: building a neural-based DP generator and integrating it into the SMT model. Experimental results on both Chinese–English and Japanese–English translation tasks demonstrate that our approach achieves a significant improvement of up to +1.58 BLEU points with 66% F-score for DP generation accuracy. Motivated by this promising result, we further exploit the DP translation approach for advanced NMT models. A novel reconstruction-based model is proposed to reconstruct the DP-annotated source sentence from the hidden states of either encoder or decoder, or both components. Experimental results on the same translation tasks show that the proposed approach significantly and consistently improves translation performance over a strong NMT baseline, which is trained on DP-annotated parallel data. To avoid the errors propagated from an external DP prediction model, we finally investigate an end-to-end DP translation model. Specifically, we improve the reconstruction-based model from three perspectives. We first employ a shared reconstructor to better exploit encoder and decoder representations. Secondly, we propose to jointly learn to translate and predict DPs. In order to capture discourse information for DP prediction, we finally combine the hierarchical encoder with the DP translation model. Experimental results on the same translation tasks show that our approach significantly improves both translation performance and DP prediction accuracy.", "source_url": "https:\/\/openalex.org\/W2951604644", "pdf_url": "https:\/\/doras.dcu.ie\/22903\/1\/Longyue_PhD_Thesis_Final.pdf", "doi": "", "handle": "", "repository_name": "Dublin City University Open Access Institutional Repository (Dublin City University)", "source_record_id": "W2951604644", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1115, "title": "Factored neural machine translation", "title_ar": "", "title_en": "", "author": "Mercedes García Martínez", "supervisor": "", "degree_type": "phd", "degree_evidence": "theses.fr: الفهرس الوطني الفرنسي لأطاريح الدكتوراه", "university": "", "university_source": "", "college": "", "department": "", "country": "France", "governorate": "", "publication_year": 2018, "language": "fr", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; machine-translation", "keywords": "", "abstract": "Traduction automatique neuronale factorisée La diversité des langues complexifie la tâche de communication entre les humains à travers les différentes cultures. La traduction automatique est un moyen rapide et peu coûteux pour simplifier la communication interculturelle. Récemment, laTraduction Automatique Neuronale (NMT) a atteint des résultats impressionnants. Cette thèse s'intéresse à la Traduction Automatique Neuronale Factorisé (FNMT) qui repose sur l'idée d'utiliser la morphologie et la décomposition grammaticale des mots (lemmes et facteurs linguistiques) dans la langue cible. Cette architecture aborde deux défis bien connus auxquelles les systèmes NMT font face. Premièrement, la limitation de la taille du vocabulaire cible, conséquence de la fonction softmax, qui nécessite un calcul coûteux à la couche de sortie du réseau neuronale, conduisant à un taux élevé de mots inconnus. Deuxièmement, le manque de données adéquates lorsque nous sommes confrontés à un domaine spécifique ou une langue morphologiquement riche. Avec l'architecture FNMT, toutes les inflexions des mots sont prises en compte et un vocabulaire plus grand est modélisé tout en gardant un coût de calcul similaire. De plus, de nouveaux mots non rencontrés dans les données d'entraînement peuvent être générés. Dans ce travail, j'ai développé différentes architectures FNMT en utilisant diverses dépendances entre les lemmes et les facteurs. En outre, j'ai amélioré la représentation de la langue source avec des facteurs. Le modèle FNMT est évalué sur différentes langues dont les plus riches morphologiquement. Les modèles à l'état de l'art, dont certains utilisant le Byte Pair Encoding (BPE) sont comparés avec le modèle FNMT en utilisant des données d'entraînement de petite et de grande taille. Nous avons constaté que les modèles utilisant les facteurs sont plus robustes aux conditions d'entraînement avec des faibles ressources. Le FNMT a été combiné avec des unités BPE permettant une amélioration par rapport au modèle FNMT entrainer avec des données volumineuses. Nous avons expérimenté avec dfférents domaines et nous avons montré des améliorations en utilisant les modèles FNMT. De plus, la justesse de la morphologie est mesurée à l'aide d'un ensemble de tests spéciaux montrant l'avantage de modéliser explicitement la morphologie de la cible. Notre travail montre les bienfaits de l'applicationde facteurs linguistiques dans le NMT.", "source_url": "https:\/\/openalex.org\/W7148632149", "pdf_url": "", "doi": "10.70675\/6b041d7cz8d47z4eb0zaefbz4170f95dfd99", "handle": "", "repository_name": "theses.fr (ABES)", "source_record_id": "W7148632149", "other_sources": "", "verification_status": "verified", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1116, "title": "Understanding, Explaining, and Aligning Neural Models for Machine Translation", "title_ar": "", "title_en": "Understanding, Explaining, and Aligning Neural Models for Machine Translation", "author": "Patrick Fernandes", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "United States", "governorate": "", "publication_year": 2025, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "machine-learning; deep-learning; nlp; predictive-analytics; llm; machine-translation", "keywords": "", "abstract": "The recent success of large neural network models trained on massive amounts of unlabeled data has, yet again, highlighted the power of scaling, in terms of both model capacity and data quantity. However, despite its successes, scaling up also adds new additional challenges: (i) the large size and complexity of these models make it difficult to understand their learning dynamics and global behavior, and how these change with the data distribution; (ii) their complexity also make it difficult to extract explanations of how they arrive at individual predictions, and to know how to leverage these explanations; (iii) this, coupled with the fact that the data used to train these models is often noisy, can lead to models that do not align with their intended use case. This thesis tackles all of the challenges above, with a particular focus on models trained for machine translation (MT). In the first part, we study the scaling behaviour of MT mod els trained to translate between multiple high-resource language pairs (LPs), showing that the way performance for individual LPs scales with model capacity is only dependent on the properties of that “task” and is invariant to cross-lingual interactions. We also investigate howmuchcontextinformationMTmodelsleverageindocument-levelMT,proposingways to increase it. In the second part, we start by proposing a novel method for extracting expla nations of how neural models arrive at their predictions, relying on the assumption that good explanations should help other models learn, and leveraging bilevel optimization\/meta learning to learn explainers that teach well. Then we showthat not onlyare reasoning models state-of-the-art systems for low-resource MT, but that their chain-of-thought rationales are good explanations to help teach smaller LLMs the same task. In the third and final part, we start by showing that, by including metrics of translation quality (trained on human feedback annotations) in the inference stage of a trained MTmodel, we can improve the qualityof the translation and align the system towards human-like translations. We then also show that we can almost completely remove the dependency on human annotations by leveraging large language models to provide (fine-grained) translation quality feedback.", "source_url": "https:\/\/openalex.org\/W7116361097", "pdf_url": "", "doi": "10.1184\/r1\/30752876", "handle": "", "repository_name": "KiltHub Repository", "source_record_id": "W7116361097", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1117, "title": "Machine translation of morphologically rich languages using deep neural networks", "title_ar": "", "title_en": "Machine translation of morphologically rich languages using deep neural networks", "author": "Peyman Passban", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Dublin City University", "university_source": "repository_host", "college": "", "department": "", "country": "Ireland", "governorate": "", "publication_year": 2017, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "deep-learning; computer-vision; nlp; image-segmentation; llm; machine-translation", "keywords": "", "abstract": "This thesis addresses some of the challenges of translating morphologically rich languages (MRLs). Words in MRLs have more complex structures than those in other languages, so that a word can be viewed as a hierarchical structure with several internal subunits. Accordingly, word-based models in which words are treated as atomic units are not suitable for this set of languages. As a commonly used and eff ective solution, morphological decomposition is applied to segment words into atomic and meaning-preserving units, but this raises other types of problems some of which we study here. We mainly use neural networks (NNs) to perform machine translation (MT) in our research and study their diff erent properties. However, our research is not limited to neural models alone as we also consider some of the difficulties of conventional MT methods. First we try to model morphologically complex words (MCWs) and provide better word-level representations. Words are symbolic concepts which are represented numerically in order to be used in NNs. Our first goal is to tackle this problem and find the best representation for MCWs. In the next step we focus on language modeling (LM) and work at the sentence level. We propose new morpheme-segmentation models by which we finetune existing LMs for MRLs. In this part of our research we try to find the most efficient neural language model for MRLs. After providing word- and sentence-level neural information in the first two steps, we try to use such information to enhance the translation quality in the statistical machine translation (SMT) pipeline using several diff erent models. Accordingly, the main goal in this part is to find methods by which deep neural networks (DNNs) can improve SMT. One of the main interests of the thesis is to study neural machine translation (NMT) engines from diff erent perspectives, and finetune them to work with MRLs. In the last step we target this problem and perform end-to-end sequence modeling via NN-based models. NMT engines have recently improved significantly and perform as well as state-of-the-art systems, but still have serious problems with morphologically complex constituents. This shortcoming of NMT is studied in two separate chapters in the thesis, where in one chapter we investigate the impact of diff erent non-linguistic morpheme-segmentation models on the NMT pipeline, and in the other one we benefit from a linguistically motivated morphological analyzer and propose a novel neural architecture particularly for translating from MRLs. Our overall goal for this part of the research is to find the most suitable neural architecture to translate MRLs. We evaluated our models on diff erent MRLs such as Czech, Farsi, German, Russian, and Turkish, and observed significant improvements. The main goal targeted in this research was to incorporate morphological information into MT and define architectures which are able to model the complex nature of MRLs. The results obtained from our experimental studies confirm that we were able to achieve our goal.", "source_url": "https:\/\/openalex.org\/W2795545650", "pdf_url": "https:\/\/doras.dcu.ie\/22200\/1\/Thesis-Hardbound-13212218-20180111-Peyman-Passban.pdf", "doi": "", "handle": "", "repository_name": "Dublin City University Open Access Institutional Repository (Dublin City University)", "source_record_id": "W2795545650", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1118, "title": "Multi-domain Neural Machine Translation", "title_ar": "", "title_en": "", "author": "Minh-Quang Pham", "supervisor": "", "degree_type": "phd", "degree_evidence": "theses.fr: الفهرس الوطني الفرنسي لأطاريح الدكتوراه", "university": "", "university_source": "", "college": "", "department": "", "country": "France", "governorate": "", "publication_year": 2021, "language": "fr", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; ai-applications; machine-translation; ai-industry", "keywords": "", "abstract": "Traduction automatique neuronale multidomaine Aujourd'hui, les systèmes de traduction automatique neuronale (NMT) constituent des systèmes de pointe en traduction automatique (TA). Cependant, ces modèles de traduction nécessitent des données d'entraînement relativement volumineuses et ont de la difficulté à traduire des textes de domaine spécifique. Un domaine peut être constitué de textes d'un sujet particulier ou de textes écrits dans un but particulier. Bien que les systèmes NMT puissent être adaptés pour une meilleure qualité de traduction dans un domaine cible étant donné un corpus de train représentatif, cette technique a des effets secondaires négatifs, notamment une fragilité contre des exemples hors domaine et un « oubli catastrophique » des domaines précédents représentés dans les données d'entraînement. De plus, un système de traduction doit faire face à de nombreux domaines possibles dans des applications réelles, ce qui rend impraticable le « un domaine un modèle ». Une solution prometteuse consiste à construire des systèmes NMT multi-domaines formés à partir des données de nombreux domaines et adaptés à plusieurs domaines cibles. Il y a deux motivations. Premièrement, les grands corpus de trains améliorent la généralisation du système NMT. Deuxièmement, les textes d'un domaine peuvent être utiles pour adapter un modèle NMT à un domaine similaire. La pénurie des données et l'effet de transfert positif hypothétique sont également deux raisons principales pour le développement des systèmes NMT multilingues. Maintenir plusieurs systèmes de traduction automatique bilingues nécessite de nombreuses ressources matérielles, car le nombre de paires de langues augmente de façon quadratique avec l'augmentation du nombre de langues. Les systèmes NMT multi-domaines et multilingues sont essentiels pour économiser des ressources pour l'industrie TA et améliorer la qualité du service TA. Cette thèse unifie d'abord l'adaptation de domaine et l'adaptation multi-domaine dans un cadre mathématique. De plus, nous passons en revue la littérature sur l'adaptation aux (multi-)domaines à travers une approche structurelle en montrant quatre cas principaux et en associant les méthodes proposées à chaque cas d'application. Deuxièmement, nous proposons une nouvelle évaluation multicritères des approches multi-domaines. Nous soulignons les exigences d'un système multi-domaines et réalisions une comparaison exhaustive d'un large ensemble de méthodes. Nous proposons également de nouvelles méthodes pour l'adaptation aux multi-domaines, y compris les plongements de mot parcimonieux, les couches parcimonieux et les adaptateurs résiduels, qui sont relativement légers et capables d'adapter un NMT modèle aux nombreux domaines. Pour équilibrer l'hétérogénéité des données d'entraînement, nous explorons et étudions les techniques à l'échantillonnage dynamique des données, qui adaptent de manière itérative la distribution de l'entraînement à une distribution de test prédéterminée. Enfin, nous nous intéressons aux approches de traduction avec des contextes augmentés, qui réutilisent des mémoires de traduction similaires pour améliorer la prédiction d'une phrase. Nous analysons et comparons plusieurs méthodes de cette ligne et démontrons qu'elles conviennent pour adapter notre système NMT à un domaine inconnu au détriment de coûts de calcul supplémentaires.", "source_url": "https:\/\/openalex.org\/W7150726110", "pdf_url": "https:\/\/theses.hal.science\/tel-03546910v1\/file\/105007_PHAM_2021_archivage.pdf", "doi": "10.70675\/17e2ca68z8662z4bb9zb875zac511563f5c6", "handle": "", "repository_name": "theses.fr (ABES)", "source_record_id": "W7150726110", "other_sources": "", "verification_status": "verified", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1119, "title": "Neural Language Models with Morphology for Machine Translation", "title_ar": "", "title_en": "Neural Language Models with Morphology for Machine Translation", "author": "Tomáš Musil", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Czech Republic", "governorate": "", "publication_year": 2017, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "deep-learning; nlp; llm; machine-translation", "keywords": "", "abstract": "Language models play an important role in many natural language processing tasks. In this thesis, we focus on language models built on artificial neural net- works. We examine the possibilities of using morphological annotations in these models. We propose a neural network architecture for a language model that explicitly makes use of morphological annotation of the input sentence: instead of word forms it processes lemmata and morphological tags. Both the baseline and the proposed method are evaluated on their own by perplexity, and also in the context of machine translation by the means of automatic translation quality evaluation. While in isolation the proposed model significantly outperforms the baseline, there is no apparent gain in machine translation. 1", "source_url": "https:\/\/openalex.org\/W2947679756", "pdf_url": "", "doi": "", "handle": "", "repository_name": "Digital Repository (National Repository of Grey Literature)", "source_record_id": "W2947679756", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1120, "title": "Multimodal neural machine translation", "title_ar": "", "title_en": "Multimodal neural machine translation", "author": "Malek Mgaidi", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Spain", "governorate": "", "publication_year": 2016, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; deep-learning; computer-vision; nlp; optimization; multimodal; video-analysis; machine-translation; speech-recognition", "keywords": "", "abstract": "Neural Machine Translation is a newly emerging approach to machine translation which attempts to build and train a large neural network that reads a sentence and outputs a notable translation. Nowadays the performance of such machines is in- creasingly in demand and Multilingual Neutral Machine Translation has emerged. There is an abundant bibliography on Multimodal Neutral Machine Translation as there is a consistent number of models which differ by the final aspects of trans- lation (adequacy, fidelity and fluency) and the multitude of inputs they can use (images, videos, text, speech or a combination of them). The GroundedTranslation was chosen in this work. As we know, by far, the state- of-the art provides some techniques such as using Long Short Term Memory and an encoder-decoder architecture for example to solve some training problems and they already have been implemented by Elliot Desmond for this retained solution. However, no investigation has been oriented toward the optimizer. This work aims to study multimodal neural machine translation architectures and its behavior un- der different optimization algorithms.", "source_url": "https:\/\/openalex.org\/W2520119691", "pdf_url": "", "doi": "", "handle": "", "repository_name": "RECERCAT (Consorci de Serveis Universitaris de Catalunya)", "source_record_id": "W2520119691", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1121, "title": "Neural Machine Translation for Arabic Language", "title_ar": "", "title_en": "Neural Machine Translation for Arabic Language", "author": "Manar Alkhatib", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; machine-translation; arabic-nlp", "keywords": "", "abstract": "", "source_url": "https:\/\/openalex.org\/W3109160103", "pdf_url": "", "doi": "", "handle": "", "repository_name": "", "source_record_id": "W3109160103", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1122, "title": "Domain adaptation for statistical machine translation and neural machine translation", "title_ar": "", "title_en": "Domain adaptation for statistical machine translation and neural machine translation", "author": "Jian Zhang", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Dublin City University", "university_source": "repository_host", "college": "", "department": "", "country": "Ireland", "governorate": "", "publication_year": 2017, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "machine-learning; deep-learning; nlp; transfer-learning; llm; machine-translation", "keywords": "", "abstract": "Both Statistical Machine Translation and Neural Machine Translation (NMT) are data-dependent learning approaches to Machine Translation (MT). The prerequisite is a large volume of training data in order to generate good statistical models. However, even if large volume of training corpora are available for MT, finding training data which are similar to the specific domains is still difficult. The MT models trained using the limited specific domain data cannot have sufficient coverage on the linguistic phenomena in that domain, which makes this a very challenging task. Because word meanings, genres or topics differ between domains, using the additional data from other domains can increase the dissimilar- ities between the training and testing data, and result in reduced translation quality. Such a challenge is defined as ‘domain adaptation’ challenge in the literature. In this thesis, we investigate domain adaptation in two different scenarios, namely a domain-awareness scenario and a domain-unawareness scenario. In a domain-awareness scenario, the domain information is given explicitly in the training data. We are interested in developing domain-adaptation techniques which transfer knowledge gained from the other domains to a desired domain. In the approach proposed here probabilistic values indicating the domain-likeness features for words are estimated by the context rather than by the words themselves. We then apply those features to the combined translation models in an MT system. We empirically show that translation quality can be significantly improved compared with previous related work. We then turn our interest to the recently proposed neural network training. We describe a domain-adaptation approach which can exploit large pre-trained word vector models. We evaluate our approach on both language modelling and machine translation tasks to demonstrate its efficiency, effectiveness and flexibility in a domain-awareness scenario. xiiIn a domain-unawareness scenario, the domain information is not given explicitly in the training data. The training data is heterogeneous, e.g. originated from tens or even hundreds of different resources without well-defined domain labels. We overcome such a challenge by deriving the topic information from the training corpora using well-estimated topic modelling algorithms. In this scenario, we pay particular attention to the most recent NMT framework. We are concerning with making a better lexical choice and improving the overall translation quality. Experimentally, we show that our model can perform better lexical choice, improve the overall translation quality and reduce the number of unknown words.", "source_url": "https:\/\/openalex.org\/W2767972135", "pdf_url": "https:\/\/doras.dcu.ie\/21949\/1\/thesis_hardbound_jian_zhang_56101775.pdf", "doi": "", "handle": "", "repository_name": "Dublin City University Open Access Institutional Repository (Dublin City University)", "source_record_id": "W2767972135", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1123, "title": "Robust Neural Machine Translation", "title_ar": "", "title_en": "Robust Neural Machine Translation", "author": "Abdul Rafae Khan", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "City University of New York", "university_source": "repository_host", "college": "", "department": "", "country": "United States", "governorate": "", "publication_year": 2020, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; nlp; transfer-learning; data-mining; machine-translation", "keywords": "", "abstract": "This thesis aims for general robust Neural Machine Translation (NMT) that is agnostic to the test domain. NMT has achieved high quality on benchmarks with closed datasets such as WMT and NIST but can fail when the translation input contains noise due to, for example, mismatched domains or spelling errors. The standard solution is to apply domain adaptation or data augmentation to build a domain-dependent system. However, in real life, the input noise varies in a wide range of domains and types, which is unknown in the training phase. This thesis introduces five general approaches to improve NMT accuracy and robustness, where three of them are invariant to models, test domains, and noise types. First, we describe a novel unsupervised text normalization framework Lex-Var, to reduce the lexical variations for NMT. Then, we apply the phonetic encoding as auxiliary linguistic information and obtained very significant (5 BLEU point) improvement in translation quality and robustness. Furthermore, we introduce the random clustering encoding method based on our hypothesis of Semantic Diversity by Phonetics and generalizes to all languages. We also discussed two domain adaptation models for the known test domain. Finally, we provide a measurement of translation robustness based on the consistency of translation accuracy among samples and use it to evaluate our other methods. All these approaches are verified with extensive experiments across different languages and achieved significant and consistent improvements in translation quality and robustness over the state-of-the-art NMT.", "source_url": "https:\/\/openalex.org\/W2997654874", "pdf_url": "", "doi": "", "handle": "", "repository_name": "CUNY Academic Works (City University of New York)", "source_record_id": "W2997654874", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1124, "title": "Syntactic and semantic features for statistical and neural machine translation", "title_ar": "", "title_en": "Syntactic and semantic features for statistical and neural machine translation", "author": "Maria Nădejde", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2018, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; machine-translation", "keywords": "", "abstract": "Machine Translation (MT) for language pairs with long distance dependencies and word reordering, such as German–English, is prone to producing output that is lexically or syntactically incoherent. Statistical MT (SMT) models used explicit or latent syntax to improve reordering, however failed at capturing other long distance dependencies. This thesis explores how explicit sentence-level syntactic information can improve translation for such complex linguistic phenomena. In particular, we work at the level of the syntactic-semantic interface with representations conveying the predicate-argument structures. These are essential to preserving semantics in translation and SMT systems have long struggled to model them. String-to-tree SMT systems use explicit target syntax to handle long-distance reordering, but make strong independence assumptions which lead to inconsistent lexical choices. To address this, we propose a Selectional Preferences feature which models the semantic affinities between target predicates and their argument fillers using the target dependency relations available in the decoder. We found that our feature is not effective in a string-to-tree system for German→English and that often the conditioning context is wrong because of mistranslated verbs. To improve verb translation, we proposed a Neural Verb Lexicon Model (NVLM) incorporating sentence-level syntactic context from the source which carries relevant semantic information for verb disambiguation. When used as an extra feature for re-ranking the output of a German→ English string-to-tree system, the NVLM improved verb translation precision by up to 2.7% and recall by up to 7.4%. While the NVLM improved some aspects of translation, other syntactic and lexical inconsistencies are not being addressed by a linear combination of independent models. In contrast to SMT, neural machine translation (NMT) avoids strong independence assumptions thus generating more fluent translations and capturing some long-distance dependencies. Still, incorporating additional linguistic information can improve translation quality. We proposed a method for tightly coupling target words and syntax in the NMT decoder. To represent syntax explicitly, we used CCG supertags, which encode subcategorization information, capturing long distance dependencies and attachments. Our method improved translation quality on several difficult linguistic constructs, including prepositional phrases which are the most frequent type of predicate arguments. These improvements over a strong baseline NMT system were consistent across two language pairs: 0.9 BLEU for German→English and 1.2 BLEU for Romanian→English.", "source_url": "https:\/\/openalex.org\/W2894807362", "pdf_url": "https:\/\/era.ed.ac.uk\/bitstream\/1842\/31346\/1\/N%c4%83dejde2018.pdf", "doi": "", "handle": "", "repository_name": "ERA", "source_record_id": "W2894807362", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1125, "title": "Document-wide Neural Machine Translation", "title_ar": "", "title_en": "Document-wide Neural Machine Translation", "author": "Sameen Maruf", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "deep-learning; nlp; machine-translation", "keywords": "", "abstract": "Machine translation (MT) is an important task in natural language processing as it automates the translation process and reduces the reliance on human translators. The goal of MT is to generate translations of a given text in a source language to that in the target language. With the advent of neural networks, the translation quality surpasses that of the translations obtained using statistical techniques. However, most of the neural translation models still perform a sentence-by-sentence translation, ignoring all extra-sentential information. This research aims to build efficient neural models for document-level translation, which incorporate global contextual information when translating sentences. Our experimental results confirm the significance of leveraging document-wide context information for improving translation quality.", "source_url": "https:\/\/openalex.org\/W3027945255", "pdf_url": "", "doi": "10.26180\/10295462", "handle": "", "repository_name": "Figshare", "source_record_id": "W3027945255", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1126, "title": "Hybrid Neural-Symbolic Machine Translation", "title_ar": "", "title_en": "Hybrid Neural-Symbolic Machine Translation", "author": "Jingyi Zhang", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Japan", "governorate": "", "publication_year": 2018, "language": "en", "research_domain": "nlp", "research_subdomain": "llm", "categories": "deep-learning; nlp; llm; machine-translation", "keywords": "", "abstract": "Machine translation (MT) investigates the use of software to translate a source sentence into the corresponding target sentence.The state-of-the-art MT approach is statistical MT, which learns statistical models from a parallel bilingual corpus to model the translation process.When this thesis started (2015), symbolic MT, which is based on symbolic translation rules, such as phrase-based MT [32], hierarchical phrase-based MT [8], and syntax-based MT [36,43], had the best translation performance.Because symbolic MT is based on context-free and ambiguous symbolic translation rules, a log-linear framework and statistical models learned from parallel or monolingual corpora, such as language models [32], reordering models [63,14], rule selection models [35,10,23], are exploited to select the most possible translation during symbolic MT decoding.Now (2018), neural MT (NMT) [4,55,68], which is based on a single large neural network and does not use any explicit translation rules, has outperformed symbolic MT on various translation tasks.Compared to symbolic MT, NMT uses distributed word representations and generally produces more fluent translations, but often sacrifices adequacy [31], such as NMT is more likely to generate completely unrelated translations, under and over translations.This thesis focuses on hybrid neural-symbolic MT methods, which aim to produce translations that are both fluent and adequate by combining the advantages of both symbolic and neural MT.The first part of our contribution is that we develop various neural models for the log-linear framework of symbolic MT, because neural models learn distributed representations for words and sentences, which can generalize better", "source_url": "https:\/\/openalex.org\/W3029827299", "pdf_url": "https:\/\/naist.repo.nii.ac.jp\/record\/10879\/files\/R014266.pdf", "doi": "10.34413\/dr.01508", "handle": "", "repository_name": "Institutional Repositories DataBase (IRDB)", "source_record_id": "W3029827299", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1127, "title": "Learning Hyperparameters for Neural Machine Translation", "title_ar": "", "title_en": "Learning Hyperparameters for Neural Machine Translation", "author": "Kenton Murray", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; deep-learning; nlp; responsible-ai; machine-translation; fairness", "keywords": "", "abstract": "Machine Translation, the subfield of Computer Science that focuses on translating between two human languages, has greatly benefited from neural networks. However, these neural machine translation systems have complicated architectures with many hyperparameters that need to be manually chosen. Frequently, these are selected either through a grid search over values, or by using values commonplace in the literature. However, these are not theoretically justified and the same values are not optimal for all language pairs and datasets. Fortunately, the innate structure of the problem allows for optimization of these hyperparameters during training. Traditionally, the hyperparameters of a system are chosen and then a learning algorithm optimizes all of the parameters within the model. In this work, I propose three methods to learn the optimal hyperparameters during the training of the model, allowing for one step instead of two. First, I propose using group regularizers to learn the number, and size of, the hidden neural network layers. Second, I demonstrate how to use a perceptron-like tuning method to solve known problems of undertranslation and label bias. Finally, I propose an Expectation-Maximization based method to learn the optimal vocabulary size and granularity. Using various techniques from machine learning and numerical optimization, this dissertation covers how to learn hyperparameters of a Neural Machine Translation system while training the model itself.", "source_url": "https:\/\/openalex.org\/W3042494786", "pdf_url": "", "doi": "10.7274\/mg74qj7616h", "handle": "", "repository_name": "Figshare", "source_record_id": "W3042494786", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1128, "title": "Decoding neural machine translation using gradient descent", "title_ar": "", "title_en": "Decoding neural machine translation using gradient descent", "author": "Emanuel Snelleman", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2016, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "deep-learning; nlp; machine-translation", "keywords": "", "abstract": "Neural machine translation is a resent approach to machine translation. Neural machine translation uses an artificial neural network to learn and perform translations. Usually an encoder-decoder architecture is used for the neural network. The encoder encodes the input sentence into a fix sized vector meant to represent the meaning of the sentence, the decoder then decodes that vector into a sentence of the same meaning in the other language. The decoder will generate the translation sequentially, and will output a probability distribution over the known words at each step of the sequence. From these probability distributions words needs to be chosen to form a sentence. The word chosen given any specific probability distribution will affect how the next probability distribution is generated. It is therefore important to chose the words well in order to get a translation as good as possible. In this thesis a way of choosing the words using gradient descent is tested. Decoding using gradient descent is compared to some other methods of decoding but gives no clear indication of working better than the other decoding methods. However, some results indicate gradient descent decoding might have some potential if further developed.", "source_url": "https:\/\/openalex.org\/W2519643131", "pdf_url": "", "doi": "", "handle": "", "repository_name": "", "source_record_id": "W2519643131", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1129, "title": "Preventing Over-Translation in Simultaneous Neural Machine Translation", "title_ar": "", "title_en": "Preventing Over-Translation in Simultaneous Neural Machine Translation", "author": "保昌 加納", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Japan", "governorate": "", "publication_year": 2024, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "deep-learning; nlp; machine-translation", "keywords": "", "abstract": "These days it is becoming common to communicate with international people as globalization expands.Although many international conferences are held, not all the speakers and listeners are good at English.Therefore, interpreters sometimes attend such conferences.For the interpretation in conferences and negotiations, we often require simultaneous interpretation which starts to translate before the speaker finishes the utterance.Simultaneous interpretation is important because the meeting time becomes much longer if we wait for the entire utterance before translation.Also, we sometimes tell jokes in negotiation, and it is funnier if the translation of a joke is done immediately without any interval.We would like to hire simultaneous interpreters for every international event, but it is highly costly.Therefore, there is a huge demand for simultaneous machine translation which can be used at a lower cost.Recent studies on machine translation are based on neural networks and they produce more fluent output than traditional statistical machine translation.Although machine translation research has evolved rapidly, simultaneous translation is still a difficult task.This is because the machine needs to output partial translation without reading the latter part of the input sentence as context.A lot of previous research on simultaneous translation uses the translation model pre-trained with bilingual full-sentence pairs.However, at the inference step, the model needs to translate only a partial input much shorter than a full sentence.Therefore, there is a gap between the training of translation models and", "source_url": "https:\/\/openalex.org\/W7144228045", "pdf_url": "https:\/\/naist.repo.nii.ac.jp\/record\/2000249\/files\/R018652.pdf", "doi": "10.34413\/dr.02042", "handle": "", "repository_name": "Institutional Repositories DataBase (IRDB)", "source_record_id": "W7144228045", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1130, "title": "Neural machine translation for multimodal interaction", "title_ar": "", "title_en": "Neural machine translation for multimodal interaction", "author": "Koel Dutta Chowdhury", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Dublin City University", "university_source": "repository_host", "college": "", "department": "", "country": "Ireland", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; deep-learning; computer-vision; nlp; cnn; multimodal; image-classification; machine-translation", "keywords": "", "abstract": "Typically it is seen that multimodal neural machine translation (MNMT) systems trained on a combination of visual and textual inputs produce better translations than systems trained using only textual inputs. The task of such systems can be decomposed into two sub-tasks: learning visually grounded representations from images and translation of the textual counterparts using those representations. In a multi-task learning framework, translations are generated from an attention-based encoder-decoder framework and grounded representations that are learned from pretrained convolutional neural networks (CNNs) for classifying images. In this thesis, I study different computational techniques to translate the meaning of sentences from one language into another considering the visual modality as a naturally occurring meaning representation bridging between languages. We examine the behaviour of state-of-the-art MNMT systems from the data perspective in order to understand the role of the both textual and visual inputs in such systems. We evaluate our models on the Multi30k, a large-scale multilingual multimodal dataset publicly available for machine learning research. Our results in the optimal and sparse data settings show that the differences in translation system performance are proportional to the amount of both visual and linguistic information whereas, in the adversarial condition the effect of the visual modality is rather small or negligible. The chapters of the thesis follow a progression starting with using different state-of-the-art MMT models for incorporating images in optimal data settings to creating synthetic image data under the low-resource scenario and extending to addition of adversarial perturbations to the textual input for evaluating the real contribution of images.", "source_url": "https:\/\/openalex.org\/W3004128329", "pdf_url": "https:\/\/doras.dcu.ie\/23747\/1\/Final_version_MSc_Thesis_2019.pdf", "doi": "", "handle": "", "repository_name": "Dublin City University Open Access Institutional Repository (Dublin City University)", "source_record_id": "W3004128329", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1131, "title": "Online Learning in Neural Machine Translation", "title_ar": "", "title_en": "Online Learning in Neural Machine Translation", "author": "Luis Cebrián Chuliá", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Spain", "governorate": "", "publication_year": 2017, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "computer-vision; nlp; machine-translation", "keywords": "", "abstract": "[EN] High quality translations are in high demand these days. Although machine translation offers acceptable performance, it is not sufficient in some cases and human supervision is required. In order to ease the translation task of the human, machine translation systems take part in this process. When a sentence in the source language needs to be translated, it is fed to the system which outputs a hypothesis translation. The human then, corrects this hypothesis (also known as post-editing) in order to obtain a high quality translation. Being able to transfer the knowledge that a human translator exhibit when post-editing a translation to the machine translation system is a desirable feature, as it has been proven that a more accurate machine translation system helps to increase the efficiency of the post-editing process. Because the post-editing scenario requires an already trained system, online learning techniques are suited for this task. In this work, three online learning algorithms have been proposed and applied to a neural machine translation sys- tem in a post-editing scenario. They rely on the Passive-Aggressive online learn- ing approach in which the model is updated after every sample in order to fulfil a correctness criterion while remembering previously learned information. The goal is to adapt and refine an already trained system with new samples on-the- fly as the post-editing process takes place (hence, the update time must be kept under control). Moreover, these new algorithms are compared with well-stablished online learning variants of the stochastic gradient descent algorithm. Results show im- provements on the translation quality of the system after applying these algo- rithms, reducing human effort in the post-editing process.", "source_url": "https:\/\/openalex.org\/W7139603174", "pdf_url": "", "doi": "", "handle": "", "repository_name": "Hispana", "source_record_id": "W7139603174", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1132, "title": "Human-assisted Neural Machine Translation: Harnessing Human Feedback for Machine Translation", "title_ar": "", "title_en": "Human-assisted Neural Machine Translation: Harnessing Human Feedback for Machine Translation", "author": "Paige Finkelstein", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "University of Washington", "university_source": "repository_host", "college": "", "department": "", "country": "United States", "governorate": "", "publication_year": 2020, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; nlp; predictive-analytics; machine-translation", "keywords": "", "abstract": "Neural machine translation (NMT) is a promising approach to the task of machine translation that has led to state-of-the-art results in many settings. However, NMT translations are still far from sufficient for many practical purposes. For this reason, there is a robust body of ongoing research on how to improve NMT systems with human feedback. This feedback can take many forms, including interactive-predictive NMT, post-editing of NMT output, and soliciting ratings or corrections of translations for the purposes of online learning. While these approaches are often effective, they are also often very time consuming and expensive. For that reason, there is important research into the question of how best to ensure that any human effort is used optimally. In this thesis, we contribute to this line of work by proposing a system that learns when it should ask for human feedback on a translation. This system makes use of an existing pre-trained NMT model, and introduces an additional feedback-requester model that learns to selectively solicit feedback from a human translator on the NMT translations. This system reduces human effort by directing attention to the most problematic sentences in a document, and the feedback-requester model itself is updated according to the translator's feedback. We also experiment with two active learning (AL) strategies for the feedback-requester model, and present a range of experiments simulating human translator use of the system and show the results over time.", "source_url": "https:\/\/openalex.org\/W3172116156", "pdf_url": "https:\/\/digital.lib.washington.edu\/bitstreams\/af687ae6-4e54-43c8-b6df-b15737b75835\/download", "doi": "", "handle": "", "repository_name": "ResearchWorks at the University of Washington (University of Washington)", "source_record_id": "W3172116156", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1133, "title": "Indonesian-English Neural Machine Translation", "title_ar": "", "title_en": "Indonesian-English Neural Machine Translation", "author": "Meisyarah Dwiastuti", "supervisor": "", "degree_type": "phd", "degree_evidence": "نص السجل: «Ph.D»", "university": "", "university_source": "", "college": "", "department": "", "country": "Czech Republic", "governorate": "", "publication_year": 2019, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; nlp; transfer-learning; machine-translation", "keywords": "", "abstract": "Title: Indonesian-English Neural Machine Translation Author: Meisyarah Dwiastuti Department: Institute of Formal and Applied Linguistics Supervisor: Mgr. Martin Popel, Ph.D., Institute of Formal and Applied Linguis- tics Abstract: In this thesis, we conduct a study on neural machine translation (NMT) for an under-studied language, Indonesian, specifically for English-Indonesian (EN-ID) and Indonesian-English (ID-EN) in a low-resource domain, TED talks. Our goal is to implement domain adaptation methods to improve the low-resource EN-ID and ID-EN NMT systems. First, we implement model fine-tuning method for EN-ID and ID-EN NMT systems by leveraging a large parallel corpus contain- ing movie subtitles. Our analysis shows the benefit of this method for the improve- ment of both systems. Second, we improve our ID-EN NMT system by leveraging English monolingual corpora through back-translation. Our back-translation ex- periments focus on how to incorporate the back-translated monolingual corpora to the training set, in which we investigate various existing training regimes and introduce a novel 4-way-concat training regime. We also analyze the effect of fine- tuning our back-translation models with different scenarios. Experimental results show that our method of implementing back-translation followed by model...", "source_url": "https:\/\/openalex.org\/W3007057041", "pdf_url": "", "doi": "10.13140\/rg.2.2.17954.45768", "handle": "", "repository_name": "Digital Repository (National Repository of Grey Literature)", "source_record_id": "W3007057041", "other_sources": "", "verification_status": "verified", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1134, "title": "Neural machine translation with limited resources", "title_ar": "", "title_en": "Neural machine translation with limited resources", "author": "Tasnim Mohiuddin", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "Nanyang Technological University", "university_source": "repository_host", "college": "", "department": "", "country": "Singapore", "governorate": "", "publication_year": 2022, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; deep-learning; computer-vision; nlp; transfer-learning; predictive-analytics; machine-translation", "keywords": "", "abstract": "With the advent of deep neural networks in recent years, Neural Machine Translation (NMT) systems have achieved state-of-the-art performance on standard translation benchmarks. NMT is a way to translate from one language to another with a single neural network in an end-to-end manner. The NMT models have emerged quickly, and within a few years of research, they have outperformed the traditional statistical systems with impressive performance. Despite the success of NMT models in standard benchmarks, there are some notable limitations. One of them is that NMT models are known to be data-hungry, i.e., they tend to work very well only when a massive amount of parallel training data (a.k.a. bitext) is available, but perform poorly when the data is limited. Except for some mainstream languages, e.g., English, French, or Chinese, most natural languages are low-resourced and lack large parallel data. Moreover, acquiring large bitext corpora is not viable in most scenarios, especially with resource-constrained conditions like low-resource languages. Researchers have made numerous endeavors to expand the success of NMT from high-resource to low-resource languages like transfer learning, data augmentation, and pivoting. However, they still require strong cross-lingual signals, i.e., lots of parallel data. One solution to this problem might be transferring cross-lingual signals through cross-lingual word embeddings (CLWEs), which can be learned from monolingual data in an unsupervised way or with the help of a small seed dictionary. CLWEs seem to be very promising in resource-constrained machine translation (MT). Most of the successful and predominant CLWE methods (a.k.a. word translation methods) learn a linear mapping function based on the isomorphic assumption, which is problematic. We hypothesize to learn the cross-lingual mapping in a projected latent space which would give the model enough flexibility to induce the required geometric structures such that it would be easier to align the embeddings. Based on this hypothesis, we propose two novel models for learning CLWEs. We empirically show that our methods are particularly very effective for low-resource languages. We then turn our attention from word- to sentence-level translation with limited resources. Specifically, we focus on data augmentation strategies widely used in NLP and Computer Vision to increase the robustness of the models in resource-constrained scenarios. We investigate the domain-mismatch issue thoroughly that hinders the all-embracing success of the existing techniques in NMT. Eventually, we introduce a novel data augmentation framework for low-resource NMT that leverages the neighboring samples of the original parallel data without explicitly using additional monolingual data. Our framework can diversify the in-domain parallel data in a controlled way. We perform extensive experiments on four low-resource language pairs comprising data from different domains. We have shown that our method is comparable to the traditional back-translation that uses extra in-domain monolingual data. Typically, NMT systems are trained on heterogeneous data from different domains, sources, topics, styles, and modalities. The quality of the data also varies a lot. Usually, during training, all the data are concatenated and randomly shuffled. However, not all of them may be useful, some data may be redundant, and some might even be noisy and detrimental to the final NMT system performance. These problems are more acute in low-resource languages compared to the high-resource ones. Consequently, we explore the possibilities of curriculum training for NMT systems, i.e., presenting the data to the NMT systems in a systematic order during training. We introduce a two-stage curriculum training framework for NMT where we fine-tune a base NMT model on subsets of data. To select the data subsets, we propose two scoring approaches --- deterministic scoring using pre-trained methods and online scoring that considers prediction scores of the emerging NMT model. Our curriculum strategies consistently demonstrate better translation quality and faster convergence (approximately 50% fewer updates) on both high- and low-resource languages.", "source_url": "https:\/\/openalex.org\/W4281623321", "pdf_url": "https:\/\/dr.ntu.edu.sg\/bitstream\/10356\/157475\/2\/Thesis-Tasnim.pdf", "doi": "10.32657\/10356\/157475", "handle": "", "repository_name": "DR-NTU (Nanyang Technological University)", "source_record_id": "W4281623321", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1135, "title": "Approximating neural machine translation for efficiency", "title_ar": "", "title_en": "Approximating neural machine translation for efficiency", "author": "Alham Fikri Aji", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "", "governorate": "", "publication_year": 2020, "language": "en", "research_domain": "deep-learning", "research_subdomain": "transformers", "categories": "machine-learning; deep-learning; nlp; transfer-learning; transformers; machine-translation", "keywords": "", "abstract": "Neural machine translation (NMT) has been shown to outperform statistical machine translation. However, NMT models typically require a large number of parameters and are expensive to train and deploy. Moreover, its large model size makes parallel training inefficient due to costly network communication. Likewise, distributing and locally running the model for a client-based NMT model such as a web browser or mobile device remains challenging. This thesis investigates ways to approximately train an NMT system by compressing either the gradients or the parameters for faster communication or reduced memory consumption. We propose a gradient compression technique that exchanges only the top 1% of the most significant gradient values while delaying the rest to be considered for the next iteration. This method reduces the network communication cost by 50-fold but causes noisy gradient updates. We also find that Transformer–the current state-of-the-art NMT architecture–is highly sensitive to noisy gradients. Therefore, we extend the compression technique by restoring the compressed gradient with locally-computed gradients. We obtained a linear scale-up in parallel training without sacrificing model performance. We also explore transfer learning as a better method of initialising the training. With transfer learning, the model converges faster and can be trained with more aggressive hyperparameters. Lastly, we propose a log-based quantisation method to compress the model size. Models are quantised to 4-bit precision with no noticeable quality degradation after re-training combined with reserving the quantisation errors as feedback.", "source_url": "https:\/\/openalex.org\/W3108682037", "pdf_url": "", "doi": "10.7488\/era\/533", "handle": "", "repository_name": "ERA", "source_record_id": "W3108682037", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1136, "title": "Non-Autoregressive Neural Machine Translation", "title_ar": "", "title_en": "Non-Autoregressive Neural Machine Translation", "author": "Jindřich Helcl", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Czech Republic", "governorate": "", "publication_year": 2022, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "deep-learning; nlp; machine-translation", "keywords": "", "abstract": "In recent years, a number of mehtods for improving the decoding speed of neural machine translation systems have emerged. One of the approaches that pro- poses fundamental changes to the model architecture are non-autoregressive models. In standard autoregressive models, the output token distributions are conditioned on the previously decoded outputs. The conditional dependence al- lows the model to keep track of the state of the decoding process, which improves the fluency of the output. On the other hand, it requires the neural network computation to be run sequentially, and thus it cannot be parallelized. Non- autoregressive models impose conditional independence on the output distri- butions, which means that the decoding process is parallelizable and hence the decoding speed improves. A major drawback of this approach is lower trans- lation quality compared to the autoregressive models. The goal of the non- autoregressive translation research is to find methods that improve the trans- lation quality, while retaining high decoding speed. In this thesis, we explore the research progress so far and identify flaws in the generally accepted eval- uation methodology. We experiement with non-autoregressive models trained with connectionist temporal classification. We find that even though our models...", "source_url": "https:\/\/openalex.org\/W7135983976", "pdf_url": "", "doi": "", "handle": "", "repository_name": "Digital Repository (National Repository of Grey Literature)", "source_record_id": "W7135983976", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1137, "title": "Efficient neural machine translation", "title_ar": "", "title_en": "Efficient neural machine translation", "author": "Jiatao Gu", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "University of Hong Kong", "university_source": "repository_host", "college": "", "department": "", "country": "Hong Kong", "governorate": "", "publication_year": 2018, "language": "en", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; machine-translation", "keywords": "", "abstract": "", "source_url": "https:\/\/openalex.org\/W2902773018", "pdf_url": "", "doi": "10.5353\/th_991044058182403414", "handle": "", "repository_name": "The HKU Scholars Hub (University of Hong Kong)", "source_record_id": "W2902773018", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:52:58", "updated_at": "2026-11-15 16:52:58" }, { "id": 1357, "title": "نموذجا HeyGen دبلجة الفيديوهات بواسطة الذكاء الاصطناعي - منصة", "title_ar": "نموذجا HeyGen دبلجة الفيديوهات بواسطة الذكاء الاصطناعي - منصة", "title_en": "", "author": "ياسين منير بلعسكري", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Algeria", "governorate": "", "publication_year": 2025, "language": "ar", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "nlp; machine-translation", "keywords": "", "abstract": "This study aims to highlight the role of machine translation and film dubbing using artificial intelligence technologies. It does so by analyzing machine translation as a tool to enhance the understanding of texts and translate them accurately and quickly, and by reviewing how artificial intelligence is used to provide more interactive and realistic viewing experiences. The study offers a comprehensive overview of how technology enhances translation and content dubbing, contributing to the expansion of cultural and artistic access.", "source_url": "https:\/\/openalex.org\/W7135255797", "pdf_url": "", "doi": "", "handle": "", "repository_name": "Dépôt Institutionnel de lUniversité de Tlemcen", "source_record_id": "W7135255797", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:53:01", "updated_at": "2026-11-15 16:53:01" }, { "id": 1373, "title": "إسهامات الذكاء الإصطناعي في الترجمة الآلية", "title_ar": "إسهامات الذكاء الإصطناعي في الترجمة الآلية", "title_en": "", "author": "محمد جزول", "supervisor": "", "degree_type": "unspecified", "degree_evidence": "", "university": "", "university_source": "", "college": "", "department": "", "country": "Algeria", "governorate": "", "publication_year": 2019, "language": "ar", "research_domain": "nlp", "research_subdomain": "machine-translation", "categories": "machine-learning; nlp; machine-translation", "keywords": "", "abstract": "ينبع هذا العمل من الرغبة في استكشاف مدى التقدم الذي حصل مؤخرا في مجال تقنيات التعلم الآلي في الترجمة الآلية. والغرض من هذه الدراسة هو اظهار التحسينات المستمرة في الذكاء الاصطناعي والذي اظهر نتائج مذهلة في مجال الترجمة", "source_url": "https:\/\/openalex.org\/W7135300530", "pdf_url": "https:\/\/dspace.univ-tlemcen.dz\/bitstreams\/c239a08b-cabb-4cac-a8de-ad6e4051aa68\/download", "doi": "", "handle": "", "repository_name": "Dépôt Institutionnel de lUniversité de Tlemcen", "source_record_id": "W7135300530", "other_sources": "", "verification_status": "needs_review", "created_at": "2026-11-15 16:53:01", "updated_at": "2026-11-15 16:53:01" } ] }