Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 20 of 20 for “"Sequence-to-Sequence Models"”.

  1. Towards Effectively Testing Sequence-to-Sequence models from White-Box Perspectives

    … progress, NLP systems, including those based on sequence-to-sequence models, confront various challenges. To tackle these, metamorphic testing methods have been employed across different NLP tasks. These methods entail task-specific adjustments at the token or sentence level. For example, in …

    waterloo Repository record for Towards Effectively Testing Sequence-to-Sequence models from White-Box Perspectives (opens in a new tab)

  2. Data Augmentation with Seq2Seq Models

    … corpus for the visual question answering system to make it more robust to paraphrases. This corpus is generated by concatenating two sequence to sequence models. In order to generate diverse paraphrases, we sample the neural network using diverse beam search. We evaluate the results on the …

    vt Repository record for Data Augmentation with Seq2Seq Models (opens in a new tab)

  3. GNET-QG: graph network for multi-hop question generation

    … answers. While recent advancements in sequence-to-sequence models have proven successful in generating natural language questions, there is an increasing need for models capable of handling more intricate contexts to produce detailed questions. Multi-hop question generation, which is …

    lethbridge Repository record for GNET-QG: graph network for multi-hop question generation (opens in a new tab)

  4. Transformer-based multi-hop question generation

    … context and, optionally, an answer, the goal is to generate a relevant and fluent natural language question. Although recent works on question generation have experienced success by utilizing sequence-to-sequence models, there is a need for question generation models to handle increasingly …

    lethbridge Repository record for Transformer-based multi-hop question generation (opens in a new tab)

  5. Multi task learning and incorporating common sense knowledge for question answering

    Question Answering (QA) system is an automated approach to retrieve correct responses to the questions asked by human in natural language. Reading comprehension (RC)in contrast to information retrieval, requires integrating information and reasoning about events, entities, and their relations …

    uiuc Repository record for Multi task learning and incorporating common sense knowledge for question answering (opens in a new tab)

  6. Neural Network Approaches to Implicit Discourse Relation Recognition

    In order to understand a coherent text, humans infer semantic or logical relations between textual units. For example, in ``I am hungry. I did not have lunch today.'' the reader infers a ``causality'' relation even if it is not explicitly stated via a term such as ``because''. The linguistic device …

    concordia Download PDF for Neural Network Approaches to Implicit Discourse Relation Recognition (opens in a new tab)

  7. Automated Methods for Comparison and Generation of Privacy Policies

    … of personal data and requires privacy policies to be more transparent for readers. There is a limited understanding of how GDPR has impacted the content of privacy policies. This study presents a comprehensive framework for evaluation of compliance of privacy policies with GDPR recommendations …

    guelph Repository record for Automated Methods for Comparison and Generation of Privacy Policies (opens in a new tab)

  8. Formality Style Transfer Within and Across Languages with Limited Supervision

    … When editing an article, professional editors take into account the target audience to select appropriate word choice and grammar. Similarly, professional translators translate documents for a specific audience and often ask what is the expected tone of the content when taking a …

    maryland Repository record for Formality Style Transfer Within and Across Languages with Limited Supervision (opens in a new tab)

  9. Comparison between rule-based and data-driven natural language processing algorithms for Brazilian Portuguese speech synthesis

    Due to the exponential growth in the use of computers, personal digital assistants and smartphones, the development of Text-to-Speech (TTS) systems have become highly demanded during the last years. An important part of these systems is the Text Analysis block, that converts the input text into …

    brazil-uerj Repository record for Comparison between rule-based and data-driven natural language processing algorithms for Brazilian Portuguese speech synthesis (opens in a new tab)

  10. An Empirical Evaluation Of Attention And Pointer Networks For Paraphrase Generation

    In computer vision, one of the common practice to augment the image dataset is by creating new images using geometric transformation, which preserves the similarity. This data augmentation was one of the most significant factors to win the Image Net competition in 2012 with vast neural networks. …

    concordia Download PDF for An Empirical Evaluation Of Attention And Pointer Networks For Paraphrase Generation (opens in a new tab)

  11. Learning deep models via optimal transport distance

    … community. Since most tasks are requiring deep models to estimate the true data distribution. For instance, GAN~\cite{goodfellow2014generative} wants to generate realistic images.In this PhD dissertation, I will discuss how to improve the distribution matching in deep learning models via optimal …

    duke Repository record for Learning deep models via optimal transport distance (opens in a new tab)

  12. Revealing the Positive Meaning of a Negation

    … and (c) psychologically more complex and harder to process. Negation often conveys positive meaning. This meaning ranges from implicatures to entailments. In this dissertation, I develop a system to reveal the underlying positive interpretation of negation. I first identify which words are …

    unt Repository record for Revealing the Positive Meaning of a Negation (opens in a new tab)

  13. WasmWalker: Path-based Code Representations for Improved WebAssembly Program Analysis

    … code in web browsers. Wasm has proven to be useful in applications including gaming, audio and video processing, and cloud computing, providing a high-performance, low-overhead alternative to JavaScript in web development. The fast and widespread adoption of WebAssembly by all major …

    waterloo Repository record for WasmWalker: Path-based Code Representations for Improved WebAssembly Program Analysis (opens in a new tab)

  14. Machine Learning Guided Discovery and Design for Inertial Confinement Fusion

    … volumes are large, and the relationship between models and experiments may be uncertain. We propose using machine learning to aid in the design and understanding of ICF implosions by integrating simulation and experimental data into a common frame-work. We begin by illustrating an early success …

    tamu Repository record for Machine Learning Guided Discovery and Design for Inertial Confinement Fusion (opens in a new tab)

  15. Proof Recommendation for the HOL4 Theorem Prover

    … knowledge, making it challenging for users to construct valid proofs. The HOL4 theorem prover, while a powerful tool in formal verification, presents usability challenges due to the intricate nature of proofs and the cognitive load placed on users. This thesis proposes an innovative solution …

    concordia Download PDF for Proof Recommendation for the HOL4 Theorem Prover (opens in a new tab)

  16. Automatic Speech Recognition without Transcribed Speech or Pronunciation Lexicons

    Rapid deployment of automatic speech recognition (ASR) in new languages, with very limited data, is of great interest and importance for intelligence gathering, as well as for humanitarian assistance and disaster relief (HADR). Deploying ASR systems in these languages often relies on cross-lingual …

    jhu Repository record for Automatic Speech Recognition without Transcribed Speech or Pronunciation Lexicons (opens in a new tab)

  17. The Application of Echo State Networks to Atypical Speech Recognition

    Automatic speech recognition (ASR) techniques have improved extensively over the past few years with the rise of new deep learning architectures. Recent sequence-to-sequence models have been shown to have high accuracy by utilizing the attention mechanism, which evaluates and learns the magnitude …

    rit Repository record for The Application of Echo State Networks to Atypical Speech Recognition (opens in a new tab)

  18. Towards End-to-end Non-autoregressive speech applications

    … speech research community, a very challenging topic researchers are interested in is the sequence-to-sequence labeling problem. Speech is a complicated signal, and many tasks aim to assign a sequence of various labels to the signal. Before, the traditional hybrid approach models this problem as …

    jhu Repository record for Towards End-to-end Non-autoregressive speech applications (opens in a new tab)

  19. Neural Sequence Modeling for Domain-Specific Language Processing: A Systematic Approach

    In recent years, deep learning based sequence modeling (neural sequence modeling) techniques have made substantial progress in many tasks, including information retrieval, question answering, information extraction, machine translation, etc. Benefiting from the highly scalable attention-based …

    vt Repository record for Neural Sequence Modeling for Domain-Specific Language Processing: A Systematic Approach (opens in a new tab)