Intro -- Title Page -- Acknowledgements -- Abstract -- Notation -- Contents -- Introduction -- Aims and Objectives -- Contributions -- Thesis Structure -- Background -- Natural Language Generation -- Neural Networks in Natural Language Processing -- Language Modelling with Neural Networks -- Neural Networks as Generative Models -- Encoder-Decoder Framework -- Summary -- Evaluation Methodology -- Evaluation Methods -- Automatic Evaluation -- Human Evaluation -- Evaluating Multilingual Summaries from the Perspective of Wikipedia Readers and Editors -- Baselines -- Random
Kneser-Ney (KN) Language Model -- Information Retrieval (IR) -- Machine Translation (MT) -- Summary -- Building Corpora of Natural Language Texts Aligned with Knowledge Base Triples -- Automatically Aligning Texts and Triples -- Wikipedia Summaries -- Knowledge Base Triples -- Aligned Corpora -- Biographies -- The D3 Corpus -- Building Multilingual Corpora -- Discussion -- Summary -- Neural Wikipedian: Generating Biographies from Knowledge Base Triples -- The Model -- Triple Encoder -- Decoder -- Property-Type Placeholders -- Model Training -- Generating Summaries -- Dataset Preparation
Modelling the Textual Summaries -- Modelling the Input Triples -- Experiments -- Training Details -- Automatic Evaluation -- Human Evaluation -- Discussion -- Conclusion -- Learning to Generate Wikipedia Summaries for Underserved Languages -- Model -- Property Placeholders -- Dataset Preparation -- Experiments -- Training Details -- Automatic Evaluation -- Community Study -- Recruitment -- Readers' Evaluation -- Editors' Evaluation -- Conclusion -- Point at the Triple: Improving Neural Wikipedian with a Pointer Mechanism -- The Model -- Decoder -- Triple Encoder
Dynamically Expanding the Vocabulary -- Summarising By Pointing and Generating -- Dataset Preparation -- Experiments -- Training Details -- Automatic Evaluation -- Human Evaluation -- Summary and Discussion -- Conclusion and Future Work -- Summary and Conclusions -- Current Limitations and Future Work -- Generation of Multi-Sentence Summaries -- The Main Entity of Interest -- Using the S3 Corpus -- Bibliography
