Feng Hou
Doctor of Philosophy, (Computer Science)
Study Completed: 2021
College of Sciences
Citation
Thesis Title
Deep Learning for Entity Analysis
Read article at Massey Research Online:
Natural (human) language texts have many mentions of entities which can be a person, location, or organization. Entity analysis identifies and analyses different aspects of entity mentions for understanding natural language. Mr Hou developed novel deep learning methods to improve the computer programs for three sub-tasks of entity analysis: classifying entity mentions into fine-grained types, linking entity mentions to concrete entities in a knowledge base, and clustering co-referent entity mentions. For fine-grained classification, three transfer learning schemes were developed to learn more efficient representations and offset label noises in the datasets. For entity linking, typed entity representations were proposed to improve the learning of contextual commonality, and anonymous entity mentions were exploited to capture the document-level coherence of entities. For clustering co-referent entity mentions, more diversified mention representations were generated to distinguish related but distinct entities.
Supervisors
Professor Ruili Wang
Professor Yi Zhou
Page authorised by Web Content Manager
Last updated on Monday 04 April 2022