FST Graduate Research Topic

Research Topic: Automated question answering based on semantic analysis and statistical learning
Suggested By: Dunwei (Grant) Wen
Last Update: September 2nd, 2009

This research project focuses on automated question answering (QA), which is a branch of natural language processing and information retrieval and regarded as the next generation of search engine. To improve the performance of QA, semantic analysis, especially semantic role labeling (shallow semantic parsing) will be used for deeper understanding of both questions and answer candidates produced by search engine. Statistical learning techniques such as semi-supervised EM and conditional random field (CRF) will be applied to question analysis and answer generation. This research will also take into consideration certain Web resources relevant to short questions and answers such as collections of emails, mailing lists, news groups, discussion forums and FAQ in organizations and public domains, which store huge amount of question-answer pair like information and knowledge. For example, there are lots of threads of posts or emails in online learning (or technical supporting) platforms between students (or clients) and their instructors (or service providers). An accordingly designed automatic question-answering (QA) system can be implemented to improve the precision in producing answers compared to those designed for open domain free text resources. The graduate students who want to pursue the research need to be familiar with text processing and Artificial Intelligence, including machine learning, natural language processing, and uncertainty reasoning techniques, or at least have background in data mining. This research topic is suitable for Project or Thesis.