Why-question generation based on causal discourse relations

I.M. Sleet, Yazeed (2010) Why-question generation based on causal discourse relations. Masters thesis, University of Malaya.

[img] PDF
Master Dissertation.pdf

Download (1MB)

Abstract

The Natural Language Understanding and Natural Language Generation are two of the grand challenges of artificial intelligence and play important roles in learning environments, information retrieval, helping systems and expert systems. In this research we want to show how the discourse and coherence relations could be used in the automatic ―wh-questions‖ generation. In particular we are interested in the causal discourse relation extraction and analysis and forming the ―why-question‖ based on the information gained from the causal discourse relations. This research consists mainly of two phases: the causal discourse relations extraction and the why-question generation. Causal discourse relation extraction phase is based on the “because” cue in a text. For example, if ―because‖ cue is found in a text in some patterns with some rules, then a causal discourse relation will be extracted. The extracted relations are shown in ―cause-effect‖ pairs such as each relation is determined by its cause and effect parts. The question generation phase depends on the results of the first part (the causal discourse relations extraction) to identify the question content. The effect part of the relation is taken to be the question‘s source. The effect part of the causal relation is parsed and analyzed to determine what we should take and what we should ignore in the question because not the whole of the effect part is important to be in the question phrase. The question generator can use the information gained from the causal relations to write the question phrase. In this study, we describe an automatic why-question generator. The system takes a text as input and returns a set of why-questions that can be answered from the given text. The whole system combined from the two phases has 78.18% precision and recall.

Item Type: Thesis (Masters)
Subjects: Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
Depositing User: MS NOOR ZAKIRA ZULRIMI
Date Deposited: 23 Jul 2013 06:32
Last Modified: 23 Jul 2013 06:32
URI: http://repository.um.edu.my/id/eprint/572

Actions (login required)

View Item View Item