ILQUA--An IE-Driven Question Answering System.
Min Wu, Michelle Duan, Samira Shaikh, Sharon Gower Small, Tomek Strzalkowski · Text REtrieval Conference · 2005
ILQUA first participated in TREC QA main task in 2003. This year we have made modifications to the system by removing some components with poor performance and enhanced the system with new methods and new components. The newly built ILQUA is an IE-driven QA system. To answer “Factoid” and “List” questions, we apply our answer extraction methods on NE-tagged passages. The answer extraction methods adopted here are surface text pattern matching, n-gram proximity search and syntactic dependency matching. Surface text pattern matching has been applied in some previous TREC QA systems. However, the patterns used in ILQUA are automatically generated by a supervised learning system and represented in a format of regular expressions which can handle up to 4 question terms. N-gram proximity search and syntactic dependency matching are two steps of one component. N-grams of question terms are matched around every named entity in the candidate passages and a list of named entities are generated as answer candidate. These named entities go through a multi-level syntactic dependency matching until a final answer is generated. To answer “Other” questions, we parse the answer sentences of “Other” questions in 2004 main task and built syntactic patterns combined with semantic features. These patterns are applied to the parsed candidate sentences to extract answers of “Other” questions. The evaluation results showed ILQUA has reached an accuracy of 30.9% for factoid questions. ILQUA is an IE-driven QA system without any pre-compiled knowledge base of facts and it doesn’t get reference from any other external search engine such as Google. The disadvantage of an IE-driven QA system is that there are some types of questions that can’t be answered because the answer in the passages can’t be tagged as appropriate NE types. Figure 1 shows the diagram of the ILQUA architecture.