Abstract:This paper describes the systems of THU QUANTA in Text Analysis Conference (TAC) 2008. We participated in the Question Answering (QA) track, and the Recognizing Textual Entailment (RTE) track. For question answering track, we enhanced the traditional question answering system by sentiment lexicon based opinion analysis. The rigid list questions are divided into two categories based on their answer types. And two snippet extraction approaches are proposed to answer squishy list questions. For RTE track, we design different strategies to recognize true entailment and false entailment. The similarity between hypothesis and text is measured to recognize true entailment. We detect the exact entity and relation mismatch to recognize the false entailment. The evaluation results show that the proposed approaches are very effective for the QA and RTE tasks. In this year's Text Analysis Conference, we participated in two tracks: the Question Answering (QA) track and the Recognizing Textual Entailment (RTE) track. This paper reports on our two developed systems for these two tracks. The TAC 2008 QA track is different from previous TREC QA. It focuses on finding answers to opinion questions. We develop the opinion question answering system by enhancing our past participated system QUANTA(3) with lexicon based sentiment analysis. We not only consider the topic relevance, but also pay attention to the sentiment match between question and answers. By analyzing the rigid list questions, we divide all the rigid list questions into two categories: the Opinion Holder Rigid List questions and Other Types Rigid List questions. Opinion Holder Rigid List questions refer to the questions whose answers are blog nicknames or blog holders. Other Types Rigid List questions refer to the questions whose answers are same to traditional factoid questions, such as person names, cities etc. The different strategies are designed for these two types of rigid list questions. We also propose two snippet extraction approaches to answer

FDUQA on TREC 2006 QA Track.

FDUQA on TREC 2004 QA Track.

FDUQA on TREC2003 QA Task.

FDUQA on TREC 2007 QA Track.

Insun05QA on QA Track of TREC 2005.

FDU at TREC-9: CLIR, Filtering and QA Tasks.

THU QUANTA at TAC 2008 QA and RTE track

A Question-Answering System over Traditional Chinese Medicine

THU QUANTA at TAC 2009 KBP and RTE Track.

FDU at TREC 2002: Filtering, Q&A, Web and Video Tasks.

Peking University at the TREC 2005 Question and Answering Track.

THUIR at TREC 2004: QA.

FDU at TREC-10: Filtering, QA, Web and Video Tasks

TREC-10 Experiments at CAS-ICT: Filtering, Web and QA.

Finding "abstract Fields" of Web Pages and Query Specific Retrieval - THUIR at TREC 2004 Web Track

Insun05QA on QA track of TREC2005

Leverage Web-based Answer Retrieval and Hierarchical Answer Selection to Improve the Performance of Live Question Answering.

FDU at TREC 2007: Opinion Retrieval of Blog Track.

At the web track of TREC 2003

WIM at TREC 2005.

THUIR at TREC2008: Enterprise track