Introduction

Question-Answering systems are systems that, given a large collection of documents and a question posed by the user, have to locate, in the entire collection, the answer to that question. Information Retrieval systems are systems that, from the user's query, return a set of documents amongst which it is possible to find the most relevant results. A question-answering system would take the result of the IR engine, analyse those documents, and return the exact answer, or at least a small span of text in which the answer appears. To locate the answer, the first step is usually to decide which is the kind of answer wanted by the user. It is common to analyse the question morpho-syntactically, and to categorise it depending on the type of answer (a date, a person name, a place, etc.) QA systems in general benefit from Information Extraction systems, to identify entities in the text. In fact, there is a kind of IE systems called user-adapted IE, which are very similar to QA systems if we consider that the user model is the query. Other related area is the automatic summary generation in answer to a question, task that has been applied to DUC-2005.

Our work in this area consisted in the participation in the Question-Answering track in the TREC-2001 competition, and the generation of user-focused summaries in DUC-2005.

Publications

On-line demos

Some external links

• Question-Answering in TREC