WWW pages of 3rd European Master School on Language and Speech

A Question-Answering System

Jordi Adell Mercado, Josep Maria Crego Clemente, Rafael Pau Oliveros and Jesús Llevadias Jané
(Universitat Politècnica de Catalunya)

The aim of a Question Answering System is to seek in a corpus an answer given a question made using Natural Language (NL). For that purpose we use different NL Processing techniques such as syntactical or morphological analysis.

Our system has four parts. The first part is the Question Processing which identifies the question type, extracts the keywords and expands them using WordNet. The next part, Information Retrieval, is divided in an offline indexation of the documents and an online document filtering in order to extract the most suitable ones.In the third part, Segmentation, with finer resolution methods, we divide these documents obtaining the most significant paragraphs. The last part searches the answer into the best paragraphs according to the question type and the keywords, using syntactical analysis.