quetallogo
MULTILINGUAL HYBRID QUESTION ANSWERING

 
 






dfkiltlogo
 
Introduction
Overview

The aim of QUETAL is the design, implementation and evaluation of a question answering (QA) system that interactively retrieves correct answers to a wide variety of natural-language queries from large volumes of textual documents. The system combines several technologies in (crosslingual) information retrieval, information extraction and deep linguistic processing from previous projects. A combination of shallow and deep processing methods is utilized for the analysis of questions including sentence fragments and non-well-formed queries. In order to provide an accurate, robust and efficient answering of questions from a nearly unrestricted number of domains, the project is exploiting at three different sources of information. Databases of facts such as product DBs, directories, structured catalogues or the results of information extraction can be queried if the answer to the input question can be expected in such a DB.

QA technologies are explored that operate on semi-structured NL documents. Meta-information (such as HTML or XML tags) for answer candidate selection is taken into account. We are investigating QA technology that can handle questions and answers going beyond the currently dominating fact-based, short-answer oriented systems.

QUETAL attempts to extend state-of-art QA functionality, by
  • Uniform NL-based access to hybrid data sources
  • Crosslingual question answering
  • Data-driven QA control
building on
  • Multi-dimensional document annotation
  • Interleaved shallow and deep NLP
  • Robust variable-depth semantic processing

Serving as the scientific core project of the Lab, Quetal also has the task of maintaining and updating the existing technological base. This base includes the language processing platforms for text analysis and generation, information extraction and information retrieval as well as the linguistic knowledge bases such as lexicons, morphological and syntactic grammars, ontologies etc.

QUETAL is funded from 2003-2005 as a three-year project by the German Federal Ministery for Research (BMBF) under contract 01 IW C02.


http://quetal.dfki.de/index.html   updated 2005-08-23 18:10:12
DFKI GmbH, Language Technology Lab