********************************************************************** CALL FOR PAPER Held in conjunction with T A L N 2 0 0 4 Workshop on QUESTION-ANSWERING --- Palais de Congr=E8s F=E8s (Maroc) April 22, 2004 http://www.lpl.univ-aix.fr/jep-taln04/ ********************************************************************** Facing a question such as =ABWhat is the most expensive car in the world? , classical search engines return the documents that are the most strongly linked to the words of the question, sometimes extract the excerpts where these words are the most numerous, but let the user browse texts to actually find an answer. This need leads to develop systems that are able to extract the parts of documents that are the most relevant in relation to a question, providing either an answer when the question is about a precise fact or a summary when it is a topical question. These functions can be implemented only if IR systems are able to analyze both queries and documents more deeply. As a consequence, question answering is at the crossing of several research fields: of course, it is grounded in Information Retrieval but it also concerns Natural Language Processing (NLP) in an important way and to some extent, fields such as Machine Learning. Most QA systems are based on a classical search engine that is enhanced by a question analysis module, a set of modules for extracting various linguistic features from documents, such as named entities, terms or syntactic relations, and a module that relies on all these data for extracting answers by mixing linguistic and numerical criteria. Moreover, the QA problem puts forward new functions, or functions that are still in an embryonic state in current IR systems: evaluating if an answer to a question exists in a document collection, achieving a synthesis from multiple or partial answers, using dialog for constructing a query, or text understanding capabilities for dealing with anaphora, inferences, or for determining if a set of several answers is coherent. More precisely, submissions will present a question answering system as a whole or will focus on one of its processes provided that it is put in the question answering context. These processes include but are not limited to: - question analysis: question typology, extraction of the question focus, of the question context or more generally, of semantic constraints - named entity recognition: fine-grained named entities, unrestricted domains - passage extraction - full or partial similarity of syntactic structures - terminological tools: extraction and recognition of terms and their variants - extraction and justification of answers: answer patterns, inferences, paraphrase =85 This workshop is particularly concerned by papers that focus on QA systems for large collections of documents or the Web but papers about QA systems for restricted domains or dedicated to knowledge bases or database will also be taken into account. Submissions can also tackle cross-domain topics in relation to Question Answering , such as: - QA and machine learning: use of machine learning for selecting and extracting answers to a question but also for building on a large scale resources that are necessary for QA systems; - multilingual and crosslingual QA: what are the difficulties for adapting an existing QA system most of them only work for English to another language; asking a question in a language and searching an answer in a collection of documents in another language; - QA and the Web: using the Web as a source of knowledge or a source of answers; what are the specific aspects of searching an answer on the Web; - multi-document QA: fusion and coherence of multiple answers. SUBMISSION: Submissions will be minimum 4 page summaries or long papers of no more than 10 pages, written in French or English, according to the style of the main conference TALN 2004. The final version will be a long paper. Submission format will be PDF, but .doc and .ps will be also admitted. Papers have to be sent to Brigitte.Grau@limsi.fr, with TALN-QA as subject. IMPORTANT DATES: Submission deadline: 15 January 2004 Notification to authors: 20 February 2004 Camera-ready: 8 March 2004 Question-Answering workshop: 22 April 2004 ORGANIZATION COMMITTEE Brigitte Grau, LIMSI, Orsay (responsable) Olivier Ferret, LIC2M, CEA, Fontenay Gabriel Illouz, LIMSI, Orsay Laura Monceaux, IRIN, Nantes Thierry Poibeau, LIPN, Villetaneuse Isabelle Robba, LIMSI, Orsay Anne Vilnat, LIMSI, Orsay SCIENTIFIC COMMITTEE Massi-Reza Amini, LIP6, Paris Patrice Bellot, LIA, Avignon Mohand Boughanem, IRIT, Toulouse Jean-Pierre Chevallet, CLIPS, Grenoble Khalid Choukri, ELDA, Paris Olivier Collin, France Telecom, Lannion Olivier Ferret, LIC2M, CEA, Fontenay Patrick Gallinari, LIP6, Paris Brigitte Grau, LIMSI, Orsay Gabriel Illouz, LIMSI, Orsay Guy Lapalme, RALI, Canada Claude de Loupy, Sinequa Jean-Luc Minel, LALICC, Paris Laura Monceaux, IRIN, Nantes Thierry Poibeau, LIPN, Villetaneuse Isabelle Robba, LIMSI, Orsay Patrick Saint-Dizier, IRIT, Toulouse Anne Vilnat, LIMSI, Orsay Pierre Zweigenbaum, STIM, AP-HP Paris Groupe LIR - LIMSI BP 133, 91403 Orsay Cedex tel. 01 69 85 80 03, fax 01 69 85 80 88 et Institut d'Informatique d'Entreprise (IIE) 18 all=E9e Jean Rostand, 91025 Evry Cedex mail : grau@iie.cnam.fr tel. 01 69 36 73 44, fax 01 69 36 73 09