IJCAI Workshop, Edinburgh, July 30th, 2005 KRAQ'05: KNOWLEDGE and REASONING for ANSWERING QUESTIONS Objectives The introduction of reasoning capabilities in question-answering (QA) systems appeared in the late 70s. A second generation of QA systems, aimed at being cooperative, emerged in the late 80s - early 90s. In these systems, quite advanced reasoning models were developed on closed domains to go beyond the production of direct responses to a query, in particular when the query has no response or when it contains misconceptions. More recently, systems such as JAVELIN, Inference WEB or Cogex, operating over open domains, integrate gradually inferential components, but not as advanced as those of the 90s. Performances of these systems in the recent TREC-QA tracks show that reasoning components do improve the response relevance and accuracy. They can also potentially be much more cooperative. However, there is still a long way before being able to produce accurate, cooperative and robust QA systems. Recent foundational, methodological and technological developments in knowledge representation (e.g. ontologies, knowledge bases incorporating various forms of incompleteness or uncertainty), advanced reasoning forms (e.g. relaxation, intensional calculus, data fusion), not necessarily based on unification, advanced language processing resources and techniques (for question processing as well as for generating responses), and recent progress in HLT make it possible to foresee the elaboration of much more accurate, cooperative and robust systems dedicated to answering questions from textual data, from e.g. online texts or web pages, operating either on open or closed domains. The workshop will be organized around a few major questions of interest to a number of AI, NLP, HLT and pragmatics people. One main question is the characterization of those reasoning procedures that need to be developed to answer questions, either on closed or on open domains. Then, are enhancing reasoning procedures and accuracy of knowledge representation sufficient conditions to improve responses ? If not, what is the role of language processing and what are the relevant paradigms (e.g. lexical inference) ? How do language and reasoning interact ? Next, what are the language models and techniques appropriate for producing responses which sound natural for the user (relevant, fluid, of an appropriate granularity, with terms the user understands, etc.). Another perspective is the role of pragmatics as a means, for example, to better capture the user's goals and intentions from his query, and therefore to better organize the response. Pragmatics is also of importance to better analyse the potential implicatures the user may draw from NL responses, in particular when the response is not direct. List of topics: - Methodologies for intelligently answering questions, - New types of questions and related KR, pragmatic and linguistic paradigms: procedural questions (how), causal questions (why), questions with comparative expressions, questions with negation, etc. - Reasoning aspects: * information fusion, * search criteria expansion models (e.g. relaxation techniques), * summarization and intensional answers, * reasoning under uncertainty or with incomplete knowledge, * Detecting and resolving query failure (due to e.g. incomplete data, misconceptions or false presuppositions) - Knowledge representation and integration: * levels of knowledge involved (e.g. ontologies, domain knowledge), * knowledge extraction models and techniques to optimize response accuracy, * coherence and integration. - Flexible and interactive systems possibly including a user model, - Pragmatic dimensions of intelligently answering questions: * user intentions, plans and goals recognition in questions, * conversational implicatures in responses, * principles for the design of cooperative systems. - Language processing: * question processing : parameters of interest for response production, * response generation (e.g. lexical choice, templates), * use of language resources for reasoning in question-answering, * explanation production (showing sources and inferences, reporting data incompleteness, etc.) - Evaluation * End-to-end evaluation of complex question types, * Intrinsic evaluation of inference methods, * Data-intensive vs knowledge-intensive methods, * portability techniques for closed domains. Submissions: The goal of this workshop is to enhance cooperation between participants with an AI background and the NLP and question-answering communities. Contributors must be opened to interactions with the different workshop areas. The programme committee will care to have a balanced number of participants from the different areas concerned: reasoning and inference, knowledge representation, NLP (in particular language generation), question-answering, human language technology and pragmatics. Although papers will obviously have a dominant theme, it is important that they contain material from at least 2 disciplines of the workshop (AI, NLP, pragmatics, ...). To encourage an athmosphere appropriate for a workshop, we plan to: - have a 15mn discussion at the end of each session, - have a panel on future directions of intelligent question-answering and on how the different disciplines can interact as optimally as possible, - have a session of demonstrations and posters. Submission format: We welcome short papers (max 5 pages), describing projects or ongoing research and long papers (max. 10 pages), that relate more established results. Papers must be sent in .pdf format. The format to use for papers and abstracts is the same as for IJCAI. Please follow the IJCAI formatting instructions and use the supplied Word templates or Latex sources. The title page (no separate title page is needed) should include the following information: Title Authors' names, affiliations, and email addresses Topic(s) of the above list, as appropriate Abstract (short summary up to 5 lines) Deadlines: March 10: paper submissions (sent to benamara@irit.fr) April 20: acceptance/rejection notification May 15: final papers due, camera-ready May 25: manuscript sent to IJCAI for printing by organizers. Publication: All accepted papers (long and short) will be published in the workshop proceedings. A book publication is under project. Registration: The registration fees include attendance at the workshop and a copy of the workshop proceedings. Registration instructions will be posted here. Workshop co-chairs and contact persons: Dr. Farah Benamara and Dr. Patrick Saint-Dizier (benamara@irit.fr, stdizier@irit.fr) Programme committee decisions will be co-chaired with: Dr. Marie-Francine Moens (marie-france.moens@law.kuleuwen.ac.be) Programme Committee: Farah Benamara, IRIT, France Johan Bos, University of Edinburgh, UK Sanda Harabagiu, University of Texas, USA Eduard Hovy, ISI, USA Daniel Kayser, LIPN, France Mark Maybury, The MITRE Corp., USA Michael Minock, University of Umea, Sweden Marie-Francine Moens, KUL, Belgium Jacques Moeschler, Geneva university, Switzerland Dan Moldovan, University of Texas, USA John Prager, IBM, USA Ehud Reiter, University of Aberdeen, UK Maarten de Rijke, University of Amsterdam, The Netherlands Gérard Sabah, LIMSI, CNRS, France Patrick Saint Dizier, IRIT, CNRS, France Manfred Stede, University of Potsdam, Germany Mathiew Stone, Center of Cognitive Science, Rutgers, USA Kees Van Deemter, University of Aberdeen, UK Ellen Voorhees, NIST, USA Bonnie Webber, University of Edinburgh, UK