linguistlist.org>
Editor for this issue: Karen Milligan <karenlinguistlist.org>
dfki.de>Call for Papers IJCAI-99 Workshop NLP-2 International Joint Congress on Artificial Intelligence KNOWLEDGE AND REASONING IN PRACTICAL DIALOGUE SYSTEMS Endorsed by SIGDIAL Applications utilizing spoken natural language interaction are becoming increasingly common, mainly due to the development of speech technology. In addition to speech recognition and speech synthesis these systems require dialogue capabilities that would allow users to be engaged in a natural and efficient interaction. For the sake of robustness and portability current practical dialogue systems tend to rely on simple models for dialogue management (e.g. state transition diagrams or dialogue grammars) and simple representations (if any) of domain or task knowledge (e.g. frames) and for certain applications, such as information retrieval from a database, these models actually appear to be sufficient. On the other hand, dialogue models developed within AI tend to emphasize the relation between utterances and speakers' goals and plans, the importance of being able to reason about other agents' beliefs and intentions, and the need for domain knowledge and discourse representation for resolving anaphoric and deictic references. Somewhere in between we also find proposals that either augment the simpler dialogue models with generic and specific domain knowledge, or restrict the role of plan inference to specific situations. The workshop aims at studying the need for knowledge and reasoning in dialogue systems from theoretical and practical perspectives. Besides the innovative aspect of research, an emphasis is also laid on the importance of implemented dialogue systems as test-beds for evaluating the usefulness of theories and ideas, and on improvements in practical system abilities supporting a more natural and efficient interaction. We primarily seek contributions which discuss one or more of the following issues: - - What is the relation between different proposed knowledge sources, such as discourse models, task models, domain models, conceptual models and user models? - - How are such models integrated and coordinated? - - What are their roles in dialogue systems and how can they improve the analysis/generation of user/system utterances? - - What types of knowledge and reasoning is useful for various kinds of applications and situations? - - How can domain models and discourse models be used to handle focus and topic shifts? Do the models support topic associations in free conversations? - - How dependent is the system's functioning on a particular domain model? Can modularity of a system be supported if its domain model is changed to a different one? - - Can domain models be automatically built? How can empirical methods be used in building domain models? - - How can we evaluate domain models and their importance to the dialogue systems? Participants v============ The workshop will be kept small, with a maximum of 40 participants. Preference will be given to active participants selected on the basis of their submitted papers. According to IJCAI rules, all workshop attendees must register for the main conference. Workshop format =============== Ample time is allowed for commenting on the papers and open group discussion, thus ensuring time for discussions that go beyond ordinary conference style commenting. The workshop is intended to be a genuinely interactive event and a forum where new ideas and insights can take form via the collaboration of experts with diverse backgrounds. Demonstration of software could be organised as part of the main IJCAI conference. Program comittee ================== Masahiro Araki, Kyoto University, Japan Rolf Carlson, KTH, Sweden Koiti Hasida, ETL, Japan Diane Litman, AT&T, USA Susan McRoy, University of Wisconsin-Milwaukee, USA Massimo Poesio, University of Edinburgh, UK Mieke Rats, Compuleer, The Netherlands Norbert Reithinger, DFKI, Germany David Sadek, France Telecom, France David Traum, University of Maryland, USA Ingrid Zukerman, Monash University, Australia Submissions =========== Potential participants are invited to submit extended abstracts on the topics outlined above. Submissions should be no longer than 6 (six) pages exclusive of references, single spaced, and be in line with the IJCAI-style sheet, obtainable from http://www.dsv.su.se/ijcai-99/ Electronic submissions (in postscript format) should be sent to Jan Alexandersson at: jan.alexanderssonMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue
jprc.com>============================= Call for Participation ============================= AAAI-99 American Association for Artificial Intelligence Workshop on Machine Learning for Information Extraction July 18 or 19, 1999 Orlando, Florida http://www.isi.edu/~muslea/RISE/ML4IE/ The dramatic growth in the number and size of on-line textual information sources has fueled increasing research interest in the information extraction (IE) problem. Given a set of text documents from some domain, an IE system automatically populates a pre-defined database by extracting relevant fragments from the documents. Manually constructed IE systems cannot adapt to domain changes, and must be adapted for each new problem domain. In consequence, various machine learning (ML) techniques---symbolic learning, inductive logic programming, wrapper induction, statistical methods, and grammar induction---have recently been applied to the IE problem. This research has led to IE systems for several genres---newswire articles, medical texts, Web pages, and Usenet posts---that automatically learn to perform IE. The purpose of this workshop is to provide a forum for exploring the commonality underlying this diversity of problem domains and approaches. The workshop has three goals: 1. to bring together communities of researchers that address the IE problem from different perspectives (e.g., the use of natural language processing for IE from grammatical text vs. extraction from semistructured documents in the context of the information integration problem); 2. to deepen the IE community's understanding of the state of the art; and 3. to identify remaining IE-related problems for which ML techniques might be appropriate. TOPICS appropriate to this workshop include: - novel or improved ML techniques for IE; - novel types of IE domains; - extraction from structured and semi-structured documents; - effective use of features such as linguistic structure, mark-up, and document formatting; and - metrics and benchmarks for evaluating IE systems. FORMAT: In the interest of promoting as much discussion as possible, the number of paper presentations will be limited in favor of panels, invited talks, and posters. Attendance is limited to 40 participants. To be invited, please submit either a short paper (up to 6 pages) or a research statement (up to two pages, including related publications). In addition to traditional research papers, we encourage the submission of both position papers and survey papers. SUBMISSION REQUIREMENTS: Papers may be submitted by emailing the URL of a postscript version to mecalifMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue