IJCAI-97 is sponsored by the International Joint Conferences on Artificial Intelligence, Inc. (IJCAII), and the Japanese Society for Artificial Intelligence (JSAI).
A number of ontology-related workshops have been held in the past years (e.g., 1993 in Padua, 1995 IJCAI, 1996 ECAI, 1997 AAAI Spring Symposium, etc.). However, none of them concentrated centrally on applications of world modeling to multilingual Natural Language Processing (NLP).
Ontologies for knowledge-based computing and especially for Natural Language Processing are steadily reaching a level of sophistication and size which make them increasingly useful to the resolution of problems in real-world NLP applications. The recent creation of an ad hoc ANSI working group on standardization of ontologies is an indication of the maturity of the field. More and more ontology-based systems are being built for multilingual applications (e.g., multilingual machine translation, multilingual information retrieval). However, most of the language-processing oriented ontologies that have been built so far have English or another language (e.g., Japanese or Spanish) as the basis (e.g., WordNet, EDR, Pangloss, etc.). Since there is a growing need for multilingual applications of these ontologies, it is natural to ask the following questions: Are any of these ontologies actually used in a multilingual setting? Can we characterize the degree of independence of an ontology from the natural language it is based on? What are the necessary properties of a truly multilingual (or universal) ontology? Is it possible to obtain a language-neutral ontology from a language-dependent ontology? What applications truly need multilingual (or language-neutral) ontologies? How do we separate language-specific (or lexical) information from ontological knowledge? How can the depth of knowledge in the ontology be balanced with the needs of an application? What are the prospects of automating ontology acquisition? What is the relationship between an ontology as the repository of general knowledge about the world and knowledge about particular individuals - people, places, organizations, events, etc.?
These and many more questions must be discussed much more widely than they have been till now. Many of the previous workshops were devoted to more formal issues in ontology building, such as the knowledge representation schemata, closures, formal properties of ontologies, and so on. Moreover, they included the discussion of small ontologies that cover a very narrow domain of problem solving; NLP typically requires a broad-coverage ontology. The hypothesis of using interlingual representations based on an ontology is at least 50 years old. It was originally formulated in the framework of machine translation. However, few systems to date have tested this hypothesis, for MT or other applications, by implementing a large-scale interlingua-based system using a language-independent ontology. This workshop will debate the benefits, costs and competitiveness of such an approach to solving semantic and cross-language problems for MT, IR, and other NLP applications.
The workshop is open to all members of the AI and NLP community. The workshop is intended for researchers and practitioners in knowledge-based NLP, artificial intelligence and computational linguistics who have been working on large scale knowledge-based resources, ontologies, multilingual lexical semantics, interlinguas, and their applications. Reports of actual work including problems and solutions in the design, construction and use of ontologies are strongly encouraged but more theoretical work (grounded on actual work on ontologies) aimed at defining the limits, constraints and directions for large-scale practical language-neutral ontologies is welcome as well.
Issues to be addressed include but are not limited to:
The workshop will include twelve presentation periods which will be divided into ten-minute presentations of positions followed by 20-minute discussions. Presentation will be short, under 15 minutes, with 20 minutes reserved for exchanges. We encourage the authors to focus on the salient points of their presentation and identify possible controversial positions. We encourage authors not to repeat as is what has been already written in the paper. There will be ample time set aside for informal and panel discussions and audience participation. Please note that workshop participants are required to register at the main IJCAI-97 conference.