AI Magazine Volume 23 Number 2 (2002) (© AAAI)

Workshop Reports

Knowledge Media Institute at The Workshop on Human Open University in England, gave the keynote entitled “Supporting Organizational Learning through the Language Technology and Enrichment of Documents.” Accord- ing to Domingue, only a small per- Knowledge Management centage of corporate training is ever applied within the workplace because organizations tend to use school- based methods of learning in con- trast to organizational learning based Mark T. Maybury on theories of learning in the work- place. Domingue described knowl- edge sharing by enriching web docu- ments with informal and formal representations, a process that cap- tures the context in which a docu- The Workshop on Human Language technologies that could enable ment is created and applied. Technology and Knowledge Manage- knowledge management functions Domingue demonstrated how this ment was held on July 6 and 7 in such as the following: enrichment facilitates retrieval and Toulouse, France, in conjunction Expert discovery: Modeling, cata- comprehension. with the meeting of the Joint Associ- loging, and tracking of distributed In addition, the group heard an ation for Computational Linguistics organizations and communities of invited talk from Hans Uszkoreit and European Association for Com- experts (DFKI Saarbruecken), scientific direc- putational Linguistics (ACL / EACL Knowledge discovery: Identifica- tor at the German Research Center ’01). Human language technologies tion and classification of knowledge for Artificial Intelligence (DFKI), head promise solutions to challenges in from unstructured multimedia data of DFKI Language Technology Lab, human-computer interaction, infor- Knowledge sharing: Awareness of, and professor of computational lin- mation access, and knowledge man- and access to, enterprise expertise guistics at the Department of Com- agement. Advances in technology and know-how putational Linguistics and Phonetics areas such as indexing, retrieval, Table 1 (from Mark Maybury’s of at Saarbrück- transcription, extraction, translation, introduction to the workshop) illus- en. Uszkoreit’s talk was entitled trates how these knowledge manage- and summarization offer new capa- “Crosslingual Language Technologies bilities for learning, playing, and for Knowledge Creation and Knowl- conducting business. These adances This article summarizes the results edge Sharing.” He described how promise to support enhanced aware- of the 6–7 July Workshop on “language technology can provide ness, creation, and dissemination of Human Language Technology and means for associating shared knowl- enterprise expertise and know-how. Knowledge Management held in edge with the relevant decision situa- Organized by the European Net- Toulouse, France. It describes invited tions by automatically linking it to keynotes, presentations, and results work of Excellence in Human Lan- the critical elements within decision of brainstorming sessions to create a gauge Technologies (Steven Krauwer, technology road map for this impor- triggers, that is, electronic documents U. Utrecht) and The MITRE Corpora- tant area. The group also articulated in the work flow that demand and tion (Mark Maybury), the workshop grand challenges in human lan- record a decision.” Uszkoreit de- brought together a group of 50 com- guage technology and solutions to scribed the role of information putational linguists, AI researchers, these challenges that could benefit extraction, automatic hyperlinking, and computer scientists from North facilities for knowledge discovery, and (human) inferencing in this pro- America, Europe, Asia, Australia, and access, and exploitation. cess. He exemplified this “automatic South Africa working in a range of relational hyperlinking” using the areas (for example, speech and lan- ment functions are supported by a example of a hypercode system guage processing, translation, sum- developed for a large German bank broad range of human language marization, multimedia presentation, to facilitate work with legacy code by technologies, including query analy- content extraction, dialog tracking) densely interlinking source code and sis-retrieval, information extraction, both to report advances in human documentation. Uszkoreit concluded language technology and their appli- question answering, machine trans- by addressing cross-lingual knowl- cation to knowledge management lation, agent-user modeling, summa- edge management, describing his and to work toward a road map for rization, presentation generation, efforts to augment general-purpose the human language technologies for and awareness-collaboration. translation systems with specialized the next decade. In part, the work- During the second day, John terminology and transfer rules for shop focused on human language Domingue, deputy director of The multilingual expert groups in a pro-

Copyright © 2002, American Association for Artificial Intelligence. All rights reserved. 0738-4602-2002 / $2.00 SUMMER 2002 99 Workshop Reports

Human Language Grand Challenges Benefits to Knowledge Discovery, Technology Access, Exploitation

Interpretation of imprecise, ambiguous, or Natural (written, spoken, gestural) access to Input or Query partial multimodal input. Facilities include information and knowledge. Decrease in access Analysis spoken query processing, visual query complexity or user training. Broaden availability of analysis (e.g., sketching), and mixed-media knowledge to users. query (e.g., text and graphics).

Natural language processing of queries and Enhancements to document-retrieval precision and Retrieval documents. Content-based retrieval of text, recall. Direct access to media, easing navigational imagery, audio, video. burden of user. Reduction of search time.

Direct access to information or knowledge elements, Segmentation, object and event Extraction identification, and extraction from including specific types that might be user preferred. Reuse of media elements enabling user-tailored multimedia sources (text, audio, video). selection or presentations.

Question analysis, response discovery and Overcome time, memory, or attention limitations generation from heterogeneous sources (e.g., Question Answering required to sift through many returned web pages multilingual; multimedia; unstructured, from a traditional search by providing direct answers structured, semistructured) to questions.

Rapid creation of translingual corpora. Effective translingual retrieval, Cross media-mode information and knowledge access Translation summarization, and translation. Access enabling broader access to global information sources verbalization of graphics, visualization of using methods such as translingual information text. retrieval.

Mixed-initiative natural interaction that Ability to tailor flow and control of interactions and Dialogue facilitate interactions. Includes error detection and deals robustly with context shift, correction tailored to individual physical, perceptual, Management interruptions, feedback, and shift of locus of control. and cognitive differences. Motivational and engaging lifelike agents.

Agent or User Unobtrusive learning; representation; and Enables tracking of user characteristics, skills, and use of characteristics, beliefs, goals, and goals to enhance interaction as well as discovery of Modeling plans of agents (including the user). experts by other users or agents.

Scaleability, cross-linguality, multimedia Increasing speed of reviewing materials. Multimedia Summarization summarization, cross-lingual summarization, large summarization. multidocument summarization.

Automated generation of coordinated Mixed media (e.g., text, graphics, video, speech, and speech, natural language, gesture, animation, nonspeech audio) and mode (e.g., linguistic, visual, Presentation nonspeech audio, generation, possibly auditory) displays tailored to the user and context. Generation delivered via interactive, animated lifelike Agents engaging and motivating to younge or less agents (includes challenges of media experienced users. selection, allocation, coordination, and realization)

Topic detection and tracking, place-based Awareness and Enhance awareness of new knowledge, as well as other asynchronous and synchronous user’s interests and expertise, and the ability of experts Collaboration collaboration environments. to exchange or integrate knowledge.

Table 1. Human Language Technology for Knowledge Management.

ject for a large multinational automo- sessions were held in ontology con- focusing on construction of a road bile manufacturer. struction, question answering, sum- map. During these sessions, the A poster session included system marization, multilingual processing, group focused on an analysis of the demonstrations and offered partici- multimedia processing, and dialogue. present situation, a vision of where pants an opportunity for rich dia- Group brainstorming sessions fol- we want to be in the future, and a logue and interaction. Major papers lowed each major technology theme, number of intermediate milestones

100 AI MAGAZINE Workshop Reports that would help in setting intermedi- and ontologies; the importance of ate goals and measuring our progress domain knowledge and the adapta- toward our goals. tion-integration of semantic re- The group outlined key challenges sources; the complexity of dealing and promising solutions in the areas with one-to-one translation of even of ontology, summarization, multi- the 200 most spoken-written lan- lingual processing, and multimedia guages (requiring 39,000 language processing. With respect to ontolo- pairs); the need for large-scale, robust gies, the group emphasized the need natural language processing and, at for tools and tasks that were reusable the same time, the importance of across domains to create and popu- fine-grained linguistic knowledge; late ontologies; the importance of a and the challenge of new application user-centered process view; the need domains such as content-driven to integrate shallow and deep meth- hypertextual authoring and crosslin- ods; the need to collaborate with gual news linking. domain ontology creators; and the The group identified resources (for need to address ontology quality, example, WORDNET, EURONET, applica- ambiguity, and usability (for exam- tion databases, text resources) as key ple, using tools for structuring, inte- to advancement, the INTERLINGUA grating, visualizing, and accessing approach as promising, the impor- massive or heterogeneous ontolo- tance of deeply annotated data com- gies). The group highlighted the bined with machine learning, the Simulating promise of the semantic web, the promise of translation memories and Organizations importance of information extraction machine learning, and the possibility of tailoring multiple ontologies to “plug-ins,” the possibility of organiz- Computational Models of ing massive documents using users and their tasks. domain-specific ontologies, the Finally, the group turned its atten- Institutions and Groups opportunity to use a “top” or core tion to multimedia challenges and ontology to bootstrap new domains, opportunities. Challenges include the Edited by Michael J. Prietula, the value of multidisciplinary collab- integration of multiple media; the Kathleen M. Carley, and Les Gasser orative teams (for example, domain nature of processing (is it centralized experts, linguists, knowledge engi- or mobile); the challenges of privacy, 6 x 9, 350 pp., ISBN 0-262-66108-X neers), the value of controlled lan- security, and scalability; the impor- tance of both remembering and for- To order, call 800-405-1619 guage management, and the promise (http://mitpress.mit.edu) of component-based methods to getting information; the need for multilingual and multisource infor- facilitate ontology decomposition, Distributed by mation extraction; and the challenge reuse, and life-cycle management. The MIT Press, of cross-document coreference resolu- With respect to summarization, the 5 Cambridge Center, group outlined challenges as includ- tion. Location-based services were Cambridge, MA 02142 ing the appropriate level-depth of highlighted as a promising future area. analysis-representation (for example, Two cross-cutting enabling semantic relations, speech acts, capabilities were identified for all the rhetorical structure), summarization Mark Maybury is exec- addressed areas. First is the need for utive director of presentation-visualization, speech for (intelligent) text annotation. Second MITRE’s Information presentation of short summaries, the is the need for large-scale annotated Technology Division in appropriate use of indicative versus corpora to enable automated training Bedford, Massachusetts. informative summaries, and the need and system evaluation. He is a member of the for action-oriented summaries (for board of directors of the ELSNET has captured the workshop example, executive-management Object Management input and will continue to revise a summaries). The group discussed a Group, secretary-treasur- technology road map. A web site to range of solutions encompassing the er of the Association of Computing share the materials and results of the analysis of information, its transfor- Machinery SIGART, and a member of the workshop has been set up.1 Intelligent User Interfaces Steering Coun- mation (including operations such as cil. Maybury has published over 50 tech- selection, aggregation, abstraction), Notes nical and tutorial articles and is editor of a and its presentation. 1. www.elsnet.org/acl2001-hlt+km.html. number of books. Maybury received a The group also identified a number doctorate in AI from Cambridge Universi- of fundamental multilingual chal- ty in England in 1991. Maybury is an lenges, including relations between international adviser to the German Min- cultures, languages, lexical resources, istry for Education and Research.

SUMMER 2002 101 Workshop Reports Intelligent Systems for Molecular Biology

ISMB-2000—San Diego, California ISMB-96—St. Louis, Missouri Proceedings of the Eighth International Conference Proceedings of the Fourth International Conference on Intelligent Systems for Molecular Biology on Intelligent Systems for Molecular Biology Edited by , Timothy Bailey, , Michael Grib- Edited by David J. States, Pankaj Agarwal, , skov, , Ilya Shindyalov, Lynn Ten Eyck, and Helge , and Randall F. Smith Weissig ISBN 1-57735-002-2 274 pp., index, $65.00 softcover ISBN 1-57735-115-0 436 pp., index, $65.00 softcover ISMB-95—Cambridge, England ISMB-99—Heidelberg, Germany Proceedings of the Third International Conference Proceedings of the Seventh International Confer- on Intelligent Systems for Molecular Biology ence on Intelligent Systems for Molecular Biology Edited by Christopher Rawlings, Dominic Clark, Russ Altman, Edited by Thomas Lengauer, Reinhard Schneider, Peer Bork, Dou- Lawrence Hunter, Thomas Lengauer, and glas Brutlag, Janice Glasgow, Hans-Werner Mewes, and Ralf Zim- ISBN 0-929280-83-0 427 pp., index, $65.00 softcover mer ISBN 1-57735-083-9 324 pp., index, $65.00 softcover ISMB-94—Stanford, California Proceedings of the Second International Confer- ISMB-98—Montréal, Quebec, Canada ence on Intelligent Systems for Molecular Biology Proceedings of the Sixth International Conference on Intelligent Systems for Molecular Biology Edited by Russ Altman, Douglas Brutlag, , Richard Lath- rop, and David Searls Edited by Janice Glasgow, Tim Littlejohn, François Major, Richard ISBN 0-929280-68-7 401 pp., index, $65.00 softcover Lathrop, , and Christoph Sensen ISBN 1-57735-053-7 234 pp., index, $65.00 softcover ISMB-93—Bethesda, Maryland Proceedings of the First International Conference ISMB-97—Halkidiki, Greece on Intelligent Systems for Molecular Biology Proceedings of the Fifth International Conference on Intelligent Systems for Molecular Biology Edited by Lawrence Hunter, David Searls, and Jude Shavlik ISBN 0-929280-47-4 468 pp., index, $65.00 softcover Edited by Terry Gaasterland, Peter Karp, Kevin Karplus, Christos Ouzounis, , and ISBN 1-57735-022-7 382 pp., index, $65.00 softcover Published by The AAAI Press, 445 Burgess Drive, Menlo Park, California 94025

To order, call (650) 328-3123 or fax to (650) 321-4457. MasterCard and VISA accepted.

102 AI MAGAZINE