linguistlist.org>
Editor for this issue: Ann Dizdar <annlinguistlist.org>
ets.org>-------------------------------------------------------- SECOND CALL FOR PAPERS ACL'97/EACL'97 Workshop July 12, 1997 Madrid, Spain -------------------------------------------------------- "From Research to Commercial Applications: Making NLP Technology Work in Practice" Success in the marketplace is one form of validation for NLP techniques and underlying theories. The broad vision of this workshop is to bring together researchers to discuss commercial or commercial-bound systems that use NLP for either text or speech. We are interested in learning about systems that show promise in re-using NLP techniques, and in the process of technology transfer for NLP applications. Another topic of interest in this workshop is industry-based practical considerations involving NLP technology. The workshop should invoke discussion about experiences and problems -- technical, logistic, or cultural -- among people working on operational and commercial NLP applications. The workshop will begin a dialogue among researchers to explore issues in technology transfer and the re-use of domain-specific systems. New applications could get leverage from using successful existing NLP technologies. The ability to re-use NLP technology for diverse applications should not only give the application a solid grounding, but should also save time and money. For example, text generation techniques are being used to build prototypes for essay analysis by Educational Testing Service. Other types of NLP technology re-use need to be identified for different applications. Closely related to the re-use of domain-specific technology is the issue of constructing general purpose tools that can be shared by the community, e.g., for tokenization, proper-noun detection, tagging, NP-identification, etc. Another purpose of the workshop is to explore industry-based practicalities that often guide the design of NLP technology. General practicalities that might be discussed are customization and trade-offs between accuracy and other requirements, such as speed, and ease of use. For example, determining the appropriate balance between reporting false positives and false negatives in information retrieval; what depth/breadth of coverage is "enough" in grammar checking; and how can adaptive systems, such as speaker-dependent speech recognizers, train themselves to the user without becoming obtrusive. Discussion of the issues above would help to create connections between both academic and industry-based research efforts to build a solid infrastructure for NLP technology re-use and lead to a deeper understanding of commercial NLP potential. WORKSHOP ORGANIZATION: Presentations will last for 20 minutes, followed by a 10 minute discussion period. Papers will be organized around themes. Ideally, we would like to include the following sessions: 1. Commercial/commercial-bound systems using NLP 2. Software re-use 3. Technology transfer SUBMISSIONS: Authors should submit a full length paper (not exceeding 3,200 words, exclusive of references) and must include a descriptive abstract of about 200 words. Electronic submissions are encouraged and should be submitted as described below. The title page should include title of the paper,names, addresses, e-mail address, telephone and fax number of all authors. Any correspondence will be addressed to the first author. FORMAT FOR SUBMISSION: Papers should be original work. Papers may be submitted either electronically or in hard copy. Electronic or hard-copy submissions must use the ACL submission style (aclsub.sty) retrievable from the ACL LISTSERV server via anonymous ftp: ftp ftp.cs.columbia.edu Name: anonymous Password: <your email address> cd acl-l/ACL97 get aclsub.sty Electronic submissions should be mailed to jbursteinMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue
research.att.com>Apologies if you receive this twice. Note that tutorial speakers are now confirmed. SECOND CALL FOR PAPERS ACL/EACL Workshop INTERACTIVE SPOKEN DIALOG SYSTEMS: Bringing Speech and NLP Together in Real Applications Madrid, Spain July 11 and 12th, 1997 Recent advances in speech technologies, natural language processing, and dialogue modeling have made it possible to build dialogue agents for a wide range of applications from voice dialing to accessing information about the weather, train schedules, cultural events or local restaurants. However, there is little research on the integration of component technologies required for these agents. This workshop will address the challenges involved in this integration: - the special requirements dialogue places on speech recognizers (such as the need for barge-in or substring confidence measures) - the special requirements dialogue places on speech synthesizers (such as producing appropriate prosody or tracking what has been produced so far) - the special requirements dialogue places on natural language generation (such as summarization, aggregration or selecting information subsets) - which ASR and TTS functionalities have the greatest impact on dialogue capabilities - the extent to which the dialogue manager can compensate for limitations of the speech recognizer or the NLP components - what tools exist for building integrated systems and what are the limitations and capabilities of these tools - what role do dialogue tagging schemes and tools for labeling play in the construction of dialogue systems - how to evaluate an integrated system (as opposed to its individual components) - how dialogue systems should be compared with one and other, across both domains and differences in component technologies. We solicit papers from text-to-speech, ASR, NLP, generation and dialogue modeling researchers, and people who are planning to work on interactive spoken dialogue systems as well as those already building integrated spoken dialogue systems. This is a TWO-DAY workshop. The first day of the workshop will start with a morning of 1 hour mini-tutorials (see below) of critical component technologies presented by leaders in the various technology fields. There will be a tutorial on the current state of the art in automatic speech recognition (ASR) by Alex Acero, on text-to-speech (TTS) by Marian Macchi and on dialogue modeling by James Allen. Since we expect to draw participants from diverse research communities, this session will provide a critical common ground for the remainder of the workshop. The remaining one and one-half days of the workshop will consist of presentations by participants on issues identified above and discussion of these issues. We plan to allow sufficient time for open discussion and interaction among the workshop participants. MINI-TUTORIALS: James Allen, University of Rochester: Dialogue Modeling Marian Macchi, Bellcore: Text to Speech Alex Acero, Microsoft Research: Automatic Speech Recognition PROGRAM COMMITTEE: Julia Hirschberg, ATT Labs Research, (co-organizer) Candace Kamm, ATT Labs Research, (co-organizer) Marilyn Walker, ATT Labs Research, (co-organizer) Rolf Carlson, KTH, Sweden Norman Fraser, Vocalis Ltd, Cambridge U.K. Lori Lamel, LIMSI, France Jacques Terken, Institute for Perception Research, The Netherlands SUBMISSIONS: Authors should submit extended abstracts of less than 3000 words. Please include a separate title page with the title of the paper, 200 word summary, 2-5 keywords, names, addresses, e-mail address, telephone and fax number of all authors. Joint submissions with the Concept-to-Speech ACL/EACL workshop are allowed. If there are sufficient joint submissions a joint session may be scheduled. Please indicate on the title page that your abstract is a joint submission. Electronic submissions are encouraged and should be submitted as described below. FORMAT FOR SUBMISSION: Papers may be submitted either electronically or in hard copy. Electronic submissions must be PLAIN ASCII TEXT. Hard-copy submissions should either be plain text or use the Applied ACL final format style (aclapp.sty) retrievable from the ACL LISTSERV server via anonymous ftp: ftp ftp.cs.columbia.edu Name: anonymous Password: <your email address> cd acl-l/Styfiles/Proceedings/aclap.sty.Z get aclap.sty.Z quit uncompress aclap.sty.Z Electronic submissions should be mailed to cakMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue