Users of search engines expect to see a number of input devices and search features to which they have become accustomed over time. When asked about their search preferences, they can usually name a list of specific features. But the question is: Given that specific features are provided, how do users expect them to function, and what are the search results they expect? The conclusion from user tests seems to be that the more freely users can enter their queries the more room there is for misunderstandings of what the search result will be. In particular, we observed the expectation of all (unsigned) query terms being ANDed by default, 'termA +termB' being read as 'termA AND termB', and very often we saw users searching conceptually by entering generics in terms of plurals where they expected to receive specific instances (in TREC-6, for example, with the queries 'Ferry sinkings' or 'Alzheimer's drug treatments' the users expected to find documents on the Estonia sinking event or the Cognex drug).
Search engines basically rely on pattern matching algorithms, looking for documents that contain the submitted query terms. However, especially with free text input, users expect a search engine to "understand" their queries.
The questions for the workshop
are:
Can we meet these expectations in a principled way? Can we derive
some general architecture and user interface design principles
for our search engines from these insights? Which devices can we
provide to augment our pattern matching algorithms on the one
hand, and help the user build effective queries on the other?
Topics may include:
We invite reports on experience with practical implementations of retrieval systems as well as presentations of research in this field. Equal time should be devoted to contributions and discussion of future directions in the design of search applications. Ideally, we can develop a consensus on possible solutions and practical steps to take.
Attendees interested in making a presentation at the workshop should submit an extended abstract of up to 2,000 words or four pages. An appropriate number of contributions will be selected for presentation. Those wishing to attend the workshop without a presentation should briefly state why this workshop is of interest to them. All submissions or requests for further information should be sent via e-mail to Birgit Schmidt-Wesche (bwesche@us.ibm.com) or to Gene Golovchinsky (gene@pal.xerox.com). Final versions should be submitted in HTML for pre-conference publication, and additionally in PDF or Postscript for the printed version of the workshop materials.
Note to authors: We are seeking preliminary or draft papers - you will retain copyright ownership and may submit your paper elsewhere for more formal, subsequent publication.
You will find the papers and further workshop information on our workshop web page at http://www.fxpal.com/SIGIR98/
Workshop chairs:
Birgit Schmidt-Wesche | IBM T.J. Watson Research Center |
PO Box 704, Yorktown
Heights, NY 10598, USA Phone: +1 914 784 6080, fax: +1 914 784 6307 bwesche@us.ibm.com |
|
Gene Golovchinsky | FX Palo Alto Laboratory |
3400 Hillview Ave. Bldg
4, Palo Alto, CA 94304, USA Phone: +1 650 813 7361, fax: +1 650 813 8071 gene@pal.xerox.com http://www.fxpal.xerox.com/people/gene/ |
Submission: | June 1, 1998 |
Notification: | June 15, 1998 |
Camera-ready copy: | July 15, 1998 |
Workshop: | August 28, 1998 |
Information about the full SIGIR program is available through the SIGIR home page. Those considering attending SIGIR or this workshop should use the SIGIR page to register their interest in participation.