2004 WhatsNewAboutSW

From GM-RKB
Jump to navigation Jump to search

Subject Headings: Semantic Web, Natural Language Database Interface

Notes

Cited By

Quotes

Abstract

  • Other Tasks
    • If we take information extraction, e.g. from a report on a type of camera, we can pull out many nuggets, with no definite limit on which. With summarising, say of a stock market report, we can produce a range of quite different, but equally plausible, summaries. It is clear that such information access, or management, tasks are very complex in how they make input-output connections, and very variable in how they determine output preferences. This seems to imply that world modelling and logical inference are the correct base for automation, though there may also be significant use of surface language information, for example in extracting subject-verb-object triples as factoids, or in selecting statements providing answers.
    • "But thus supposing we continue to believe that we do need model and inference, there are lessons to be learnt from work on natural language access to databases. Databases have formal models and database search is logical. But there are a lot of hidden semantics in the guise of user knowledge, as work on informal natural language access (in the 1980s) made very clear. A natural language interface needs a domain model as well as the abstract data model, both to bridge the gap between the user’s approach and the database reality, and to trap the user’s inadvertent excursions outside the bounds of the database. Even simple conversions from relationships between things to attributes of things can require a surprising amount of apparatus: a domain model can be much more substantial than the data model that is required to characterise the database itself. But such a domain model will still be very application-particular, where the SW explicitly requires generality."

References

  • T. Berners-Lee, J. Hendler and O. Lassilia. ‘The Semantic Web’, Scientific American, May 2001.
  • M. Gorman and P.W. Winkler (Eds.) Anglo-American cataloguing rules, 2nd ed, 1988 revision, Chicago, ILL: American Library Association, 1988. http://newsblaster.cs.columbia.edu,


 AuthorvolumeDate ValuetitletypejournaltitleUrldoinoteyear
2004 WhatsNewAboutSWKaren Spärck Jonesnew about the Semantic Web. Some questionshttp://www.acm.org/sigs/sigir/forum/2004D/sparck jones sigirforum 2004d.pdf What's