Affiliations: Central Research Laboratory, Hitachi, Ltd., 1-280,
Higashi-Koigakubo, Kokubunji, Tokyo 185- 8601, Japan, Tel.: +81 42 323 1111,
Fax: +81 42 327 7700, E-mail: tabata@crl.hitachi.co.jp | Department of Knowledge Engineering and Computer
Sciences, Doshisha University, Kyotanabe, Kyoto 610-0313, Japan, Tel.: +81 774
65 6297, Fax: +81 774 65 6821, E-mail: smitsumo@mail.doshisha.ac.jp
Abstract: This paper presents the conceptual framework for a new kind of
information acquisition system and the required enabling technologies. This
system is for information probing - as opposed to information
retrieval - and is intended to allow the acquisition of information without
requiring the assumption of a commonality of interpretation between authors and
users. The concept of the skeleton was devised as an
abstraction of the relations between terms and provides a model of the
assertions made by authors in documents and in glossaries. This model has
produced four technological results. The first is a scheme for the
assertion-based acquisition of information. The assertions of authors and of
users are represented as respective sets of defined_by relations; document
information is then selectively acquired on the basis of measures of proximity
that are used to compare the assertions. In contrast to traditional methods of
retrieval, which rely on the frequency of occurrence of specific terms, our
approach uses assertions as clues in the search for information. The second is
the provision of an automatically generated form of metadata. This has a
logical structure that reflects the assertions made by authors in documents,
and allows the user-centric assignment of descriptors. The third is the dynamic
acquisition of term knowledge which is tailored to individual users. Glossary
skeletons are generated automatically from users glossaries and
constitute bodies of declarative knowledge with regard to the definitions and
usage of terms. This knowledge is used in formulating the queries used to probe
for information. The fourth is a system architecture in which data structures
are dynamically evolved on the basis of user-specific perspectives.