Dissertation Information for Robert Burgin NAME: - Robert Burgin
- (Alias) Robert Edward Burgin
DEGREE:
- Ph.D.
DISCIPLINE:
- Library and Information Science
SCHOOL:
- University of North Carolina, Chapel Hill (USA) (1990)
ADVISORS: - Martin Dillon
COMMITTEE MEMBERS: - Robert M. Losee - Steve Weiss - Judith Wood - William M. Shaw Jr.
MPACT Status: Fully Complete
Title: Improved syntactic disambiguation in the FASIT automatic indexing environment
Abstract: The development, implementation, and evaluation of an improved disambiguation mechanism within the FASIT automatic indexing system are examined in this dissertation. The enhanced disambiguation mechanism relies on a recursive transition network to improve the indexing representation produced by FASIT by enhancing the system's ability to disambiguate syntactic tags assigned to the words of a text and thereby to improve retrieval results.
FASIT with the recursive transition network was compared with FASIT alone and with single-term indexing on retrieval effectiveness in experiments with two databases. Both versions of FASIT showed noticeable improvement over single-term indexing with a database of articles on cystic fibrosis, and both showed material improvement over single-term indexing with a subset of the Inspec database. However, in neither case did FASIT with the enhanced disambiguation mechanism provide even noticeable improvement over FASIT alone.
The failure of FASIT with the recursive transition network to improve upon the retrieval results of FASIT alone appeared to be due primarily to the interaction of two factors: the inability of the recursive transition network to correctly disambiguate more than 40 per cent of the sentences requiring further disambiguation; and the stepwise reduction of the effectiveness of further disambiguation in the FASIT environment.
A retrieval experiment using a complete manual disambiguation of a sample of documents from the cystic fibrosis database produced no noticeable difference between FASIT alone and FASIT with the manual disambiguation. These results suggest that retrieval performance is not noticeably affected by improvements to FASIT's disambiguation process due to the diminution of the effectiveness of disambiguation caused by FASIT's flexibility in allowing ambiguous tags to be parts of tag combinations selected as index terms.
These experiments provide evidence that little improvement in retrieval performance can be expected, regardless of the degree to which FASIT's disambiguation mechanism alone is enhanced. Therefore, it may be more promising to explore improvements to FASIT apart from enhancements to the disambiguation process, such as changes in the way FASIT selects index terms and the incorporation of linguistic considerations beyond the syntactic information now used by the program.
|