Dissertation Information for Elizabeth DuRoss Liddy NAME: - Elizabeth DuRoss Liddy
- (Alias) Elizabeth D. Liddy - (Alias) Liz Liddy
DEGREE:
- Ph.D.
DISCIPLINE:
- Library and Information Science
SCHOOL:
- Syracuse University (USA) (1988)
ADVISORS: - Jeffrey H. Katzer
COMMITTEE MEMBERS: - Robert Oddy - Susan Monica Bonzi - Joe Grimes
MPACT Status: Fully Complete
Title: The discourse-level structure of natural language texts: An exploratory study of empirical abstracts
Abstract: Discourse linguistics theory suggests that natural language texts communicate in part due to the predictable structure they exhibit. In an exploratory study, two research questions were investigated: (1) Do informative abstracts reporting on empirical work possess a discourse-level structure? (2) Are there lexical clues in empirical abstracts which reveal this discourse-level structure? The first question was addressed by means of four tasks, employing techniques used in cognitive psychology research, to delineate the structure of empirical abstracts based on the internalized notions of twelve professional abstractors. Research Question 2 employed a discourse linguistic approach in the analysis of a sample of 276 empirical abstracts from ERIC and PsycINFO. Based on the results of these investigations, a linguistic model of an empirical abstract was constructed and tested in a two stage validation procedure using sixty-eight abstracts and four abstractors.
Results indicate that expert abstractors do possess an internalized structure of empirical abstracts, whose components and relations were confirmed repeatedly over the four tasks. There appear to be fifteen common components which were freely generated most frequently; rated the highest in terms of typicality; and existed in the strongest semantic relationships with other common components. Substantively the same structure revealed by abstractors was manifested in the sample of abstracts. 344 re-occurring lexical clues which indicated the presence of 92% of the components' presence in the abstracts were identified. Abstractors validated the structure predicted by the model at an average level of 86%.
Results of this exploratory investigation strongly support the notion of a detectable structure in the text-type of empirical abstracts. Such a structure may be of use in a variety of information tasks, such as automatic extracting; question-answering systems; and improving precision of retrieval. The techniques developed for analyzing natural language texts for the purpose of providing more useful representations of their semantic content offer potential for application to other types of natural language texts.
|