ERIC Number: ED027026
Record Type: RIE
Publication Date: 1967-Jul-1
Reference Count: N/A
Automatic Processing of Metallurgical Abstracts for the Purpose of Information Retrieval. Final Report.
Melton, Jessica S.
Objectives of this project were to develop and test a method for automatically processing the text of abstracts for a document retrieval system. The test corpus consisted of 768 abstracts from the metallurgical section of Chemical Abstracts (CA). The system, based on a subject indexing rational, had two components: (1) a stored dictionary of words classed according to their subject content and (2) a set of stored rules applied to the text on a sentence-by-sentence basis. Testing consisted of comparing the performance of the system to that of the mechanical searching system of the Documentation Service of the American Society for Metals (ASM). It was found that: (1) sentence-by-sentence scanning of the text of CA yields by far too many retrievals and (2) that the structuring of the file does not make an appreciable difference in the search results. Both of these findings are directly relevant to any system of processing text by computer for document retrieval. They also have less direct implications as to the justification of other syntax-like devices, such as role indicators, in a document retrieval system. Appendices include a description of computer operations, selections from the dictionary, and structured sentences answering a sample question. (Author/JB)
Descriptors: Abstracts, Automation, Comparative Analysis, Computer Programs, Dictionaries, Indexing, Information Processing, Information Retrieval, Information Systems, Input Output Analysis, Search Strategies, Sentences, Vocabulary
Clearinghouse for Federal Scientific and Technical Information, Springfield, Va. 22151 (PB-175-965, MF-$0.65, HC-$3.00).
Publication Type: N/A
Education Level: N/A
Sponsor: National Science Foundation, Washington, DC.
Authoring Institution: Case Western Reserve Univ., Cleveland, OH. Center for Documentation and Communication Research.
Identifiers: Chemical Abstracts