A Feasibility Study of Automatic Indexing and Information Retrieval

Comparison was made between the natural language words and phrases contained in titles and full abstract text and the indexer-assigned, thesaurus-controlled, descriptors used to define information contained in articles abstracted for inclusion in Petroleum Abstracts. Exact as well as word synonym and phrase concept synonym "matches" were recorded. The study indicates that for the system as a whole only 40 percent of the total assigned descriptors was contained in the title and text of the abstract. This amount of the optimum information needed for retrieval is considered to be an unacceptable level of recovery.