Automatic indexing using the SLC-II system

摘要

This paper presents an outline of the concepts and facilities of SLC-II, a basic support for R & D in automatic documentation and language translation, and gives a detailed description of its use in automatic indexing. The design and running of a full-scale pilot project on automatic indexing of INIS Atomindex material, done in cooperation with the IAEA and Inspec, will be described, as well as a parallel experiment in the effectiveness of “thesaurus enrichment” by semiautomatic means.The SLC-II system aims to provide the user with a programming language using terms current in his application area, providing him with the maximum possibilities for linguistic analysis, in particular syntactic analysis, while removing from him the problems of data organization, dictionary access, program control and execution and input/output.The first part of the experiment is designed to compare the effectiveness of automatic indexing with “manual” (or ldintellectual”) indexing. In order to minimize the differences between an automatic and a manual system, the terms are chosen from a pre-existent thesaurus, namely the INIS Thesaurus. The data consists of 6 issues of INIS Atomindex. The main difference between the two systems is that the computer only “sees” the abstract while the indexer usually has access to the complete paper.The second part of the experiment aims at comparing automatic indexing using a preexistent thesaurus as it stands, with automatic indexing using the same thesaurus “enriched”. By enrichment we mean the augmentation of the thesaurus with terms selected from the texts being processed.For practical reasons this part of the experiment was restricted to a single subject area, isotope separation. The test collection consists of material from ENDS.The paper was originally presented at the Sixth Cranfield International Conference on Mechanized Information Storage and Retrieval Systems (26–29 July 1977).