| The research described here accomplished several goals. First, we built a microcomputer based Arabic bibliographic Information Retrieval System (Arabic-IRS), that interprets queries and retrieves relevant abstracts, with a reasonable response time. This system will be used in the Library at Mut'ah University. Second, we repeated Al-Kharashi's (1991) experiments with consistent results. We found that the system functions better with roots than with stems as index terms and better with both roots and stems than with words. Third, we demonstrated that using abstracts gives superior results no matter what index terms are used. Fourth, we discovered that a relational thesaurus used interactively gives the same good results as using roots as index terms.; The system is designed to use the thesaurus interactively. The user is shown words related to the query and asked to press the Enter Key to add those desired. The chosen words are then added to the query. The results show that the architecture design described for Arabic-IRS is effective and it runs well on a microcomputer.; Testing the system involves several steps (a) Preparing 120 Arabic documents with abstracts to be used as the database for the system, (b) Accessing and processing Arabic and English text, (c) Collecting 32 queries from experts, (d) Asking students of Computer Science to make Relevance Judgments, (e) Indexing and sorting Arabic terms, and (f) Retrieving documents using different Boolean operators.; Standard measures of recall and precision were used in the evaluation process. Tests of significance were carried out using nonparametric statistical methods, the Signed Pair, and the Wilcoxon tests.; We conclude that certain standard information retrieval techniques are valid for Arabic as well as for English, but that simple suffix-chopping is not appropriate because of the complex nature of Arabic morphology and to obtain good results it is necessary to use a thesaurus or complex morphological analysis. |
