Arabic Document Indexing for Improved Text Retrieval

Arabic document indexing is a challenging process due to the complex morphological nature of the Arabic language. Methods of document indexing in the literature relied on applying morphological schemes to extract terms. These morphological schemes mainly depend on root extraction and stemming. This paper proposes a simple document indexing method based on selecting only definite words (that have the prefix AL, or it is acceptable to have this prefix). The words that preceding (and/or) succeeding these definite words are also considered. The proposed method is tested using the TREC-2001/2002 Arabic test collection. The proposed method outperforms selecting all terms, either without stemming, or stemmed by the Light10 stemmer, for example, indexing documents by selecting definite words and words that come after them enhances the Mean average Precision of the Light10 by 4.4%, and at the same time decreases the index size by 6.1%.