Applying Arabic Stemming Using Query Expansion

Abstract

The process of conflating different Arabic word formats to their stem or root is called stemming. Indexing text collection using stems or roots has been reported to be superior to using original word formats. However, the Arabic text stemming has negative effects on words. It conflates words with different meaning under one index term. This occurs frequently in Arabic when using stems and it becomes more frequent when choosing roots to index the collection. Furthermore, search engine whose index is constructed using a particular stemmer makes it a stemmer-dependent engine. In this paper, we show how we can still use stemming to reach the same results without indexing the stemmed text. Original words have been indexed, and a stemmer used to extract word variants and add them to the user query. Apart from making our search operation stemmer-independent, we proof that our approach is as good as light stemming and is significantly better than root stemming.