A conceptual indexing approach for Arabic texts

Classical IR systems are often based on lexical matching using approaches that rely on purely statistical methods founded on distributions of keywords to calculate the similarity between the query and the documents of the corpus. The relevance of a document according to a query is based on the similarity of vocabulary and not according to the thematic similarity between both. Indeed, a document selected by the IRS relevant shares the same words with the query. Among the research that tried to overcome the limitations of conventional IR, there is an approach called (Concept-Based Information Retrieval) that is characterized by the notion of conceptual space in which documents and queries are represented by opposition to the space of simple words used in conventional IR. The use of external knowledge such as WordNet was the subject of several works in IR, often geared towards improving the precision (WSD: Word-Sense Disambiguation, Conceptual Indexing) and/or improving the recall (query expansion).