Semantic Query Structuring to Enhance Precision of an Information Retrieval System: Application to the Medical Domain
Abstract
Most Information retrieval systems represent a query, also a document, as a bag of indexing terms without any relation between each other. This bag-based representation causes a problem for specialists when they deal with a specific domain like medical one. We present an alternative to the bag of indexing terms representation depending on semantic query structuring, in order to fulfill this need of precision in a specific domain. This structure of a query is obtained by grouping indexing terms using pre-defined categories called dimensions. These dimensions represent the different aspects that could appear in a query or a document. By using this notion, the relevant document to a given query should not only has a maximum number of shared indexing terms but also have a similar structure. Experimental results show precision improvement related to the granularity of dimensions and its distribution over the whole corpus.