Document retrieval from digital libraries based on user's query is highly affected by the terms appeared in the query. In many cases, there are some documents in the digital libraries that do not share exactly the same terms with the query, but they are related to the user's need. We address this challenge in this paper by introducing a new subject-based retrieval approach in which, apart from ranking documents based on the terms in the query, a new subject-based scoring scheme is defined between the query and a document. We define this score by introducing a new vector space model in which a vectorized subject-based representation is defined for each document and its keywords, and the terms in the query, as well. We have tested the new subject-based scoring scheme on a database of scientific papers obtained from Web of Science. Our Experimental results show that in 83% of times users prefer the proposed scoring scheme with respect to the classic scoring ones.
展开▼
