Abstract:

Lattice-based information retrieval methods have the advantage of providing efficient results. However, the general process of such methods is based on a matching step between each concept of the lattice and the considered query in order to identify the most relevant concept. The matching step is based on the comparison of strings which reflects a relatively high run-time cost. This can be seen as a time-consuming task, mainly when dealing with large corpora. In fact, the complexity of a lattice may grow rapidly with the corpus size. To overcome this deficiency, we introduce a new method of lattice-based information retrieval using Formal Concept Analysis (FCA) and prime numbers, and we propose an algorithm, called PNBLSearch, to answer users' queries. We evaluate the reliability of the proposed method using a standard document collection. We have also compared the outcomes provided by our method with those provided by an existing lattice-based information retrieval algorithm [11], called BR-Explorer. Experimental results showed that our method reached better results than the Br-Explorer algorithm.