Retrieving Best Entry Points in Semi-Structured Documents

Focused structured document retrieval tries to make use of the concept of best entry point (BEP) which is intended to define from a user’s perspective the starting-point from which browsing relevant document components should be optimally initiated [5]. In this paper we describe a simple, efficient and effective method for providing BEPs candidates in XML documents. Experiments conducted within the framework of INEX 2006 evaluation campaign ranked the proposed approach on the 1 place out of 77 official submissions for the Best In Context Task. Secondly we compare the effectiveness of the approach with a standard 'flat' document retrieval system that returns document snippets as BEPs. The experimental results on the Wikipedia collection [3] show that BEPs are considered as useful features from the users points of view.