Home // International Journal On Advances in Internet Technology, volume 6, numbers 1 and 2, 2013 // View article

Entity Ranking as a Search Engine Front-End

Authors:
Alexandros Komninos
Avi Arampatzis

Keywords: Web entity ranking, entity search, information retrieval, threshold optimization.

Abstract:
— In this paper, we present a Web application for entity ranking. The application accepts as input a query in natural language and outputs a list of the most relevant entities according to the query. The system uses Web documents as data and performs extraction, formatting and ranking of entities in real time. An experiment is conducted to determine the most efficient ranking method among eleven alternatives. The experiment suggests that the total frequency of an entity in a retrieved set of documents has less to say on the entity's relevance than the number of retrieved documents it occurs in. Furthermore, for small retrieved sets such as the top-10, document rank information seems to play a little role. Four algorithms are tested for estimating the correct amount of results in the ranked list and provide a threshold. The best results are achieved by the maximum entropy algorithm applied to the distribution of scores provided by a multiplicative combination of logarithmic entity frequency and document frequency.

Pages: 68 to 78

Copyright: Copyright (c) to authors, 2013. Used with permission.

Publication date: June 30, 2013

Published in: journal

ISSN: 1942-2652