2008 | OriginalPaper | Buchkapitel
Improving Expertise Recommender Systems by Odds Ratio
verfasst von : Zhao Ru, Jun Guo, Weiran Xu
Erschienen in: Information Retrieval Technology
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Expertise recommenders that help in tracing expertise rather than documents start to apply some advanced information retrieval techniques. This paper introduces an odds ratio model to model expert entities for expert finding. This model applies odds ratio instead of raw probability to use language modeling techniques. A raw language model that uses prior probability for smoothing has a tendency to boost up “common” experts. In such a model the score of a candidate expert increases as its prior probability increases. Therefore, the system would trend to suggest people who have relatively large prior probabilities but not the real experts. While in the odds ratio model, such a tendency is avoided by applying an inverse ratio of the prior probability to accommodate “common” experts. The experiments on TREC test collections shows the odds ratio model improves the performance remarkably.