A novel model of selecting high quality pseudo-relevance feedback documents using classification approach for query expansion

In this paper, we propose a new high quality pseudo-relevance feedback documents selection approach that uses machine learning based classifier for selecting a set of good feedback documents for boosting the effectiveness of Query Expansion (QE). Our proposed classification technique utilizes very small amount of labelled data set for training purpose that is very appropriate to select a set of good documents as feedback in our case. Support vector machine classifier is applied for implementing a classifier. Our experimental analysis confirmed that proposed approach improved the effectiveness of QE´s on standard TREC-3 ad-hoc data collection.