Quasi-Synchronous Dependence Model for Information Retrieval

Incorporating syntactic features in a retrieval model has had very limited success in the past, with the exception of term dependencies. This paper presents a new term dependency modeling approach based on a dependency parsing technique used for both queries and documents. Our model is inspired by a quasi-synchronous stochastic process for machine translation [21]. It describes four different types of syntactic relationships between dependent terms and allows inexact matching between documents and queries to deal with possible syntactic transformations. We also propose a machine learning technique for predicting optimal parameter settings for a retrieval model incorporating the syntactic relationships. The results on TREC collections show that the quasi-synchronous dependence model can improve retrieval performance and outperform a strong state-of-art baseline when we use predicted optimal parameters.