Improving retrieval feedback with multiple term-ranking function combination

Abstract

and VITTORIO GIANNINI WIND Telecomunicazioni S.p.A ________________________________________________________________________In this paper we consider methods for automatic query expansion from top retrieved documents (i.e., retrieval feedback) which make use of various functions for scoring expansion terms within Rocchio's classical reweighting scheme. An analytical comparison shows that the retrieval performance of methods based on distinct term-scoring functions is comparable on the whole query set but considerably differs on single queries, consistent with the fact that the ordered sets of expansion terms suggested for each query by the different functions are largely uncorrelated. Motivated by these findings, we argue that the results of multiple functions can be merged, by analogy with ensembling classifiers, and present a simple combination technique based on the rank values of the suggested terms. The combined retrieval feedback method is effective not only with respect to unexpanded queries but also to any individual method, with notable improvements on the system's precision. Furthermore, the combined method is robust with respect to variation of experimental parameters and it is beneficial even when the same information needs are expressed with shorter queries.

A Survey of Automatic Query Expansion in Information Retrieval

“…Using a combined recall/precision measure, the overwhelming majority of recent experimental studies agree that AQE results in better retrieval effectiveness, with improvements of the order of 10% and larger (e.g., [Mitra et al 1998], [Carpineto et al 2002], [Liu et al 2004], ). Such findings are important to support the claim that AQE is an effective technique, but this may be not sufficient for the cases when we are primarily interested in precision.…”

“…In particular, it has been shown that the ordered sets of expansion terms suggested for each query by the different functions are largely uncorrelated [Carpineto et al 2002]. However, several experiments suggest that the choice of the ranking function does not have a great impact on the overall system performance as long as it is used just to determine a set of terms to be used in the expanded query ( [Salton and Buckley 1990], [Harman 1992], ).…”

A survey of Web clustering engines

“…As there is no clustering algorithm that is intrinsically better or more valid than the others, provided that the structure of the data set is not incompatible with the model of the algorithm [Estivill-Castro 2002], it may be convenient to combine the existing algorithms. Method combination works well for classification [Kittler et al 1998] and query expansion [Carpineto et al 2002], and there is evidence that this is also the case for clustering. Fred and Jain [2005] have shown that merging the results of multiple clustering helps identify clusters with non predefined shapes or structures.…”

SemIndex: Semantic-Aware Inverted Index

“…Clearly, the standard inverted index which only supports exact matching cannot deal with these cases. Various approaches combining different types of data and se- Code R (1977): This CBS adventure series managed to combine elements of "Adam-12", "Emergency" and "Baywatch" at the same time... O 3 Sound of Music, The (1965): Maria had longed to be a nun since she was a young girl, yet when she became old enough discovered that it wasn't at all what she thought... · · · · · · mantic knowledge have been propose to enhance query processing (cf. Related Works).…”

“…Our work can be seen as another alternative to consider the semantic gap by enclosing semantic knowledge directly into an inverted index, so that main tasks can be done before query processing, -User involvement: Most existing works introduce some predefined parameters (heuristics) to rewrite queries such that users are only involved in the query refinement (expansion, filtering, etc.) process after providing the first round of results [3,4,14,20]. In our work, we aim at allowing end-users to write, using the same framework, classical queries but also semantically enriched queries according to their needs.…”