Combining Approaches to Information Retrieval

Abstract

The combination of different text representations and search strategies has become a standard technique for improving the effectiveness of information retrieval. Combination, for example, has been studied extensively in the TREC evaluations and is the basis of the "meta-search" engines used on the Web. This paper examines the development of this technique, including both experimental results and the retrieval models that have been proposed as formal frameworks for combination. We show that combining approaches for information retrieval can be modeled as combining the outputs of multiple classifiers based on one or more representations, and that this simple model can provide explanations for many of the experimental results. We also show that this view of combination is very similar to the inference net model, and that a new approach to retrieval based on language models supports combination and can be integrated with the inference net model.

Probabilistic Score Normalization for Rank Aggregation

“…for practical purposes. However, in order to combine the scores, the values should be first made comparable across input systems [2], which usually involves a normalization step [6]. In this poster we propose an aggregation model where the source scores are normalized to a common ideal score distribution, and then merged by a linear combination.…”

Combining Relevance Information in a Synchronous Collaborative Information Retrieval Environment

“…Therefore, we will outline how the RF process can be extended to allow for a weighted combination of multi-user relevance information in a collaborative relevance feedback process. Combination of evidence is an established research problem in IR (Croft, 2002), in our work we are interested in investigating the combination of multi-user relevance information within the relevance feedback process. In our work we use the probabilistic model for retrieval which is both theoretically motivated, and proven to be successful in controlled TREC experiments first shown in (Robertson et al, 1992).…”

“…Combining the outputs from multiple ranking algorithms has become a standard method for improving the performance of IR systems' ranking (Croft, 2002).…”

Enhancing the Effectiveness of Ligand‐Based Virtual Screening Using Data Fusion

“…An early IR investigation by Belkin et al [53] combined the results of multiple text-database searches that had been conducted in response to a single user query but that employed different indexing and searching strategies. Each such strategy yielded a ranking of the text database that was being searched and the set of rankings was then combined using simple arithmetic fusion rules, such as taking the largest rank (MAX), the smallest rank (MIN, as in the Merck study mentioned previously) or the sum of ranks (SUM); this work soon led to many other studies and data fusion is now a standard approach in IR (as reviewed by Croft [54] and by Hsu and Taksa [55]). …”