An exploration of the principles underlying redundancy-based factoid question answering

Abstract

The so-called "redundancy-based" approach to question answering represents a successful strategy for mining answers to factoid questions such as "Who shot Abraham Lincoln?" from the World Wide Web. Through contrastive and ablation experiments with Aranea, a system that has performed well in several TREC QA evaluations, this work examines the underlying assumptions and principles behind redundancy-based techniques. Specifically, we develop two theses: that stable characteristics of data redundancy allow factoid systems to rely on external "black box" components, and that despite embodying a data-driven approach, redundancy-based methods encode a substantial amount of knowledge in the form of heuristics. Overall, this work attempts to address the broader question of "what really matters" and to provide guidance for future researchers.

An Unsupervised Model of Redundancy for Answer Validation

“…A cross-model comparison demonstrates that the Umodel statistically significantly outperforms previous redundancy-based models [3], [16] (as described in Sect. 4.5).…”

“…Approaches based on this observation are called redundancy-based models. Some pioneering studies [3], [4], [16], [22]- [24] have investigated redundancy from the Web for the AV in QA.…”

From document to entity retrieval

“…Moreover, it is concerned with the extraction of appropriate answers from the text content of highly ranked passages or documents (see e.g. Lin, 2007). Both problems are lying outside the scope of the entity ranking task, since entity occurrences are recognized and tagged beforehand and requests on the answer type are assumed to be given explicitly.…”

“…Hence, they rank entities directly by their number of mentions. Others sum up the relevance scores of text fragments that contain string-identical answer candidates (Lin, 2007). Another recent study addresses the issue of answer identity in more detail by incorporating similarity scores between candidate answers in the calculation of the individual answer scores.…”

“…The unweighted indegree IDG represents a pure centrality score depending solely on the structural properties of a vertex. Moreover, the unweighted indegree is equivalent to the so-called voting model often applied in question answering (Lin, 2007) and also in expert finding (Macdonald and Ounis, 2006). The degree of an entity vertex is equal to the number of votes it gets by mentions from document vertices.…”