Co-occurrence retrieval: a general framework for lexical distributional similarity

Weeds, Julie and Weir, David (2005) Co-occurrence retrieval: a general framework for lexical distributional similarity. Computational Linguistics, 31 (4). pp. 439-475. ISSN 0891-2017

Full text not available from this repository.

Abstract

Techniques that exploit knowledge of distributional similarity between words have been proposed in many areas of Natural Language Processing. For example, in language modeling, the sparse data problem can be alleviated by estimating the probabilities of unseen co-occurrences of events from the probabilities of seen co-occurrences of similar events. In other applications, distributional similarity is taken to be an approximation to semantic similarity. However, due to the wide range of potential applications and the lack of a strict definition of the concept of distributional similarity, many methods of calculating distributional similarity have been proposed or adopted.In this work, a flexible, parameterized framework for calculating distributional similarity is proposed. Within this framework, the problem of finding distributionally similar words is cast as one of co-occurrence retrieval (CR) for which precision and recall can be measured by analogy with the way they are measured in document retrieval. As will be shown, a number of popular existing measures of distributional similarity are simulated with parameter settings within the CR framework. In this article, the CR framework is then used to systematically investigate three fundamental questions concerning distributional similarity. First, is the relationship of lexical similarity necessarily symmetric, or are there advantages to be gained from considering it as an asymmetric relationship? Second, are some co-occurrences inherently more salient than others in the calculation of distributional similarity? Third, is it necessary to consider the difference in the extent to which each word occurs in each co-occurrence type?Two application-based tasks are used for evaluation: automatic thesaurus generation and pseudo-disambiguation. It is possible to achieve significantly better results on both these tasks by varying the parameters within the CR framework rather than using other existing distributional similarity measures; it will also be shown that any single unparameterized measure is unlikely to be able to do better on both tasks. This is due to an inherent asymmetry in lexical substitutability and therefore also in lexical distributional similarity.

Item Type: Article
Additional Information: Originality: This paper examines measures of lexical distributional similarity, presenting a novel framework that both generalises existing measures, as well as providing the basis for understanding how these measures compare with one another. Rigour: The framework is formally specified, and extensive experimental evaluations are performed. Significance: Measuring the distributional similarity of words is emerging as one of the field's hotest topics, and a substantial number of rather different methods have been adopted, and applied to a wide variety of tasks. The challenge that this paper addresses is a lack of understanding as to which measures are most effective for which tasks. This is the first paper to present a clear basis for understanding the nature of the space of possible measures, providing a methodology for establishing measures that are tailored to particular applications and will outperform the competition. Impact: Total citations in Google Scholar for this paper and its preceding conference paper are 23.
Schools and Departments: School of Engineering and Informatics > Informatics
Depositing User: Julie Weeds
Date Deposited: 06 Feb 2012 19:17
Last Modified: 28 Mar 2012 08:25
URI: http://sro.sussex.ac.uk/id/eprint/19932
📧 Request an update