2010 AutomaticallyGeneratingTermFreq

(Murthy et al., 2010) ⇒ Karin Murthy, Tanveer A. Faruquie, L. Venkata Subramaniam, K. Hima Prasad, and Mukesh Mohania. (2010). “Automatically Generating Term-frequency-induced Taxonomies.” In: Proceedings of the ACL 2010 Conference Short Papers.

Subject Headings:

Notes

Cited By

Quotes

Abstract

We propose a novel method to automatically acquire a term-frequency-based taxonomy from a corpus using an unsupervised method. A term-frequency-based taxonomy is useful for application domains where the frequency with which terms occur on their own and in combination with other terms imposes a natural term hierarchy. We highlight an application for our approach and demonstrate its effectiveness and robustness in extracting knowledge from real-world data.

1 Introduction

Taxonomy deduction is an important task to understand and manage information. However, building taxonomies manually for specific domains or data sources is time consuming and expensive. Techniques to automatically deduce a taxonomy in an unsupervised manner are thus indispensable. Automatic deduction of taxonomies consist of two tasks: extracting relevant terms to represent concepts of the taxonomy and discovering relationships between concepts. For unstructured text, the extraction of relevant terms relies on information extraction methods (Etzioni et al., 2005).

2 Related Work

One approach for taxonomy deduction is to use explicit expressions (Iwaska et al., 2000) or lexical and semantic patterns such as is a (Snow et al., 2004), similar usage (Kozareva et al., 2008), synonyms and antonyms (Lin et al., 2003), purpose (Cimiano andWenderoth, 2007), and employed by (Bunescu and Mooney, 2007) to extract and organize terms. The quality of extraction is often controlled using statistical measures (Pantel and Pennacchiotti, 2006) and external resources such as wordnet (Girju et al., 2006). However, there are domains (such as the one introduced in Section 3.2) where the text does not allow the derivation of linguistic relations.

Supervised methods for taxonomy induction provide training instances with global semantic information about concepts (Fleischman and Hovy, 2002) and use bootstrapping to induce new seeds to extract further patterns (Cimiano et al., 2005). Semi-supervised approaches start with known terms belonging to a category, construct context vectors of classified terms, and associate categories to previously unclassified terms depending on the similarity of their context (Tanev and Magnini, 2006). However, providing training data and hand-crafted patterns can be tedious. Moreover in some domains (such as the one presented in Section 3.2) it is not possible to construct a context vector or determine the replacement fit.

Unsupervised methods use clustering of word-context vectors (Lin, 1998), co-occurrence (Yang and Callan, 2008), and conjunction features (Caraballo, 1999) to discover implicit relationships. However, these approaches do not perform well for small corpora. Also, it is difficult to label the obtained clusters which poses challenges for evaluation. To avoid these problems, incremental clustering approaches have been proposed (Yang and Callan, 2009). Recently, lexical entailment has been used where the term is assigned to a category if its occurrence in the corpus can be replaced by the lexicalization of the category (Giuliano and Gliozzo, 2008). In our method, terms are incrementally added to the taxonomy based on their support and context.

Association rule mining (Agrawal and Srikant, 1994) discovers interesting relations between terms, based on the frequency with which terms appear together. However, the amount of patterns generated is often huge and constructing a taxonomy from all the patterns can be challenging. In our approach, we employ similar concepts but make taxonomy construction part of the relationship discovery process.

6 Conclusion

In this paper, we presented a novel approach to generate a taxonomy for data where terms exhibit an inherent frequency-based hierarchy. We showed that term frequency can be used to generate a meaningful taxonomy from address records. The presented approach can also be used to extend an existing taxonomy which is a big advantage for emerging countries where geographical areas evolve continuously.

While we have evaluated our approach on address data, it is applicable to all data sources where the inherent hierarchical structure is encoded in the frequency with which terms appear on their own and together with other terms. Preliminary experiments on real-time analyst’s stock market tips 5 produced a taxonomy of (TV station, Analyst, Affiliation) with decent precision and recall.

References

,

	Author	volume	Date Value	title	type	journal	titleUrl	doi	note	year
2010 AutomaticallyGeneratingTermFreq	Mukesh Mohania Karin Murthy Tanveer A. Faruquie L. Venkata Subramaniam K. Hima Prasad			Automatically Generating Term-frequency-induced Taxonomies