Tsui, E.; Wang, W.M.; Cheung, C.F.; Lau, A.S.M.: ¬A concept-relationship acquisition and inference approach for hierarchical taxonomy construction from tags (2010)
0.02
0.01799386 = product of:
0.03598772 = sum of:
0.03598772 = product of:
0.07197544 = sum of:
0.07197544 = weight(_text_:core in 4220) [ClassicSimilarity], result of:
0.07197544 = score(doc=4220,freq=2.0), product of:
0.25797358 = queryWeight, product of:
5.0504966 = idf(docFreq=769, maxDocs=44218)
0.051078856 = queryNorm
0.27900314 = fieldWeight in 4220, product of:
1.4142135 = tf(freq=2.0), with freq of:
2.0 = termFreq=2.0
5.0504966 = idf(docFreq=769, maxDocs=44218)
0.0390625 = fieldNorm(doc=4220)
0.5 = coord(1/2)
0.5 = coord(1/2)
- Abstract
- Taxonomy construction is a resource-demanding, top-down, and time consuming effort. It does not always cater for the prevailing context of the captured information. This paper proposes a novel approach to automatically convert tags into a hierarchical taxonomy. Folksonomy describes the process by which many users add metadata in the form of keywords or tags to shared content. Using folksonomy as a knowledge source for nominating tags, the proposed method first converts the tags into a hierarchy. This serves to harness a core set of taxonomy terms; the generated hierarchical structure facilitates users' information navigation behavior and permits personalizations. Newly acquired tags are then progressively integrated into a taxonomy in a largely automated way to complete the taxonomy creation process. Common taxonomy construction techniques are based on 3 main approaches: clustering, lexico-syntactic pattern matching, and automatic acquisition from machine-readable dictionaries. In contrast to these prevailing approaches, this paper proposes a taxonomy construction analysis based on heuristic rules and deep syntactic analysis. The proposed method requires only a relatively small corpus to create a preliminary taxonomy. The approach has been evaluated using an expert-defined taxonomy in the environmental protection domain and encouraging results were yielded.
