Clickthrough data has been proposed for numerous uses, and this paper describes how a special form of clickthough data, coselection data, can form non-ambiguous clusters that can then be used to detect semantic similarity between query terms. This semantic similarity assessment can be applied to distinct terms in the same language, giving rise to synonyms, or in different languages, indicating possible translations. It can determine alternative names or descriptors for items which do not occur in traditional thesauri, such as phrases, proper nouns or technical terms. The semantic similarity is calculated without any use of external reference materials and without any analysis of content.
|Cite as: Caon, G., Truran, M. and Ashman, H. (2013). Finding synonyms and other semantically-similar terms from coselection data. In Proc. The Web 2013 (AWC 2013) Adelaide, Australia. CRPIT, 144. Ashman, H., Sheng, Q.Z. and Trotman, A. Eds., ACS. 35-42 |
(local if available)