Revealing Topic-based Relationship Among Documents using Association Rule Mining.
Kritsada Sriphaew, Thanaruk Theeramunkong · 2005
With a large volume of electronic documents, find-ing documents the contents of which are same or simi-lar in their topics has recently become a crucial aspect in textual data mining. Towards revealing so-called topic-based relationship among the documents, this paper pro-poses a method to exploit co-occurring unigrams and bi-grams among documents to extract a set of topically simi-lar documents with association rule mining techniques. To evaluate effectiveness of the method, a collection of well-organized scientific research publications is employed. The experimental result indicates that any two documents with referential links can be found with the accuracy of 60-80% in the case of unigrams, and 80-90 % in the case of bigrams. An analysis of discovered association rules is also given. KEY WORDS association rule mining, data mining, knowledge dis-covery, document, topic-based, relationship. 1