Discovering knowledge from large amount of textual data is an important problem. Especially, application of association rule mining to textual data has been studied excessively. Many works has successfully found relationships between words that reflects syntactical rules, co-occurences, or phrases. These rules are useful for understanding the linguistic nature, but in real life, the relationships between the topics or contents are important and useful, such as what kind of topic tends to appear in same paper or books. Our objective is to find relationships between contexts or topics. In this paper, we propose an approach to use passages to take in some level of semantics in rule mining. We show some preliminary results to show its potential and give discussions on the problem for further improvement.
展开▼