Relation discovery by named entity recognition from Tibetan websites

Discovering the significant relations embedded in the Web pages would be very useful for community discovery. In this paper, we propose an unsupervised method for relation discovery from Tibetan Web pages, which is based on co-occurrences of named entities in the pages. In order to find the relation, a rule-based named entity recognition algorithm has been proposed. Our experiment shows that the algorithm has got a high precision and recall by using 30.2 megabyte plain text from three large Tibetan Web sites. And we also give a relation strength formula combining the co-occurrence frequency and personal information, thus the relation in the Web pages can easily be fond according to the value of the relation strength.