Evaluation of the Document Categorization in "Fixed-point Observatory"

“Fixed-point observatory” is a prototype to support users to grasp recent trends in the fields of their interest from large-scale information. It consists of content-based categorizer, named-entity-based categorizer and multiple-document summarizer. We have evaluated the content-based categorizer, which adopts the simple “bag-of-words” model. Though the quality seems be sufficient for rough classification, it might be improved to use the categorizer in other applications.