Using Text Clustering for Intelligence Classification

In this paper, we discuss how text mining methods could be used in a mixedinitiative interaction approach to intelligence analysis. We describe how simple methods from text mining can be used to help intelligence analysts determine where a specific report or analysis fits into the knowledge base (KB), i.e., how it should be classified and which, if any, other documents in the KB it should be linked to. The method works by comparing the vector space model representation of the new information document with those of all documents previously stored in the knowledge base. Those documents that are sufficiently similar to the new piece of information are displayed to the user, who can then choose to place links between them. Using a computer tool such as the one suggested here potentially allows the analyst to spend more time analyzing intelligence reports rather than searching for and classifying them. In previous work, we have discussed how the MilWiki, an improved implementation of the open-source MediaWiki system, could be used as a knowledge base for military purposes. To illustrate the text classification method described in this paper, it has been implemented for MilWiki. To simulate new pieces of information, the prototype allows the user to download articles from the Wikipedia. The method, as well as the collaborative work process used in a wiki, could be implemented in any content management systems. In addition to describing the text classification method, we also give a brief introduction to text mining and the vector space model of documents.