A simple probabilistic approach to classification and routing
James Leistensnider, Jon Wilding · 2002
The Language Exploitation Group in Management and Data Systems has developed classification and routing software which calculates the most likely category of a document and creates ranked lists of documents relevant to user-defined categories. The basic algorithm used is to represent categories as multinomial distributions based upon information gathered from training sets of relevant documents, and to use probability to classify and route new documents into these categories. The software has been proven by its entry in the international TREC5 competition, and the software is being extended with both contract and IR and D efforts.