There is a positive correlation between news reports on a company's financial outlook and the company's attractiveness as an investment. However, because of the volume of such reports, it is impossible for financial analysts or investors to track and read all of them.

A system that automatically classifies news reports that reflect positively or negatively on a company's financial outlook would greatly benefit analysts and investors. In the application domain of stock portfolio management (see Warren), software agents that evaluate the risks associated with the individual companies of a portfolio should be able to read, classify and weigh electronic news articles, to give investors an indication of the financial outlook of a company.

To accomplish this task, we treat the unsupervised reading and understanding of news articles as an automatic text classification problem. In this project, we are developing an automatic text classifying technique resulting in software agents that we call "Domain Experts." Domain Experts use a sampling algorithm -- based on Weighted Majority algorithm -- that make use of frequently co-occurring phrases as their feature vector.

We call the sampling technique used by Domain Experts "self-confident" sampling. Briefly, "self-confident" sampling is a technique for sampling more promising data from unlabeled data sets to improve a classifier's performance. The "self-confident" sampling is a kind of pseudo labeling method that predicts a label for unlabeled data on the basis of the entropy value of unlabeled data, and the trainer's confidence, which is acquired during training phase.

Instructions for access to TextMiner:

To receive access to TextMiner, please print the the CMU License Agreement (.pdf). Click here to download Adobe Reader.