Automated categorization of real-time newswire stories. Hooked on lexiconics: how I taught my Sun to read

Modern international trade activities rely heavily on thousands of daily information artifacts reporting the state of the world's trading blocks. These information artifacts often require handling by human operators for indexing, sorting, and categorization. Intervention by human operators costs precious hours in the dissemination of these artifacts to end users. The paper describes the information recognition capability that the California Agricultural Technology Institute (CATI) developed as part of its Advanced Technologies Information Network (ATI-Net). The capability includes software using statistical analysis of previously human-recognized documents in order to seed information recognition databases. The recognition databases are used by automated recognition software to classify and store information artifacts without human intervention. This software is discussed with reference to two ATI-Net projects: automatic storage of newspaper articles into categories of interest to the public, and the assignment of Department of Commerce industry codes to international trade lead reports. Each of these projects takes advantage of several years of previously human-recognized information artifacts in order to automate the recognition process.