Oracle Systems Corp has announced that Context, its core technology for unstructured text analysis, will be integrated into all its existing products by the end of the year. Context is a natural language processor designed to understand the content of documents (CI No 2,424, 2,447). Oracle’s workgroup products marketing manager Jenny Edmonson says the product will be used to develop programs that sift through information, picking up only the relevant material for the user, and sees the technology being attractive to people in the fields of market research, news-wire monitoring, knowledge bases and patent research. The system works by accessing a 600,000-word dictionary with up to 1,000 pieces of linguistic information a word. The program can pick up on the relationships between words to produce a bias towards different meanings and could recognise the difference between ‘fence’, meaning a divider between two sections of land, and the same word used as a verb to describe the divestment of stolen goods. The technology analyses a document at five levels: grammar, theme, syntax, connotation and discourse, before feeding the results through to the application. Using these data collection layers, Context obtains a view of the rise and fall of themes and subjects. Using a slide-bar metaphor, users can reduce blocks of text from verbosity down to key words, and can access various levels in between. Oracle is setting up a third party programme to develop applications using Context, and is working with outsiders to develop lexicons for specific markets such as the pharmaceutical and legal sectors. Context is to be included in three new product releases by year-end: those of SQL Text Retrieval, to be renamed Context Server, in Oracle Office for intelligent sorting of mail messages, and in the Media Server product, for evolving video on demand into newspapers on demand.
