Autonomy First to Fully Automate XML Tagging; Autonomy Agentware Knowledge Server Now Features Automatic XML Tagging Module.

SAN FRANCISCO--(BUSINESS WIRE)--June 2, 1998--Autonomy, Inc. today announced that the Autonomy Agentware Knowledge Server now includes an automatic XML tagging module in addition to its automatic categorization, hyperlinking, data visualization and employee profiling features. Autonomy is the first company to fully automate XML tagging for knowledge management applications.

Autonomy CEO Mike Lynch believes XML standards can be very useful in helping companies turn their vast repositories of word processing documents, email messages, news articles and other archives into a coherent and accessible base of knowledge. "Many knowledge management experts recommend developing a company-wide system of XML tags for specific subjects and categories," Lynch said. "Unfortunately, most of these knowledge management systems depend on the work of 'elves', those expensive helping hands who manually read, sort, and tag documents so they can be stored for retrieval," said Lynch.

Requiring employees to tag their own documents is not the answer, either. "The goal of a comprehensive knowledge management system is to make sure that today's e-mail message or hot memo can help someone who may need it a year down the line," Lynch said. "But put yourself in the shoes of the average employee. You've just written a potentially useful email message. It is 4:30 in the afternoon and you still have 12 items on your 'to-do' list. Are you really going to take the time to look up the appropriate category and insert the relevant tag just in case a colleague in another office might someday find your email message useful?" he asked.

According to Lynch, the cultural barriers are tremendous. "That's why we set out to build a knowledge management system that works as a byproduct of normal business operations."

Agentware Knowledge Server, the core engine of Autonomy's knowledge management solution, uses advanced pattern recognition technology to:

-- Automate the categorization, XML-tagging, hyperlinking and presentation of information, virtually eliminating the need for manual labor in the process.

-- Summarize documents and recommend related articles and documents via hypertext links. Because these links are automatically inserted at the time a document is retrieved, they can include references to documents and articles written long after the original document was published.

-- Provide powerful data visualization technologies that present a unified view of disparate data sources across the enterprise -- including email messages, word processing files, PowerPoint presentations, Excel spreadsheets, PDF files, Lotus Notes archives, intranet file servers, SQL/ODBC databases, live chat/IRC, newsfeeds, and the expertise profiles of other employees. (Visualizer module)

-- Profile each employee's area of expertise based on the issues they research on the intranet, the ideas in the documents and email messages they submit to the system and the topics they follow in their Knowledge Update (another component of Autonomy's knowledge management solution).

The Technology Behind Agentware

Autonomy's Agentware products employ Adaptive Probabilistic Concept Modeling (APCM) algorithms to analyze, sort and cross-reference unstructured data. The method is based on Bayesian statistical probability theorems, Claude Shannon's principles of information theory, and neural networks.

Autonomy's APCM-based software identifies key concepts in text along with the associated frequency and relationship of terms most closely correlated with the idea. Referred to as Concept Agents, the software abstracts can then be used to locate other instances of the pattern of terms and contextual relationships that represent a given concept.

The effectiveness of Concept Agents improves over time as their focus on a specific topic becomes more comprehensive and sophisticated. The agent will actually become skilled at recognizing a relevant topic independently of the language used to describe it, identifying articles on the same topic that might have appeared in a tabloid and a scholarly journal.

The Company

Autonomy develops software that automatically organizes large, unstructured volumes of information into personally relevant channels to help corporate intranets and commercial websites serve the changing needs and interests of users.

The intelligent pattern recognition technology that fuels Autonomy's software was originally developed by Autonomy's sister company, Neurodynamics, for use in British intelligence and defense applications. That technology is at the heart of the British police force's Holmes2 system, named after Sir Arthur Conan Doyle's legendary detective. The Holmes2 system helps the police solve crimes by matching fingerprints or by finding similarities and connections in disparate crime witness statements or police reports. The same technology is now being used to find connections in commercial data.

Autonomy was founded in March, 1996 and remains a privately held company. Autonomy is headquartered in San Francisco, California, with additional offices in Roseland, New Jersey, and Cambridge, England. Its current customers include News Corp., The Royal Mail, Barclays Bank, Virgin, British Telecom, Shell International, Semi-Tech, Macmillan Publishing and Unilever.

CONTACT: Edelman Worldwide for Autonomy

Marivi Lerdo/Tiffany Spencer, 415/433-5381

[e-mail]

[e-mail]