Date: Friday, 21-Nov-97 04:25:59 GMT Server: NCSA/1.3 MIME-version: 1.0 Content-type: text/html Last-modified: Friday, 16-Feb-96 18:12:00 GMT Content-length: 5552 DMS - Hawkeye



Timely news delivered quickly in electronic form is a critical resource in today's business. But sifting through the flood of information can be a daunting task. Even with both automatic and manual profiling, the real-time routing of news to users is often out of control. The problems are all too familiar.

Profiles for screening news are hard to set up and hard to calibrate to get an acceptable balance between not missing anything and not having to look at everything.


If you don't want to become roadkill on the Information Superhighway, you need to monitor and control your critical information resources with a tool that:
HAWKEYE is an automated tool that analyzes and classifies text information from electronic message streams. It is set apart from other text information products because it operates in real time, handling high volumes of dynamic data with unknown characteristics.

It employs technology unlike that of conventional text retrieval systems, which typically handle static archived data. HAWKEYE reads message files from an input source, checks for duplicates, classifies them by context, extracts names and other indexing terms, looks for natural groupings in the messages, maintains a moving window on the messages for review, and writes out messages with routing tags to an output directory.

It also checks constantly on its own operation so that it can automatically warn you of unexpected conditions.



FEATURES

HAWKEYE provides these unique features:

SOPHISTICATED TECHNOLOGY IN A SYSTEM THAT'S EASY TO USE

HAWKEYE uses a technology based on N-Grams, which are simply word fragments from two to four characters long. Messages containing similar N-Grams will be related in content. N-Grams are unique among document indexing technologies in that they can be validated rigorously through controlled experiments with all kinds of text data.

For each message it processes, HAWKEYE computes a vector of its N-Gram frequencies as a statistical description. HAWKEYE then creates a profile by assigning numerical weights of importance to selected N-Grams. The match of a profile with a message can be measured by adding the products of the frequencies of message N-Grams times their profile weights.

You can build a profile simply by selecting a sample of the kind of text you want to match. HAWKEYE's unique method of setting match thresholds for profiles is always consistent and easy to understand. Unlike other profiling systems, the match thresholds are not arbitrary numbers you have to set by trial and error.

You can observe and direct the operation of HAWKEYE from a "drag-and- drop" graphical interface requiring little keyboard entry. HAWKEYE avoids "queries" in the usual sense. Instead, you produce statistical profiles of target information simply by providing examples of what to look for in incoming data.

For additional information on HAWKEYE's innovative technology, contact us at dmsweb@partech.com


PAR Home 
Page PMC Home RRC Home PGSC Home PVSC Home DMS Home Webmaster Email