PureSight Technology - Active Content Recognition (ACRTM)
PureSight core technology is a sophisticated, proprietary Active Content Recognition (ACRTM) solution. ACRTM is able to identify Web site content and decide whether to allow access to it or not.
Every requested site is inspected by PureSight's intelligent algorithm to ensure its compliance with corporate, institutional, or parental usage policies. Puresight's ACRTM driven solutions have the ability to dynamically analyze every piece of Internet content and accurately determine the context and meaning without the need for cumbersome databases or crude and inaccurate keyword filters. The result is complete and reliable Web coverage with unmatched content recognition and categorization accuracy.
ACRTM filtering components:
Parsing (from Web page to parameters to Raw Data Vector)
Any Web page requested by a user is received packet by packet and sent to the HTML Parser. The parser is the first component of the ACRTM that inspects the incoming data. When packets arrive, the parser breaks down the HTML code into hundreds of parameters, creating a long vector called a Raw Data Vector (RDV). The ACRTM engine recognizes different types of parameters and takes into account the words used in the page and the basic layout and format of the page.
Extracting Features from Raw Data Vector to create a Processed Data Vector
The Raw Data Vector is processed by the PureSight Feature Extractor, which is the first layer of artificial intelligent algorithms. The Feature Extractor reduces the RDV from hundreds of parameters and creates another vector containing 20-25 specific features, called the Processed Data Vector (PDV).
Clustering - Assigning the PDV to a Content Category
The PDV is processed by a second layer of AI algorithms called the Clustering mechanism. The Clustering mechanism is a neural network that analyzes the combinations and relationships in the Processed Data Vector, and extracts a mathematical coordinate for the HTML page.