|
The Predictive Analytical
Engine
At the core of its technology is the Predictive Analytical EngineTM
(PAETM), a learning system that
categorizes, organizes, and delivers information, drawing connections between
digital information items that were otherwise not apparent. The PAE works
transparently to detect user requests, process them, and enrich the responses
to these requests with Predictive InformationTM.
The Predictive Database
The PAE draws its information resources from the Predictive DatabaseTM,
which holds metadata of information items that are aggregated from internal
enterprise sources as well as external sources such as the Internet. These
sources may already be categorized, coming from an existing Enterprise
taxonomy, the Effective-i's taxonomy that holds popular categorized
information, Web sources or information resulting from employee collaboration.
On the other hand, the sources may be uncategorized; in which case the PAE,
through sophisticated concept extraction algorithms, automatically categorizes
unstructured information into the PAE taxonomy.
The PAE categorization capability is based on concept extraction from documents
using a proprietary algorithm, which is a unique flavor of Bayesian
probability. This algorithm successfully deals with plain and rich text. In the
case of rich text, the algorithm takes into consideration the data with meta
tags in the categorization process. Extraction from HTML / XML based
categorized sources is made feasible by a technology which parses any page with
similar items, either defined in a taxonomy or a result of a search process,
and dynamically creates a set of items in the PAE's internal data
representation. The PAE also includes an API to enable the enterprise's
information managers to manage the Predictive Database.
Predictive Relations
Unique pattern recognition technology, developed by Effective-i, enables the
PAE to analyze accumulated information access behavior, seek common patterns,
and then create Predictive RelationsTM.
The knowledge derived from these Predictive Relations--new connections between
different categories in the PAE taxonomy—is leveraged among other users in
the company.
The PAE is triggered by an event, such as a user reading a document or e-mail.
This event initializes a process that extracts the relevant Predictive
Information from the PAE taxonomy, according to the information item's
classification and the user's personal requirements and access rights.
|
