Categorizer performs automated text analysis and categorization. Categorizer configured as a stand-alone product that perform email categorization interfacing the enterprise email server or as a module for text analysis that is interfaced with organizational professional applications.

System layout:

PIC5

Categorizer is based on Machine Learning and other language dependent semantic technologies. Categorizer recognition level is 90% – 95%. Once Categorizer isn’t sure that it recognizes the right category (5% of the cases) it transfers the message for manual categorization.

Domain Semantic Feature is a set of dictionaries (corpus) that is assigned to the domain of the organization (telecom, banking, education, etc). This module is supplied with the categorizer, and it serves as the basic corpus for the training module.

Training Module is a module that the system administrator has to run before operations in order to train the system with the type of text, sentences or emails (hereafter “Sentences”) that are typical to the organization. After training the system should be able to analyze sentences from the organization domain.

Retraining Module. The system admin can classify manually Sentences that weren’t classified properly by the system during operations and he can add those Sentences to the system Corpus. This activity that is done on the first months of the system operations improves even more the system accuracy level.

Supported Languages: English, Hebrew

Interfaces: The system includes an interface to MS Exchange system. Should an interface with other systems is required the customer can use Uniport web service format to develop the interface.