Natural Language Processing (NLP)

We are attempting to make NLP more effective by adding AI techniques that are not frequently used in this domain, and by processing in the space of concepts, rather than words.

Although the mapping from words to concepts is ambiguous, the benefits of the presence of structure in the concept domain outweighs any ambiguity problems.

Introducing Concept Strings

Concept strings combine natural language processing (NLP) and text mining techniques to create a new data structure and set of algorithms useful in many text recognition applications.

A concept string represents a piece of text. However a concept string representation also contains all the NLP and concept information that can be gleaned about the text. Concept strings can be compared, stored, indexed referenced and analyzed. Scientio has created algorithms that can index, look up, and compare sections of concept strings in linear time. We're using this technology in chat bots and new recognition engines for many applications.

Recognizing text

Because concept strings hold grammatical and conceptual data about a piece of text, they will match with other strings having the same meaning.

In many security and compliance applications it's necessary to create 'template text' - examples of the kind of text sought in traffic. With concept strings, the 'reach' of each template is dramatically increased, - a template doesn't just match a given piece of text, it matches many other pieces of text with the same meaning.

This reduces the work involved in recognizing conversations containing particular topics, and reduces the risk of missing variations of such text.