Cognal’s Cognitive Natural Language technology uniquely combines elements from four different scientific disciplines: Semantics/Ontology, Natural Language Understanding, automatic Query Generation, and Cognal’s patent-pending “Teach the Computer™” Machine Learning. This unique combination of innovative technology provides the basis for Cognal’s disruptive methodology.

Technology Highlights

Cognal’s NLQuery Technology combines four technology disciplines

  • Ontology / Semantics
  • Natural Language Understanding
  • Automatic SQL Generation
  • Teach the Computer Machine Learning

Teach the Computer™

The most novel part of the company’s Cognitive Natural Language Technology is a patent pending machine-learning feature, “Teach the Computer”

Cognal Labs provides an end user “Cognal Studio” for admins to use to correct a user’s request that failed to get an answer (or returned a “wrong answer”), Admins can test the corrected query manually or through an NL Request to see that it produces the correct result, and save the corrected query (along with the way the user asked for it originally). Cognal’s technology employs a patent-pending “Semantic Equivalence” as its machine-learning mechanism to achieve ever-more accurate answers.

The age-old problem of Natural Language products is the fact that there are many ways to ask for the same information. Cognal’s technology addresses that issue head-on by capturing every failed NL Request and then providing a cadre of trained end-users a method of “teaching the computer” the meaning of the failed query through the company’s patent-pending Semantic Equivalence methodology.

Below are some example OntoloNet ontologies that might contain “Cognalized elements” used in Cognalizing data sources

  • Common Terms Model
  • General Business Model
  • Agency Model
  • Auto Insurance Model
  • Classified Ads Model
  • eCommerce Model
  • Sports Model
  • Major League Baseball Model

Measuring ‘Natural Language Understanding’ (NLU) for Data Search

The Achilles Heel of NLU development is that there are many ways to say the same thing. So the best metric for grading NLU products is the ratio of “successful answers” (i.e. exact facts returned) to “number of requests”. “Hit or miss” products that have a poor ratio for database search (less than 40%), like all key-word search products and also like siri, will not be used for getting exact facts from databases with their current search algorithms, although of course these non-database search products will remain popular for giving valuable information in unstructured search or general document search.

Cognal Labs’ Cognitive Natural Language Understanding methodology includes a cognitive “Teach the Computer” feature that allows a power user to correct a user’s information request by using the company’s open source NLStudio product to manually drag-drop concept elements from a visible tree that will be guaranteed to produce the correct answer, and then save that “corrected” request for the next time a user says something similar.