Text analytics: versatile and growing

Learning the context of words

Most text analytics software uses statistical, semantic or rule-based analyses or a combination of those. Nathan, a software product from ai-one, is technology for building machine learning applications. It is designed to function the way the human brain learns.

Nathan ingests documents to form a representational model of information using a combination of keywords and associations. The system creates a lightweight ontology that shows the relationships among all words. One of the most common uses is to "fingerprint" documents to find, compare and match texts that have similar ideas. Olin Hyde, VP for business development at ai-one, explains, "The system excels in cases where traditional keywords fail, such as when an idea can be expressed with many different combinations of words."

Described as a "learning platform" because it is designed to assimilate and understand information, Nathan is offered in several forms: NathanAPP as a platform-as-a-service (PaaS) product, NathanNODE for deployment in private clouds, and NathanCORE for embedding in computers and other devices, including smartphones.

"Nathan uses the same type of autonomic parallel processing of neurons that the human brain uses, so it can analyze text more quickly and accurately," Hyde adds. "It learns the context of each word through associations-without any training or human intervention.".

Ai-BrainDocs is an application using NathanAPP to find similar ideas in large collections of documents. ISC Consulting Group has developed a prototype for a cybersecurity case using ai-BrainDocs to thwart information leaks.

"Ai-BrainDocs can go very quickly through thousands of documents at a time," says Dan Squillaro, VP of information technology at ISC. "It lets us understand the content raw documents even if they are not tagged or indexed."