Intuview Fights Terrorism With Natural Language

Today London Business School hosted the Global Security Challenge finals, where six homeland security startups , selected from hundred of applicants, competed for a grand prize of $500K.

Among the six was Israel‘s Intuview – creator of “artificial intuition” software for real-time exploitation of Arabic language documents to determine contents of intelligence value. I spoke with Intuview’s founder and CEO, Dr. Shmuel Bar about his software contribution to the war against terror.

Since the 9/11 attacks , it was evident to law enforcement and intelligence units that terror organizations require a high volume of communication to carry on an attack. A combination of extremist
writings, exhortations, religious sermons and instructions provide clues and evidence to subsequent actions. However, in reality, there is a big gap between the collection of written intelligence seized in Iraq, Afghanistan and so on to the agencies ability to translate, comprehend it and link it to terrorist intentions.

Intuview’s flagship product is IntuScan, an advanced text analysis system that seamlessly extracts intelligence data from documents written in Arabic. Intuscan provides the law enforcement officers on the field with intelligence reports by extracting entities and hidden meanings in the text. The product uses a combination of a multi-engine approach and manual analysis.

Dr. Bar shies away from the spotlight. He dealt with Islamic terror since the 80s, and it was clear to him that Intelligence organizations are missing a lot of data by not analyzing the documents in Arabic. He explored whether he can develop a set of tools to produce not only categorization of documents but also to analyze what the documents are talking about – is it a personal document? Is it an official memo?

The connection between algorithms and terrorism

Intuview’s technology is based on Natural Language Processing (NLP) in Arabic. The company gathered more than 8,000 documents related to radical Islam and terror organizations, plus a collection 10,000 documents that are irrelevant, serving as a training set. It turns out that all the current NLP tools in Arabic are designed for modern Arabic – so Intuview built NLP for neo-classic Arabic  – the dialect used in most suspect documents. Rather than using the ‘bag of words’ system, IntuScan finds synonyms for each word in the document and then runs all the permutations of the words in a sentence and compares it to verses from the Koran. The technology enables a learning system, in which the matches are used to build a proprietary ontology.

For example,  in a document praising Osama Bin Laden by quoting verses from the Koran and the Haddif– the words themselves are irrelevant. Intuview takes the meaning, which becomes an ontological signature, which the system uses to compare with the data set of high risk documents. IntuScan can also be used to disambiguate Islamic names – using the different permutations. He can tell if the person is Shiite or Sunni for example by looking at the names and expressions.

Advanced Entity Extraction

An officer in the filed doesn’t have the luxury of time. Within minutes of spotting a suspect, he need to decide on one of three courses of action: Red, Amber, Green – shoot him, talk to him, let him go. Therefore, the rangers in the field need a very simple solution.  Intuview supports multiple document types and can process 150 documents in three minutes using a regular laptop. . An average document is 30-40 pages.

Apart from analyzing documents for names and places in Arabic, Farsi, Pashto, Estonian and several other Arabic dialects, IntuScan can identify the time when the document was written by extracting events such as the death of a certain member of the organization or by referencing a milestone mentioned in the news. The chase after bad guys can become difficult at times – the moment the government started checking laptops for counter terrorism, the number of suspects with digital devices that came into JFK went down by 60%.

Business Model

After conducting several pilots with government agencies in America, Europe and Asia, Intuview’s NLP system is currently available for licensing. Clients will be able to purchase a module for each language for approximately ~$40K. The system can be highly customized – the client can select to receive data specifically for a region, such as the south of Iraq or for a certain dialect.

Team

Started in May 2005, Intuview has team of developers and analysts with 30 years of experience in semantics and the Arabic language. To be part of the team, a Doctoral degree is a minimum requirement. The board of directors includes the former head of the Israeli Mossad and a former head of the CIA.

Following a seed round in 2006, the company is now ready to raise capital to increase market penetration. Good luck to Shmuel and the team in making the world a bit safer for us.