DOCEX for Homeland Security

IntuView for Homeland Security is a decision-support expert system for document exploitation through real-time summarization and interpretation of domain-specific documents in all supported languages. IntuView instantly assesses any such document or batch of documents, determines whether it/they contain content of a terrorist nature or of intelligence value, provides a first-tier Intelligence Analysis Report of the main requirement-relevant elements in the document and feeds the results into the user’s database for integration with other intelligence data.

 

IntuView is now available for English, French, Spanish, Arabic, Farsi, Urdu, Pashto (limited), Bahasa (Malay and Indonesian), Russian, Czech (limited) and Hebrew. We are planning, subject to client demand, on Turkish.

 

One installation of the IntuView machine can process ±1 million short texts per day. The main purpose is to provide the intelligence and Counter-terrorism community with an ability to understand radical Islamic and terrorist-related texts without the need to have well versed Arabic literate analysts on hand. IntuView does not “translate” but “interprets” – provides the meaning of the document that lies beyond the prima facie statements in the text. The system is based on our experience that translation – human and automated – does not really provide a satisfactory solution. Translators who are not casing officers with intimate knowledge of the religious, social, organizational and political milieu of the documents will inevitably output large amounts of irrelevant material (false positives) putting a strain on the analytical team, and overlook key information. Ultimately, delays in the process of intelligence generation frequently may bear a cost in human lives. I am adding this since I believe that in your capacity, this tool may be of interest for scanning information on the Web or on databases for relevant terrorist financing information.

 

The output of IntuView™ is structured data in three forms:

An English language report with the look and feel of a human-authored document

Data sent to a local database which can be examined later or migrated to other applications and integrated into other databases.

Statistical representation of different features of the batch being analyzed so that the user can see at any given time the predominant character of the material being processed.

 

The information that IntuScan extracts from the documents include:

  • Metadata of the document as it appears in the digital information on the file.

  • Information derived from the “understanding” of the text regarding the authorship of the document:

  • Affiliation of the document to a Sect (Sunni/Shiite), ideological stream (Mainstream Sufi, Jihadist, etc.) and organization (e.g. al-Qaeda).

  • Probable date of writing or a period during which the document was probably written

  • Risk score or relevancy score of the document configurable to the user’s requirements.

  • Key topics dealt with in the document.

  • Apparent positive or hostile attitudes of the text towards persons or groups (e.g. “anti-American; anti-Shiite; pro-ISIS)

  • Abstract of the main issues treated in the document with breakdown For example, if the document supports suicide attacks, the summary will indicate that and will provide the theological basis that it presents for that support.

  • References to religious texts with links to the probable interpretation accorded to them by the author of the text (for example: if the text is an al-Qaeda document and it quotes a verse in the Koran or Hadith which is commonly interpreted by the organization as having a given hermeneutics, the summary will identify it as such.

  • Names of known persons mentioned in the text who appear in the IntuScan™ database (or in the user’s database). These names are identified not only in their full forms as they appear in the database but in alternative forms which are automatically derived using sophisticated algorithms for name analysis.

  • Entity extraction of names of people, places and organizations including “contextual” information derived from the text (e.g. the title of a person, references to his being alive or dead, traditional statements of high regard in a document whose author is identified as belonging to a certain political camp provide the system information on such a person).

  • Identification whether a document refers to “recipes” for building explosives or chemical-based weapons. The algorithms infer from the materials, the procedures, safety measures and proportions of chemical substances what explosive substance the recipe may intend to create. The algorithms also infer what use may the product be put to (obviously if the amount or proportions indicate a high sensitivity explosive, there would be constraints in moving it, amounts may indicate usage in vests, car bombs or for closed areas such as personal assassination or aircraft.