THE ASH PROJECT

Among several projects in progress, ThinkinGolem is employing resources on the ASH Project, devoted to the development of and advanced natural language processing system useful as a core module for several applications, such as text mining and synthetic characters.

ASH (standing for "Activation-based Semantic Hyperspace") will represents an improvement in the field of natural language processing technology. Despite to a strong distinction between statistical processing techniques and formal logic-based intelligent techniques, the ASH technology is based on a different point of view. It arises from the integration of different Cognitive Psychology models such as the Hyperspace Analogue to Language (HAL) developed by Burgess and Lund at the end of 90s to provide a model for linguistic cognitive processing, and the Spreading Activation Theory developed by Collins and Loftus around the half of 70s to provide a theory for conceptual processing in Semantic Networks. The particular integration of those models provides a framework for supporting a true conceptual linguistic processing, strongly based on the context in which language symbols are used. Markovian statistical techniques to process and analyze a lexicon with an opportune filtering of input/output messages, plus pattern recognition and reinforcement learning techniques derived from the Artificial Intelligence domain, reinforce further that cognitive substrate yelding it very powerful.

A good feature of the ASH technology is its partially independence from the semantic of linguistic "platform" on which base the natural language processing takes place. This means that the system is able to learn and represent the conceptual framework of any language simply analyzing a typical conversational lexicon. That kind of technology can be used as a core natural language processing module for interesting scientific and commercial applications, like the following examples:

  • intelligent text mining (for Web search agents, intelligent search engines);
  • natural language human-machine interaction (intelligent interfaces, intelligent agents, personal digital assistant, chatterbot);
  • story-telling (narrative intelligence, documentation integration, automatic reporting by text mining, Web avatars);
  • contextual information storing (information systems, knowbots).

The ASH model, resulting from the particular integration of memory models derived from the Cognitive Psychology domain, statistical analysis techniques, and Artificial Intelligence models, represents one of the main technology developed by ThinkinGolem on its own. At the present time, the first Research & Development phase is concluded, and an online demo to test the ASH system will be available on this site as soon as possible.

This project on multidimensional representation of language, that is currently at the first stadium of development, was presented officially for the first time at the Master in Sources and Methods for Social Research, Faculty of Statistics, University of Rome "La Sapienza" gaining great consents. It is possible to download a copy of the seminar presentation in MS Power PointTM format.

If you are interested in customisation and/or commercialisation of ASH, please contact us.

BACK