Software Architectures for Natural language Processing

  Speaker 

  :

Rajeev Sangal, Ph.D 1980,University of Pennsylvania
Director
Language Technology Research Centre
International Institute of Information Technology – Hyderabad

Research Area Includes:

Natural Language Processing,Artificial Intelligence, Machine Translation and Speech Processing .

  Abstract

  :

New software architectures are needed for building systems in which modules or sub-system provide partial answers, and at times may not succeed in producing even that. For example, a natural language sentence analyzer may fail to produce a parse tree for a sentence. However, the applications or the subsequent modules which use the output, should continue to work albeit with reduced accuracy or effectiveness.

In this talk, we will present a software architecture together with a data representation which can deal with partial information. A uniform data representation for natural language processing which allows partial analysis to be represented at multiple levels is used by the modules. When a module does not find requisite information at a given level, it operates at a lower less detailed level of analysis. A software architecture which puts it all together with a tool called Dashboard to support the architecture will also be outlined.
http://www.iiit.ac.in/%7Esangal/