A multi-agent system for natural language understanding

Natural language processing (NLP) has many applications such as database user interfaces, machine translation, knowledge acquisition and report abstraction. Several approaches have been used in dealing with NLP. We describe an ongoing research project about understanding natural language input using multiagent system techniques. The system combines between a lexical structural approach and a cognitive structural approach. The lexical structural contains the English vocabulary as agents with all their linguistic information. The cognitive structural consists of 6 modules: speech-to-text, text-to-speech, morphological, semantic, discourse, and query analyzers. These agents are communicating with each other to construct agent subsocieties representing the user input. The query analyzer agent constructs an agent subsociety that communications with the input agent subsocieties to answer the user questions. We start with a brief introduction to multiagent systems and natural language processing. The descriptions of the project, its different modules, and an example of text understanding are presented.