The ontology of such a system contains morphological, syntactic and pragmatic rules for handling words, as well as the ontology that contains semantics. The goal is to enable the user to communicate with the system using free text.

Consider as an example an e-commerce application where a sentence entered by the website visitor contains a request to book a hotel. An agent is assigned to each candidate meaning for each word of this sentence and these agents will negotiate with each other until they agree the precise meaning of the text, within the given context.

The agents negotiate to maximise the sum of notional money available to them for obtaining the correct meaning of the enquiry as well as by earning commission by booking a room. Each word 'knows' its contribution and pursues financial gains so that the system behaves as a virtual market. In this market, word meanings search for each other, compare their characteristics as 'sellers' and 'customers' and pay for services rendered. For example, when in the search for meaning, several agents of meaning compete, and their chances seem to be equal, it is the wealthiest word meaning that wins by being able to pay the asking price with ease.

The wealth of a word meaning accumulates through the frequency of its occurrence. If the user has given a particular meaning to a word many times before, and therefore that meaning has won many previous competitions, it becomes wealthy.

If a word is misunderstood, the wrongly assigned meaning loses its commission (or even pays penalties). In time, wrongly attributed word meanings will not be able to compete with other word meanings and will loose their credibility. As the opponent meanings of words improve or loose their 'scores' (commission earned), the system evolves and adapts itself to each particular user or group of users.

Semantic ontology network (fragment)	Morphology stage of analysis
Syntax stage	Semantics

Constructed scene	Pragmatics

Back to MagentA Products

MagentA Home