Increasing bid expressiveness for effective and balanced e-barter trading

We present a novel knowledge-based approach for automated electronic barter trade systems. An e-barter is basically a closed e-marketplace, where agents may exchange (buy/sell) goods -or equivalent trade dollars- only with other participants to the e-barter. Obviously, in such systems one of the major issues is keeping exchanges as balanced as possible. If the description of goods or services to be exchanged is simple and limited to a well defined set, e.g., oil, wheat, transport, etc., then an exchange based only on price and quantity is enough. But, what if goods or services to be exchanged are described in a complex way? Is it a suitable exchange the one involving mobile phones supporting video streaming with a QWERTY keyboard if the agent is looking for smart phones? Those two descriptions, although very different form a syntactic point of view, are very similar with respect to their meaning (semantics). How could an agent manage and exploit the knowledge on a given domain to deal with such a semantic information and optimize exchanges? We focus on how to find most promising matches, in a many-to-many matchmaking process, between bids (supplies/demands), taking into account not only the price and quantities as in classical barter trade systems, but also a semantic similarity among bid descriptions while keeping exchanges balanced. To this aim we use a logical language to express agent preferences, thereby enhancing bid expressiveness. We also define a logic-based utility function that allows to evaluate the semantic similarity between bids. Finally we illustrate the optimization problem we solve in order to clear the market