We extended tree similarity based match-making from the buyer/seller situation to a scenario of bilateral bartering and multi-agent ring bartering. Our work is built on top of the AgentMatcher tree similarity algorithm for node-labelled, arc-labelled, arc-weighted trees. A representation of these trees in a multi-dimensional space was developed to allow efficient indexing and pruning in large tree databases. The concept of risk was introduced to control the process of bartering ring construction. We have tested our system on the Teclantic.ca portal, where it allows researchers and companies from Atlantic Canada to share technologies as well as to be contacted by investors.
Match-Making for Bartering Scenarios (MCS Thesis Sebastien Mathieu)
Match-Making for Bartering Scenarios (MCS Thesis Sebastien Mathieu) - PowerPoint presentation