A Distributed Architecture System for Recognizing Textual Entailment

Solving complex problems has become a usual fact in the natural language processing domain, where it is normal to use large information databases like lexicons, semantic relations, dictionaries. This paper describes the steps followed in building the system participating in the RTE3 competition. The system architecture is based on the peer-to-peer model, using the SMB protocol for file transfer. The system functionality was optimized by using a caching mechanism and a quota for synchronizing the termination of all processes. After the peer-to-peer network is configured, one computer becomes the initiator and builds the list of available neighbors. Subsequently, the initiator has the following additional roles: split the initial problem into sub-problems, send the sub-problems to the list of neighbors for solving, receive the partial output files and build the final solution.