Linking External Sources to Legal Contracts using Semantic Similarity

This project addresses the problem of automated processing of signed contracts in corporate environment. Theindustrial collaborator, the PACTA company, helps companies keep track of their signed contracts acrossteams, departments, and organizations. The project is focused on the task of identifying external documentsources that may affect the contracts, such as relevant legal documents. Our approach is based on segmentationof documents and contracts, and semantic similarity using web n-gram corpora. This word n-gram basedapproach would be fully unsupervised, and it is extensible to language other than English. This approach isbased on the existing expertise of the Dalhousie research team. The ultimate goal is to accelerate and improvesearch ability to find most appropriate candidate documents that are strongly related to the contract, reducinghuman effort in the process. The project will have a clear and measurable impact on PACTA's functionality andvalue proposition. The collaboration with Dalhousie University will be of benefit to both parties in that it willallow PACTA to deliver a new and innovative functionality to users at a faster pace and Dalhousie Universityto get access to a corpus of commercial agreements for future research and development. The result of theresearch contained in this proposal will enable the company to expand the services offering to its clients, andthereby enhance its competitive position and increase export potential of the services.