Grant: Collaborative Research and Development Grants

Given the pivotal role of structured knowledge extraction from the Web for SideBuy, this proposed project will address the challenge by exploring the development of tools and techniques for harvesting entities and relationships from Web sources. Our work will concentrate on formulating a hybrid framework that combines symbolic (semantics-driven) and statistical (data-driven) techniques for extracting information from the Web. In our approach, information will be extracted based on a well-defined covering ontology that is properly described and extensively annotated with domain knowledge. This ontology will serve as the backbone of the symbolic computation aspect of our framework, whose structure and semantics will guide the information extraction process. The process will also be augmented with statistical techniques that mine patterns and relations within Web content in order to derive useful information. The main goal of our work will be to develop semi-supervised techniques that are scalable to Web. The focus on a hybrid approach for information extraction that involves both symbolic and statistical computations is the strong research focus of this proposal.