Automated Semantic Query Formulation for Document Retrieval

Introduction to the Semantic Web is the chances for easier and effective access to the constantly increasing heterogeneous data on the Web. Currently, the data is able to be retrieved semantically rather than through traditional keyword based searches, which usually return lots of irrelevant information. However, one of the main challenges of the Semantic Web is that data are stored in a structured RDF triple format and are retrieved using complex structured triple represented queries, such as SPARQL, instead of preferred natural language queries and this problem remains subject to research. The proposed AutoSDoR, meaning Automated Semantic Document Retrieval, enables the semantic formulation of natural language queries to structured triple representation based on the machine learning approach in order to retrieve documents from the structured RDF triple format. Additionally the research goes beyond small fragment queries, such as in FREyA to paragraph length query. Automatic disambiguation of query terms that are not covered in WordNet is also proposed, which contributes to the increase in precision and recall of the retrieved document.