Document Interrogation: Architecture, Information Extraction and Approximate Answers

We present an architecture for structuring and querying the contents of a set of documents which belong to an organization. The structure is a database which is semi-automatically populated using information extraction techniques. We provide an ontology-based language to interrogate the contents of the documents. The processing of queries in this language can give approximate answers and triggers a mechanism for improving the answers by doing additional information extraction of the textual sources. Individual database items have associated quality metadata which can be used when evaluating the quality of answers. The interaction between information extraction and query processing is a pivotal aspect of this research.