An Approach to Automatic Construction of a Hierarchical Subject Domain for Question Answering Systems

We propose a new statistical algorithm for automatic construction of subject domains that can be used in e-mail or Web question answering systems and ontology generating. The domain hierarchy is extracted from electronic texts written in a natural language, e.g., in English. During the text processing, the quality and quantity of information presented in the texts are being evaluated and then the hierarchical relationships between the pieces of texts are established depending on the derived data. Using this approach, we have created a question answering system which executes hierarchy navigation based on a query analysis including evaluation of the user’s conversance with the subject domain. In combination, these steps result in comprehensive and non-redundant answers.