Integrating a query language for structured and semi-structured data and IR techniques

The authors describe the basic ideas and concepts behind the Information Retrieval Query Language (IRQL) that is used as one of the back-ends in the GETESS project. The front-end provides a user interface which is embedded in a dialogue system. This dialogue system allows queries to be formulated in a user friendly (i.e. exploiting a limited range of natural language) and interactive way. Access to the analyzed data is provided by IRQL. The principal focus of IRQL development is the integration of concepts of information retrieval, database query languages, and query languages for semi-structured data. Therefore, we will be able to exploit the structure of documents, if known, and can additionally use information retrieval techniques regardless of whether the structure is known or not. Our approach develops a query language that is compatible with the recently adopted SQL99 standard and information retrieval clauses (e.g. Boolean retrieval). Our data model extends the object-relational model and additionally supports an abstraction of attributes. That is, we can use attribute-independent queries as well as attribute-dependent ones as in RDBMSs. We evaluate IRQL queries by mapping them to queries supported by existing systems such as object-relational DBMSs, full-text DBMSs, or conventional search engines, and post processing the results supplied by these systems, if necessary.