Satisfying Varying Retrieval Requirements in Case Based Intelligent Directory Assistance

Intelligent directory assistance is the ability to retrieve relevant subscriber records in the presence of mismatches between the query and the subscriber record. The challenges are to provide inexact retrieval over very large data volumes within a reasonable timeframe as well as the ability to tailor retrieval to differing query requirements. In this paper we discuss a case-based approach to intelligent directory assistance which focusses on a layered approach to retrieval. The case-base is multiply indexed where each level indexes the case-base at a level of dissimilarity. We introduce the concepts of Levels of Lenience, Degree of Lenience and Search templates and demonstrate how these enable retrieval to be fast and yet tailored to query requirements. A production quality system based on these ideas has been implemented at a leading telecommunications service provider and is giving excellent