tech,16-1-E89-1016,ak approaches that have been used in the <term> evaluation </term> of <term> Natural Language systems </term>
tech,18-1-E89-1016,ak used in the <term> evaluation </term> of <term> Natural Language systems </term> . We conclude that previous approaches
tech,22-2-E89-1016,ak use , e.g. solving a task requiring <term> data retrieval </term> . This raises questions about the
other,5-3-E89-1016,ak </term> . This raises questions about the <term> validity </term> of such approaches . In the second
other,5-5-E89-1016,ak task . We evaluate the demands that <term> task dialogues </term> collected using this technique ,
tech,16-5-E89-1016,ak technique , place upon a prototype <term> Natural Language system </term> . We identify three important requirements
other,16-6-E89-1016,ak the task that we gave our subjects : <term> operators </term> specific to the task of <term> database
other,22-6-E89-1016,ak operators </term> specific to the task of <term> database access </term> , <term> complex contextual reference
other,25-6-E89-1016,ak task of <term> database access </term> , <term> complex contextual reference </term> and <term> reference </term> to the structure
other,29-6-E89-1016,ak complex contextual reference </term> and <term> reference </term> to the structure of the <term> information
other,35-6-E89-1016,ak reference </term> to the structure of the <term> information source </term> . We discuss how these might be satisfied
tech,9-7-E89-1016,ak these might be satisfied by future <term> Natural Language systems </term> . Recently , we initiated a project
hide detail