Considering the subjectivity to rationalise evaluation approaches: The example of Spoken Dialogue Systems

We present in this paper a reading grid aiming at helping the evaluator take into account the subjectivity factor when designing evaluation protocols. Actually, however most contributions to Spoken Dialogue Systems evaluation tend to objectify their approach with rationalising purpose, we believe that subjectivity needs to be considered for valuable evaluations. The first section shows how closely evaluation processes are dependant on their contexts and on the evaluators' perspectives. We then present an anthropocentric framework that establishes the evaluator as a mediator between the consideration of contextual elements and a rationalising corpus of evaluation procedures. We finally anticipate the benefits our framework brings at both individual and community levels.