Edinburgh Napier University

  Evaluation of complex, collaborative dialogue systems is a difficult task.Traditionally, developers have relied upon subjective feedback from the user,and parametrisation over observable metrics. However, both models place somereliance on the notion of a task; that is, the system is helping to userachieve some clearly defined goal, such as book a flight or complete a bankingtransaction. It is not clear that such metrics are as useful when dealing witha system that has a more complex task, or even no definable task at all, beyondmaintain and performing a collaborative dialogue. Working within the EU fundedCOMPANIONS program, we investigate the use of appropriateness as a measure ofconversation quality, the hypothesis being that good companions need to be goodconversational partners . We report initial work in the direction of annotatingdialogue for indicators of good conversation, including the annotation andcomparison of the output of two generations of the same dialogue system