We describe and evaluate a personalized conversational item
recommendation system.  In this setting, the system directs
conversations that help users find items, and adapts its actions
according to information from past conversations with a given user.
We present results demonstrating the effectiveness of our approach.