We study the performance of methods based on text and entity features on a conversational assistance dataset from TREC CAST. Our methods do not make use of any question answering or dialogue tracking component. We apply methods which make use of entities in context to another entity and entity pairs.
