other,19-6-E06-1035,ak | on <term> models </term> that combine <term> | lexical-cohesion and conversational features | </term> , but do not change the general preference | #11564 We also find that the transcription errors inevitable in ASR output have a negative impact on models that combinelexical-cohesion and conversational features, but do not change the general preference of approach for the two tasks. |