FlowQA: Grasping Flow in History for Conversational Machine\n Comprehension

Conversational machine comprehension requires the understanding of the\nconversation history, such as previous question/answer pairs, the document\ncontext, and the current question. To enable traditional, single-turn models to\nencode the history comprehensively, we introduce Flow, a mechanism that can\nincorporate intermediate representations generated during the process of\nanswering previous questions, through an alternating parallel processing\nstructure. Compared to approaches that concatenate previous questions/answers\nas input, Flow integrates the latent semantics of the conversation history more\ndeeply. Our model, FlowQA, shows superior performance on two recently proposed\nconversational challenges (+7.2% F1 on CoQA and +4.0% on QuAC). The\neffectiveness of Flow also shows in other tasks. By reducing sequential\ninstruction understanding to conversational machine comprehension, FlowQA\noutperforms the best models on all three domains in SCONE, with +1.8% to +4.4%\nimprovement in accuracy.\n