No that's not what I meant: Handling Third Position Repair in Conversational Question Answering

The ability to handle miscommunication is crucial to robust and faithful conversational AI. People usually deal with miscommunication immediately as they detect it, using highly systematic interactional mechanisms called repair. One important type of repair is Third Position Repair (TPR) whereby a speaker is initially misunderstood but then corrects the misunderstanding as it becomes apparent after the addressee’s erroneous response (see Fig. 1). Here, we collect and publicly release REPAIR-QA1, the first large dataset of TPRs in a conversational question answering (QA) setting.

In this paper, we focus on an important class of repairs that has, to our knowledge, been neglected in the NLP community, likely due to the unavailability of data: Third Position Repair (TPR; (Schegloff, 1992); aka repair after next turn). These occur when the addressee initially misunderstands the speaker (Fig. 1 at T1, the trouble source turn), responds based on this misunderstanding (at T2), which in turn reveals the misunderstanding to the addressee who then goes on to correct the misunderstanding (at T3).