Backchannels, which refer to short and often affirmative or empathetic responses from a listener during a conversation, play a crucial role in effective communication. In this paper, we introduce CABP(Context-Aware Backchannel Prediction), a sequential and attentive context approach aimed at enhancing backchannel prediction performance. Additionally, CABP leverages the pretrained wav2vec model for encoding audio signal. Experimental results show that CABP performs better than context-free models, with performance improvements of 1.3% and 1.8% in Korean and English datasets, respectively. Furthermore, when utilizing the pretrained wav2vec model, CABP consistently demonstrates the best performance, achieving performance improvements of 4.4% and 3.1% in Korean and English datasets.

Produced in the form of small injections such as “Yeah!” or “Uh-Huh” by listeners in a conversation, supportive verbal feedback (i.e., backchanneling) is essential for natural dialogue. Highlighting its tight relation to speaker intent and utterance type, we propose a multi-task learning approach that learns textual representations for the task of backchannel prediction in tandem with dialogue act classification. We demonstrate the effectiveness of our approach by improving the prediction of specific backchannels like “Yeah” or “Really?” by up to 2.0% in F1. Additionally, whereas previous models relied on well-established methods to extract audio features, we further pre-train the audio encoder in a self-supervised fashion using voice activity projection. This leads to additional gains of 1.4% in weighted F1.