Few-Shot Emotion Recognition in Conversation with Sequential Prototypical Networks

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 6858–6870 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 6858 Few-Shot Emotion Recognition in Conversation with Sequential Prototypical Networks Gaël Guibon1,2, Matthieu Labeau1, Hélène Flamein2, Luce Lefeuvre2, and Chloé Clavel1 1LTCI, Télécom-Paris, Institut Polytechnique de Paris 2Direction Innovation & Recherche SNCF {gael.guibon,matthieu.labeau,chloe.clavel}@telecom-paris.fr {ext.gael.guibon,helene.flamein,luce.lefeuvre}@sncf.fr Abstract Several recent studies on dyadic human- human interactions have been done on con- versations without specific business objectives. However, many companies might benefit from studies dedicated to more precise environ- ments such as after sales services or customer satisfaction surveys. In this work, we place ourselves in the scope of a live chat customer service in which we want to detect emotions and their evolution in the conversation flow. This context leads to multiple challenges that range from exploiting restricted, small and mostly unlabeled datasets to finding and adapt- ing methods for such context. We tackle these challenges by using Few-Shot Learning while making the hypothesis it can serve con- versational emotion classification for different languages and sparse labels. We contribute by proposing a variation of Prototypical Net- works for sequence labeling in conversation that we name ProtoSeq. We test this method on two datasets with different languages: daily conversations in English and customer service chat conversations in French. When applied to emotion classification in conversations, our method proved to be competitive even when compared to other ones. The code for Proto- Seq is available at https://github.com/ gguibon/ProtoSeq. 1 Introduction There has been a recent surge in research focusing on analyzing dyadic human to human interactions. Many of these studies (Poria et al., 2017; Zadeh et al., 2018a,b; Majumder et al., 2019) focus on emotion recognition in conversations (ERC) taking into account multiple data modalities. Moreover, most of the progress made in ERC has been done without factoring in constraints corresponding to specific but prominent industrial applications, like customer service. This is partly due to studies fo- cusing on using artificial datasets (Li et al., 2017; Busso et al., 2008) made of mock-up conversations to facilitate result replication and comparison. A few existing studies address customer service appli- cations (Mundra et al., 2017; Yom-Tov et al., 2018; Maslowski et al., 2017) and show the difficulties to deal with such in-the-wild and domain-specific data. In this work, we focus on data from a live chat support in which we want to detect emotions and their evolution in the conversational flow. This set- ting corresponds to a human dyadic conversation, albeit with a specific business-related objective. We make the hypothesis that the emotion flows of the visitor and the operator will bring information on the quality of the service and help operators better assist customers. This hypothesis is close to rel- evant studies on the importance of emotions and empathy in dyadic call center conversations (Alam, 2017; Alam et al., 2018). This specific setting leads to multiple challenges: indeed, it is difficult and costly to label this kind of data — and even then, these …

Content truncated to 500 words