IndicSpeech: Text-to-Speech Corpus for Indian Languages

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 620–627 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 620 EDA: Enriching Emotional Dialogue Acts using an Ensemble of Neural Annotators Chandrakant Bothe, Cornelius Weber, Sven Magg, and Stefan Wermter Knowledge Technology, Department of Informatics, University of Hamburg, Vogt-Koelln-Str. 30, 22527 Hamburg, Germany www.informatik.uni-hamburg.de/WTM/ {bothe,weber,magg,wermter}@informatik.uni-hamburg.de Abstract The recognition of emotion and dialogue acts enriches conversational analysis and help to build natural dialogue systems. Emotion interpretation makes us understand feelings and dialogue acts reflect the intentions and performative functions in the utterances. However, most of the textual and multi-modal conversational emotion corpora contain only emotion labels but not dialogue acts. To address this problem, we propose to use a pool of various recurrent neural models trained on a dialogue act corpus, with and without context. These neural models annotate the emotion corpora with dialogue act labels, and an ensemble annotator extracts the final dialogue act label. We annotated two accessible multi-modal emotion corpora: IEMOCAP and MELD. We analyzed the co-occurrence of emotion and dialogue act labels and discovered specific relations. For example, Accept/Agree dialogue acts often occur with the Joy emotion, Apology with Sadness, and Thanking with Joy. We make the Emotional Dialogue Acts (EDA) corpus publicly available to the research community for further study and analysis. Keywords: Emotional Dialogue Acts Corpus, Conversational Analysis, Automated Neural Ensemble Annotation and Evaluation 1. Introduction With the growing demand for human-computer/robot in- teraction systems, detecting the emotional state of the user can substantially benefit a conversational agent to respond at an appropriate emotional level. Emotion recognition in conversations has proven valuable for potential applications such as response recommendation or generation, emotion- based text-to-speech, personalization. Human emotional states can be expressed verbally and non-verbally (Ekman et al., 1987; Osgood et al., 1975). However, while build- ing an interactive dialogue system, the interface needs di- alogue acts. A typical dialogue system consists of a lan- guage understanding module which requires to determine the meaning and intention in the human input utterances (Wermter and L¨ochel, 1996; Berg, 2015; Ultes et al., 2017). Also, in discourse or conversational analysis, dialogue acts are the main linguistic features to consider (Bothe et al., 2018a). A dialogue act provides an intention and performa- tive function in an utterance of the dialogue. For example, it can infer a user’s intention by distinguishing Question, Answer, Request, Agree/Reject, etc. and performative func- tions such as Acknowledgement, Conversational-opening or -closing, Thanking, etc. The dialogue act information together with emotional states can be very useful for a spo- ken dialogue system to produce natural interaction (Ihasz and Kryssanov, 2018). The research in emotion recognition is growing, and many datasets are available, such as text-, speech- or vision- based, and multi-modal-based emotion data. Emotion ex- pression recognition is a challenging task, and hence mul- timodality is crucial (Ekman et al., 1987). However, few conversational multi-modal emotion recognition datasets are available, for example, IEMOCAP (Busso et al., 2008) or SEMAINE …

Content truncated to 500 words