COSMIC: COmmonSense knowledge for eMotion Identification in Conversations

Abstract

In this paper, we address the task of utterance level emotion recognition in conversations using commonsense knowledge. We propose COSMIC, a new framework that incorporates different elements of commonsense such as mental states, events, and causal relations, and build upon them to learn interactions between interlocutors participating in a conversation. Current state-of-theart methods often encounter difficulties in context propagation, emotion shift detection, and differentiating between related emotion classes. By learning distinct commonsense representations, COSMIC addresses these challenges and achieves new state-of-the-art results for emotion recognition on four different benchmark conversational datasets. Our code is available at https://github.com/ declare-lab/conv-emotion.

Emotion Recognition on Edge Devices: Training and Deployment

“…Conversations are acted and the actors either improvise or follow a script depending on the particular conversation. Like [ 60 ], we also use the six most common emotions, namely happy, sad, neutral, angry, excited, and frustrated, and the same data splits. In the future, however, we plan to use the Hourglass of Emotions [ 61 ] as categorization model.…”

“…For additional comparison, state-of-the-art method COSMIC reports an F1 score of 0.6521 [ 60 ]. However this method uses a commonsense knowledge extractor trained on large knowledge bases.…”

Pronunciation-Enhanced Chinese Word Embedding

“…Text classification is widely used to evaluate the effectiveness of word embeddings in NLP tasks [36]. We adopted the Fudan dataset, which contained documents on 20 topics, for training 11 and testing 12 . Following [29], we selected 12,545 (6424 for training and 6121 for testing) documents on five topics: environment, agriculture, economy, politics, and sports.…”

“…Word embedding (also known as distributed word representation) denotes a word as a real-valued and lowdimensional vector. In recent years, it has attracted significant attention and has been applied to many natural language processing (NLP) tasks, such as sentiment classification [1][2][3][4][5], sentence/concept-level sentiment analysis [6][7][8][9][10], question answering [11,12], text analysis [13,14], named entity recognition [15,16], and text segmentation [17,18]. For instance, in some prior studies, word embeddings in a sentence were summed and averaged to obtain the probability of each sentiment [19].…”

“…http://downl oad.csdn.net/downl oad/githu b_36326 955/97479 2712 http://downl oad.csdn.net/downl oad/githu b_36326 955/97479 2913 https ://githu b.com/cjlin 1/libli near 14 https ://www.aitec hclub .com/data-detai l?data_id=29…”