Emotion Classification

Emotion classification, or emotion categorization, is the task of recognising emotions to classify them into the corresponding category. Given an input, classify it as 'neutral or no emotion' or as one, or more, of several given emotions that best represent the mental state of the subject's facial expression, words, and so on. Some example benchmarks include ROCStories, Many Faces of Anger (MFA), and GoEmotions. Models can be evaluated using metrics such as the Concordance Correlation Coefficient (CCC) and the Mean Squared Error (MSE).

Benchmarks

Libraries

Datasets

Subtasks

Most implemented papers

GoEmotions: A Dataset of Fine-Grained Emotions

Content

Real-time Convolutional Neural Networks for emotion and gender classification

Multimodal Speech Emotion Recognition Using Audio and Text

NTUA-SLP at SemEval-2018 Task 1: Predicting Affective Content in Tweets with Deep Attentive RNNs and Transfer Learning

Classifying and Visualizing Emotions with Emotional DAN

EmoTxt: A toolkit for emotion recognition from text

DialogueRNN: An Attentive RNN for Emotion Detection in Conversations

Speech Emotion Recognition Using Multi-hop Attention Mechanism

DialogueGCN: A Graph Convolutional Neural Network for Emotion Recognition in Conversation

Emotion Classification in a Resource Constrained Language Using Transformer-based Approach