M3ED*

该数据集是一个包含来自56部不同电视剧的双向对话的话题密度数据集，旨在评估多模态话题信息在对话中的有效性。数据集按照语句的比例7:1:2被划分为训练集、验证集和测试集，确保测试集中的电视剧领域在训练集中出现过。该数据集包含了来自56部不同电视剧的对话，其任务是进行多模态对话情感检测。

This dataset is a topic density dataset containing bidirectional dialogues from 56 distinct TV series, aiming to evaluate the effectiveness of multimodal topic information in conversational contexts. The dataset is divided into training, validation and test sets according to the 7:1:2 utterance ratio, ensuring that the TV series domains in the test set have appeared in the training set. This dataset includes dialogues from 56 different TV series, and its task is multimodal conversational sentiment detection.

搜集汇总

背景与挑战

背景概述

M3ED是一个多模态、多场景、多标签的情感对话数据库，发布于ACL 2022。它包含文本、音频和视觉模态的情感数据，提供了多种特征提取方法和不同模型的实现，适用于情感分析和对话系统研究。