POM电影评论数据集

POM电影评论数据集由巴黎电信学院和巴黎-萨克雷大学创建，包含1000个视频评论，旨在分析评论者的说服力。数据集大小为1000个视频，每个视频平均包含15.1个句子，22.5个tokens。创建过程中，研究者对视频评论进行了细粒度的意见标注，包括意见的粗粒度和细粒度标注。该数据集主要应用于细粒度多模态意见预测，旨在解决如何从视频评论中准确预测不同意见成分的问题。

The POM Movie Review Dataset was created by Télécom Paris and Université Paris-Saclay. It contains 1000 video reviews and is designed to analyze the persuasiveness of commenters. The dataset consists of 1000 videos, with each video averaging 15.1 sentences and 22.5 tokens. During the dataset construction, researchers conducted fine-grained opinion annotation on the video reviews, covering both coarse-grained and fine-grained annotations of opinions. This dataset is primarily utilized for fine-grained multimodal opinion prediction, aiming to address the problem of accurately predicting various opinion components from video reviews.

搜集汇总

背景与挑战

背景概述

POM电影评论数据集是Persuasive Opinion Multimedia (POM)语料库的注释变体，专为意见预测任务设计，包含表达级和词级的细粒度意见注释。该数据集是多模态的，整合了音频、视频和文本特征，并提供了视频标签（如个性、说服力）以及丰富的意见标签（如极性、目标），适用于情感分析和意见挖掘研究。