SPOT(Sentiment Polarity Annotations Dataset)

SPOT 数据集包含来自 Yelp'13 和 IMDB 集合 ([1][2]) 的 197 条评论，并使用分段级别的极性标签（正面/中性/负面）进行注释。已在 2 个粒度级别上收集注释： 句子基本话语单元 (EDU)，即由最先进的 RST 解析器生成的子句从句 该数据集旨在帮助情感分析研究，特别是评估试图在细粒度、分段级别的基础上预测情绪的方法。

The SPOT dataset contains 197 reviews sourced from the Yelp'13 and IMDB corpora ([1][2]), annotated with segment-level polarity labels (positive/neutral/negative). Annotations have been collected at two granularity levels: elementary discourse units (EDUs), which are clausal clauses generated by a state-of-the-art RST parser. This dataset is designed to facilitate sentiment analysis research, particularly for evaluating methods that attempt to predict sentiment at the fine-grained, segment-level.

搜集汇总

背景与挑战

背景概述

SPOT数据集包含197条来自Yelp'13和IMDB的评论，并标注了分段级别的情感极性标签（正面/中性/负面），注释粒度涵盖句子基本话语单元和子句从句。该数据集旨在为细粒度情感分析研究提供评估支持。