allegro/klej-polemo2-out
收藏资源简介:
PolEmo2.0-OUT数据集是一个包含来自四个领域(医学、酒店、产品和大学)的在线消费者评论的数据集。该数据集在完整评论和单个句子的层面上进行了人工标注,包含超过8000条评论,其中约85%来自医学和酒店领域。数据集被用于形成两个任务,其中第二个任务是测试模型在领域外评论(产品和大学领域)上的表现。由于这些领域的原始测试集较少(每个领域50条评论),因此使用了原始的领域外训练集中的900条评论进行测试,并创建了新的开发和测试集分割。最终,该任务包含1000条评论,与领域内测试数据集的1400条评论规模相当。
The PolEmo2.0-OUT dataset is a collection of online consumer reviews spanning four domains: medicine, hospitality, products, and universities. This dataset is manually annotated at both the full review and individual sentence levels, with over 8,000 reviews in total, of which approximately 85% come from the medicine and hospitality domains. The dataset is used to construct two tasks, where the second one aims to evaluate model performance on out-of-domain reviews from the products and universities domains. Given the small size of the original test sets for these two domains (only 50 reviews per domain), 900 reviews from the original out-of-domain training set were repurposed for testing, and new splits for development and test sets were created. Ultimately, this task consists of 1,000 reviews, which is comparable in scale to the 1,400 reviews in the in-domain test dataset.
数据集概述
名称: PolEmo2.0-OUT
语言: 波兰语 (pl)
许可: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)
多语言性: 单语
规模: 1K<n<10K
来源: 原始数据
任务类别: 文本分类
任务ID: 情感分类
数据集描述
PolEmo2.0是一个包含四个领域(医药、酒店、产品和大学)的在线消费者评论数据集。数据集包含超过8000条评论,其中约85%来自医药和酒店领域。数据集中的评论和句子均由人工标注。
任务详情
输入: 句子
输出: 句子情感标签(zero: 中性, minus: 负面, plus: 正面, amb: 模糊)
度量标准: 准确度
数据分割
| 子集 | 数量 |
|---|---|
| 训练 | 5783 |
| 测试 | 722 |
| 验证 | 723 |
类别分布
| 类别 | 情感 | 训练 | 验证 | 测试 |
|---|---|---|---|---|
| minus | 正面 | 0.379 | 0.334 | 0.368 |
| plus | 负面 | 0.271 | 0.332 | 0.302 |
| amb | 模糊 | 0.182 | 0.332 | 0.328 |
| zero | 中性 | 0.168 | 0.002 | 0.002 |
引用信息
@inproceedings{kocon-etal-2019-multi, title = "Multi-Level Sentiment Analysis of {P}ol{E}mo 2.0: Extended Corpus of Multi-Domain Consumer Reviews", author = "Koco{ }, Jan and Mi{l}kowski, Piotr and Za{s}ko-Zieli{ }ska, Monika", booktitle = "Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL)", month = nov, year = "2019", address = "Hong Kong, China", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/K19-1092", doi = "10.18653/v1/K19-1092", pages = "980--991", abstract = "In this article we present an extended version of PolEmo {--} a corpus of consumer reviews from 4 domains: medicine, hotels, products and school. Current version (PolEmo 2.0) contains 8,216 reviews having 57,466 sentences. Each text and sentence was manually annotated with sentiment in 2+1 scheme, which gives a total of 197,046 annotations. We obtained a high value of Positive Specific Agreement, which is 0.91 for texts and 0.88 for sentences. PolEmo 2.0 is publicly available under a Creative Commons copyright license. We explored recent deep learning approaches for the recognition of sentiment, such as Bi-directional Long Short-Term Memory (BiLSTM) and Bidirectional Encoder Representations from Transformers (BERT).", }




