TachyonTRSA1
收藏资源简介:
Tachyon TRSA1 是一个土耳其语情感分析数据集,包含 1494 条文本评论,每条评论被人工标注为积极、消极或中性情感类别。数据分布为:积极 655 条、消极 541 条、中性 298 条。数据集结构包括两个字段:label(表示情感标签,取值为 Negative、Neutral、Positive)和 text(存储原始评论文本)。该数据集旨在解决土耳其语情感分析任务中缺乏可靠验证/开发集的问题,由三位标注者共同创建。它适用于文本分类任务,特别是情感分析场景,可用于模型训练、评估或基准测试。数据集语言为土耳其语,规模属于中等(1K 到 10K 条样本)。
Tachyon TRSA1 is a Turkish sentiment analysis dataset containing 1,494 textual reviews, each manually annotated with a sentiment label from three categories: Positive, Negative, and Neutral. The sample distribution across classes is as follows: 655 Positive samples, 541 Negative samples, and 298 Neutral samples. The dataset structure includes two fields: "label", which denotes the sentiment label with valid values of Negative, Neutral, and Positive, and "text", which stores the original review texts. Co-created by three annotators, this dataset aims to address the shortage of reliable validation/development sets for Turkish sentiment analysis tasks. It is suitable for text classification tasks, especially sentiment analysis scenarios, and can be used for model training, evaluation, or benchmark testing. The dataset is in Turkish, and its scale falls into the medium category with 1,000 to 10,000 samples.
数据集概述
- 数据集名称: Tachyon TRSA1 DATASET
- 数据集规模: 1494条文本(1K-10K之间)
- 语言: 土耳其语(tr)
- 任务类别: 文本分类(text-classification)
- 标签类型: 情感分析(sentiment-analysis)
- 情感类别: 正面(Positive)、负面(Negative)、中性(Neutral)
数据分布
| 情感标签 | 样本数量 |
|---|---|
| 正面(Positive) | 655 |
| 负面(Negative) | 541 |
| 中性(Neutral) | 298 |
数据集结构
- 字段:
label: 情感标签(Negative、Neutral、Positive)text: 评论文本
数据集创建
- 标注方式: 由三名人工标注员共同完成。
- 创建理由: 当时缺乏可靠的验证/开发集(val/dev set)。
- 标注人员: @vinnipukh、@cosmicproc
- 数据集作者: @vinnipukh、@cosmicproc
- 数据集联系方式: https://vinnipukh.neocities.org/
相关实验
- 基于该数据集的实验详见GitHub仓库:https://github.com/vinnipukh/PreliminaryLLMExperiments
许可证
- 许可证类型: 无限制公共领域(Unlicense)




