DSL-13-SRMAP/TeSent_Benchmark-Dataset

TeSent_Benchmark-Dataset是一个大规模的泰卢固语情感分析基准数据集，专注于通过人工注释的理由实现情感分析的可解释性。该数据集包含22,505个句子，每个句子由三个注释者独立注释，提供主要情感标签（正面、负面或中立）、理由（正面或负面标签必须，中立标签可选）以及可选的次要情感标签。数据集支持细粒度的解释性、捕捉情感模糊性、多注释者设计，提供了标准的训练集、验证集和测试集分割，适用于可解释的情感分类和理由提取。

TeSent_Benchmark-Dataset is a large-scale Telugu sentiment analysis benchmark dataset focused on explainability through human-annotated rationales. The dataset contains 22,505 sentences, each annotated independently by three annotators providing a primary sentiment label (Positive, Negative, or Neutral), rationale (required for positive/negative, optional for neutral), and an optional secondary sentiment label. The dataset supports fine-grained explainability, captures ambiguity in sentiment, features a multi-annotator design, and includes standard train/val/test splits suitable for explainable sentiment classification and rationale extraction.