Bengali-English Code-Mixed Corpus

本研究创建了首个公开的Bengali-English代码混合语料库，用于情感分析。该数据集包含5000条从Twitter收集并经过筛选的代码混合数据，每条数据都带有语言和极性标签。数据集的创建过程涉及使用混合系统（结合规则基础和监督模型）进行语言和情感标签的自动标注，随后由一组标注者进行人工验证。该数据集主要用于解决印度多语言环境下的情感分析问题，特别是在社交媒体中普遍存在的代码混合现象。

This study created the first publicly available Bengali-English code-mixed corpus for sentiment analysis. The dataset contains 5000 filtered code-mixed samples collected from Twitter, with each sample annotated with both language and polarity labels. The dataset construction process utilized a hybrid system combining rule-based and supervised models to automatically perform language and polarity label annotation, followed by manual validation conducted by a panel of annotators. This dataset is primarily developed to address the challenges of sentiment analysis in India's multilingual context, especially the code-mixing phenomenon that is prevalent in social media.