bermaneh/codeswitching-sentiment-bias-exp2-results-v1
收藏官方服务：
资源简介：
该数据集是实验2的完整结果，涉及掩码标记语言预测。实验通过掩码处理代码转换句子中的关键词，并使用Qwen2.5-7B-Instruct模型来填充空白，测量模型在完成代码转换句子时的语言偏好。数据集包含多个列，如句子ID、原始句子、掩码句子、真实单词、真实语言等，并提供了详细的实验结果和指标。
Full results of Experiment 2: Masked Token Language Prediction. The experiment masks the top SHAP-contributing word in code-switched sentences and uses the Qwen2.5-7B-Instruct model to fill in the blank, measuring the models language preference when completing code-switched sentences. The dataset includes various columns such as sentence_id, original_sentence, masked_sentence, ground_truth_word, ground_truth_language, etc., and provides detailed experimental results and metrics.
提供机构：
bermaneh
