tomg-group-umd/fictionalqa_training_splits

FictionalQA Training Splits数据集是一个由机器生成的数据集，主要用于文本生成和问答任务，包括封闭域问答、封闭书籍问答和开放书籍问答。数据集包括多种配置，具有不同的特征和分割方式，例如event_split_fictions_webtext_train_ds_valratio0.33_seed1234、event_split_fictions_webtext_train_ds_valratio0.33_seed1234_mcq_topk10、event_split_fictions_webtext_train_ds_valratio0.33_seed1234_mcq_topk4等。每个配置都有其自己的特征和分割，包括事件ID、小说ID、文本、问题ID、跨度答案、自然答案、输入、目标、目标跨度、备选目标等。数据集分为训练和验证分割，具有不同数量的示例和文件大小。

The FictionalQA Training Splits dataset is a machine-generated dataset primarily used for tasks such as text generation and question-answering, specifically closed-domain QA, closed-book QA, and open-book QA. The dataset includes various configurations with different features and splits, such as event_split_fictions_webtext_train_ds_valratio0.33_seed1234, event_split_fictions_webtext_train_ds_valratio0.33_seed1234_mcq_topk10, event_split_fictions_webtext_train_ds_valratio0.33_seed1234_mcq_topk4, and others. Each configuration has its own set of features and splits, including event_id, fiction_id, text, question_id, span_answer, natural_answer, input, target, target_span, alt_targets, and more. The dataset is divided into train and validation splits, with varying numbers of examples and file sizes.