mayankkeshari/hotpotqa-sentence-retrieval-with-attn-scores

该数据集名为hotpotqa-sentence-retrieval-with-attn-scores，可能基于HotpotQA数据集，专注于句子检索任务，并包含注意力分数。数据集包含多种特征，如查询ID、段落ID、句子哈希、查询、答案、查询类型、查询级别、段落、段落索引、标题、段落中的句子数量、句子、句子索引、句子字符起始和结束位置、相关性标签、来源分割、原始分割、段落重新排序分数、句子注意力分数及其归一化值等。数据集分为训练集、验证集和测试集，分别包含420、420和431个样本。

The dataset is named hotpotqa-sentence-retrieval-with-attn-scores, likely based on the HotpotQA dataset, focusing on sentence retrieval tasks with attention scores. It includes various features such as query_id, passage_id, sentence_hash, query, answer, query_type, query_level, passage, passage_idx, title, num_sentences_in_passage, sentence, sentence_idx, sentence_char_start, sentence_char_end, relevant, source_split, original_split, passage_reranker_score, sentence_attn_score, and sentence_attn_score_normalized. The dataset is divided into train, validation, and test splits, containing 420, 420, and 431 examples respectively.