TekGen
收藏资源简介：
数据集是 KELM 语料库的一部分 这是 Wikipedia text--Wikidata KG 对齐语料库，用于训练数据到文本生成模型。请注意，这是通过远程监督生成的语料库，不应作为评估的黄金标准。它由 3 个文件组成： https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/quadruples-train.tsv https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021 /quadruples-validation.tsv https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/quadruples-test.tsv 每个文件每行包含一个示例。每个示例都是一个包含三个字段的 json 对象：三元组：形式为三元组的列表（主题、关系、对象）。例如。 （X 人，获得的奖项，Y 奖）。如果三元组有一个子属性，那么它是四元组。例如。 （X 人，Y 奖，于 Z 日收到）。序列化三元组：连接在一起的三元组，用于 T5 的输入。格式为“<subject><relation><object> “有些主体有多重关系，例如”<subject><relation1><object1><relation2><object2><relation3><object3> ”。有关这些关系如何分组的更多详细信息，请参阅论文。句子：与这些三元组对齐的维基百科句子。实体的名称、别名和 Wikidata Id 可以在 https://storage.googleapis 中找到。 com/gresearch/kelm-corpus/updated-2021/entities.jsonl。
This dataset is part of the KELM corpus. It is a Wikipedia text--Wikidata Knowledge Graph (KG) aligned corpus used for training data-to-text generation models. Please note that this is a corpus generated via distant supervision and should not be treated as a gold standard for evaluation. It consists of three files: https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/quadruples-train.tsv https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/quadruples-validation.tsv https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/quadruples-test.tsv Each file contains one example per line. Each example is a JSON object with three fields: 1. Triples: A list of triples formatted as (subject, relation, object). For example, (Person X, award received, Award Y). If a triple has a sub-property, it becomes a quadruple, e.g., (Person X, Award Y, received on Date Z). 2. Serialized triples: Concatenated triples for input to T5. The format is "<subject><relation><object>". For cases where a subject has multiple relations, the format follows "<subject><relation1><object1><relation2><object2><relation3><object3>". Refer to the paper for more details on how these relations are grouped. 3. Sentence: A Wikipedia sentence aligned with these triples. The names, aliases, and Wikidata IDs of entities can be found at https://storage.googleapis.com/gresearch/kelm-corpus/updated-2021/entities.jsonl.
