text2cypher-2024v1
收藏资源简介:
text2cypher-2024v1数据集由Neo4j公司创建,旨在将自然语言查询转换为Cypher查询语言,以扩展非技术专家用户对知识图谱的利用。该数据集包含44,387个实例,分为39,554个训练实例和4,833个测试实例。数据集通过整合和清洗多个公开可用的数据集构建,确保了数据的高质量和多样性。创建过程包括数据收集、清洗和标准化,最终形成了适用于模型微调和评估的结构化数据集。该数据集主要应用于自然语言处理和图数据库查询领域,旨在解决非专家用户在查询复杂知识图谱时的技术障碍。
The text2cypher-2024v1 dataset, developed by Neo4j, is engineered to translate natural language queries into Cypher query language, with the goal of expanding non-technical users' utilization of knowledge graphs. This dataset comprises 44,387 total instances, split into 39,554 training instances and 4,833 test instances. It is constructed by integrating and cleaning multiple publicly available datasets, ensuring high data quality and diversity. The development workflow encompasses data collection, cleaning and standardization, ultimately yielding a structured dataset suitable for model fine-tuning and evaluation. This dataset is primarily applied in the domains of natural language processing and graph database querying, aiming to address the technical barriers faced by non-expert users when querying complex knowledge graphs.
Neo4j-Text2Cypher (2024) Dataset
数据集概述
Neo4j-Text2Cypher (2024) Dataset 是一个整合了公开可用数据集的实例,经过清理和组织,以便更顺畅地使用。每个条目至少包含一个“问题、模式、Cypher”三元组,总共有44,387个实例,其中39,554个用于训练,4,833个用于测试。
数据集信息
-
特征:
question: 用户提出的文本问题,类型为字符串。schema: 数据库模式,类型为字符串。cypher: 输出的Cypher查询,类型为字符串。data_source: 数据集来源的别名,类型为字符串。instance_id: 每行的增量索引,类型为字符串。database_reference_alias: 数据库的别名(如果可用),类型为字符串。
-
数据分割:
train: 训练集,包含39,554个样本,大小为88,717,369字节。test: 测试集,包含4,833个样本,大小为11,304,360字节。
-
下载大小: 8,169,979字节
-
数据集大小: 100,021,729字节
-
配置:
default: 默认配置,包含训练集和测试集的数据文件路径。
-
许可证: Apache-2.0
-
任务类别:
text2text-generation
-
语言:
en
-
标签:
neo4jcyphertext2cypher
-
数据集名称: Neo4j-Text2Cypher Dataset (2024)
-
数据集规模: 10K < n < 100K
字段描述
| 字段 | 描述 |
|---|---|
question |
用户提出的文本问题。例如,“What is the total number of companies?” |
schema |
数据库模式。 |
cypher |
输出的Cypher查询。例如,“MATCH (c:Company) RETURN COUNT(c)” |
data_source |
数据集来源的别名。例如,"neo4jLabs_synthetic_gpt4turbo" |
database_reference_alias |
数据库的别名(如果可用)。例如,None, "neo4jlabs_demo_db_stackoverflow" |
instance_id |
每行的增量索引。 |
数据来源
数据来源如下:
| 名称 | 别名 | 链接 |
|---|---|---|
| Functional Cypher Generator (neo4j-labs/text2cypher) | neo4jLabs_text2cypher_functionalCypher_parametricTrainerWithRepeats | Link |
| Synthetic gemini demodbs (neo4j-labs/text2cypher) | neo4jLabs_text2cypher_gemini | Link |
| Synthetic gpt4o demodbs (neo4j-labs/text2cypher) | neo4jLabs_text2cypher_gpt4o | Link |
| Synthetic gpt4turbo demodbs (neo4j-labs/text2cypher) | neo4jLabs_text2cypher_gpt4turbo | Link |
| Synthetic opus demodbs (neo4j-labs/text2cypher) | neo4jLabs_text2cypher_claudeopus | Link |
| Rag-Eval datasets | neo4j_rageval_[movies,products]_text2cypher_results | N/A |
| Neo4j-Text2Cypher’23 datasets | neo4j_text2cypher2023-[train, test] | N/A |
| Crowdsourcing dataset | neo4j_crowdsourced_text2cypher_raw | N/A |
| HF-iprahara/text_to_cypher | hf_iprahara_text_to_cypher | Link |
| HF-dfwlab/cypher | hf_dfwlab_cypher_eng-to-cypher | Link |
| HF-vedana17/text-to-cypher | hf_vedana17_text-to-cypher_dataset | Link |
| Cy-Spider | cySpider_semanticParser4Graph_data_folder | Link |




