devrev/search

**DevRev搜索数据集**包含了对DevRev公共文章进行转换和匿名化处理的用户查询，以及与这些查询相关的回答文章块。该数据集旨在为企业搜索和重排名系统提供基准测试。数据集还包括用于检索的测试查询和基础文章块。数据集由DevRev研究团队策划，主要用于评估企业支持场景下的搜索、检索和重排名模型。数据集包含三个主要部分：`annotated_queries.parquet`（带有注释的文章块查询对）、`knowledge_base.parquet`（来自DevRev客户支持文档的文章块）和`test_queries.parquet`（用于评估的保留查询）。所有用户查询均经过匿名化处理，不包含任何个人或敏感信息。

The **DevRev Search Dataset** contains transformed and anonymized user queries on DevRevs public articles, paired with relevant article chunks that answer them. It is designed to benchmark enterprise search and reranking systems. The dataset also includes test queries along with the underlying article chunks used for retrieval. Curated by Research@DevRev, it is intended for evaluation of search, retrieval, and reranking models in enterprise support contexts. The dataset consists of three main components: `annotated_queries.parquet` (queries paired with annotated article chunks), `knowledge_base.parquet` (article chunks from DevRevs customer-facing support documentation), and `test_queries.parquet` (held-out queries for evaluation). All user queries were anonymized and stripped of any identifying or sensitive data.