0601p/Traveling_Namuwiki_Actions

Traveling Namuwiki Actions 是一个基于 Namuwiki 页面链接构建的邻接列表数据集，专为图导航任务设计。每个数据行包含一个页面标题（源页面）和其链接的页面标题列表（动作），这些链接可用于在图导航中选择下一个动作。该数据集源自 Hugging Face 数据集 heegyu/namuwiki，仅存储页面标题之间的链接关系，不包含完整的文章文本。在预处理过程中，已移除指向文件、类别、图像、外部URL和自链接的链接，并删除了没有出链的页面。数据集以 JSONL 格式存储，其中 actions 字段已排序和去重，总共有 562,217 个源页面和 20,715,178 个链接，平均每个页面有 36.85 个链接。该数据集可用于构建内存中的图结构，支持图遍历和导航任务，但需要注意，某些链接的页面标题可能未作为源标题出现，应视为终端页面处理。Traveling Namuwiki Actions is an adjacency-list dataset built from Namuwiki page links, designed for graph-navigation tasks. Each row contains a page title and the list of linked page titles that can be used as next actions in graph navigation. The dataset is derived from the Hugging Face dataset heegyu/namuwiki, storing only page-title links without full article text. During preprocessing, links to files, categories, images, external URLs, and self-links were removed, and pages with no outgoing links were excluded. It is stored in JSONL format, with actions sorted and deduplicated, totaling 562,217 source pages and 20,715,178 links, averaging 36.85 links per page. This dataset can be used to build in-memory graphs for traversal and navigation tasks, but note that some linked page titles may not appear as source titles and should be handled as terminal pages.