DeepPavlov/wizard_of_wikipedia_ru

该数据集是一个多语言信息检索数据集，包含三个主要部分：语料库（corpus）、查询（queries）和查询-文档相关性标注（qrels）。语料库配置包含文档的ID、标题和文本，支持英语和俄语版本（如title_ru、text_ru字段），用于存储检索文档。查询配置包含查询的ID、文本（以列表形式存储内容和角色）、人物角色和主题，同样支持英语和俄语版本（如text_ru、persona_ru、topic_ru字段），可能用于模拟用户查询或对话场景。相关性标注配置连接查询和语料库文档，包含查询ID、语料库ID和分数，用于训练和评估检索模型。数据集分为训练、验证和测试分片，总大小约为1GB，适用于信息检索、对话系统或多语言NLP任务。

This dataset is a multilingual information retrieval dataset comprising three main components: corpus, queries, and query-document relevance judgments (qrels). The corpus configuration includes document IDs, titles, and texts, with support for both English and Russian versions (e.g., title_ru, text_ru fields), serving as a collection of documents for retrieval. The queries configuration includes query IDs, text (stored as a list with content and role), persona, and topic, also supporting English and Russian versions (e.g., text_ru, persona_ru, topic_ru fields), potentially for simulating user queries or dialogue scenarios. The relevance judgments configuration links queries to corpus documents with query IDs, corpus IDs, and scores, used for training and evaluating retrieval models. The dataset is split into train, validation, and test subsets, with a total size of approximately 1GB, suitable for information retrieval, dialogue systems, or multilingual NLP tasks.