nassimjp/Pashto-Online-Conversations

这是一个名为پښتو آنلاین خبرې اترې - Pashto Online Conversations的数据集，专为AI模型训练而设计。它包含超过100,000条自然、简单且有趣的普什图语在线对话，涵盖了日常聊天、笑话、虚构故事以及在线朋友间的友好交流。数据集强调自然性，对话源自真实日常语言；具有多样性，从简单幽默到深度哲学讨论；保证对话的完整性和双向性；并确保与普什图文化、传统和价值观相符。数据以JSONL格式存储，每条记录包含一个对话列表，每个对话有from（发送者类型，如human或gpt）和value（对话文本）字段。数据集按主题分类，包括幽默笑话、电影电视、音乐、食物烹饪、一般聊天、艺术DIY、旅行探险和智力讨论等。数据经过严格质量控制，包括去重、清理、标准化，并支持断点续传。数据集由英语在线对话翻译成普什图语，并保持了对话的自然性、表情符号和文化适应性。它适用于文本生成任务，采用Apache 2.0许可证。

This is a dataset named پښتو آنلاین خبرې اترې - Pashto Online Conversations, designed for AI model training. It contains over 100,000 natural, simple, and engaging Pashto online conversations, covering daily chats, jokes, fictional stories, and friendly exchanges among online friends. The dataset emphasizes naturalness, with conversations derived from real daily language; diversity, ranging from simple humor to deep philosophical discussions; completeness, ensuring two-way and full dialogues; and cultural alignment, matching Pashto culture, traditions, and values. Data is stored in JSONL format, with each record containing a list of conversations, each with from (sender type, e.g., human or gpt) and value (dialogue text) fields. The dataset is categorized by topics, including humor/jokes, movies/TV, music, food/cooking, general chat, art/DIY, travel/adventure, and intellectual discussions. It undergoes rigorous quality control, including deduplication, cleaning, standardization, and supports resume capability. The dataset is a translated version of English online conversations into Pashto, preserving naturalness, emojis, and cultural adaptation. It is suitable for text-generation tasks and is licensed under Apache 2.0.