DiyRex/emobooks-dataset

**emoBooks**数据集是一个精心策划的对话样本集合，旨在训练AI模型进行**情感感知的书籍推荐**。该数据集专注于理解用户情绪，并提供“Singlish”（音译的僧伽罗语）中的相关书籍建议。数据集遵循**匹配/切换**逻辑：- **匹配**：推荐与用户当前情绪状态相符的书籍。- **切换**：推荐帮助用户转换到更积极情绪状态的书籍（例如，从悲伤到快乐）。数据集采用标准的**ChatML/OpenAI格式**，非常适合对Llama-3、Mistral或Gemma等模型进行指令微调。每个条目都是一个包含`messages`列表的JSON对象。数据集涵盖8种主要情绪状态和多种书籍类型，包括儿童故事、小说、翻译作品、传记和短篇故事。该数据集是通过自定义流程合成的，旨在用于微调大型语言模型（LLM）以用于推荐系统、情感感知对话AI研究以及开发双语（僧伽罗语/英语）文学助手。

The **emoBooks** dataset is a curated collection of conversational samples designed to train AI models for **emotion-aware book recommendations**. It focuses on understanding user emotions and providing relevant book suggestions in "Singlish" (transliterated Sinhala). The dataset follows a **Match/Switch** logic: - **Match**: Recommend books that align with the users current emotional state. - **Switch**: Recommend books that help transition the user to a more positive emotional state (e.g., from Sadness to Joy). The dataset is provided in a standard **ChatML/OpenAI format**, making it ideal for instruction fine-tuning of models like Llama-3, Mistral, or Gemma. Each entry is a JSON object with a `messages` list. The dataset covers 8 primary emotional states and various book genres, including Childrens Stories, Novels, Translations, Biographies, and Short Stories. This dataset was synthetically generated using a custom pipeline and is intended for fine-tuning Large Language Models (LLMs) for recommendation systems, research in emotion-aware conversational AI, and developing bilingual (Sinhala/English) literary assistants.