thorwhalen/newsmood-data
收藏资源简介：
newsmood-data数据集是newsmood项目生成的数据产物，用于评估美国股票上的新闻嵌入交易信号。数据集包含约36万篇新闻文章的嵌入向量（使用OpenAI的text-embedding-3-small模型生成的1536维向量）和规范新闻DataFrame，涵盖2025年1月至2026年5月（约17个月）的时间跨度。此外，还包括14个SPY-universe股票代码（如AAPL、AMD、AMZN等）的每日OHLCV（开盘价、最高价、最低价、收盘价、成交量）数据，以及原始JSON搜索快照。数据集主要用于金融、新闻、嵌入和交易时间序列分析，支持通过Python库或huggingface_hub直接访问。
The newsmood-data dataset consists of data artifacts produced by the newsmood project for evaluating news-embedding trading signals on US equities. It includes approximately 360,000 news article embeddings (1536-dimensional vectors generated using OpenAIs text-embedding-3-small model) and a canonical news DataFrame, covering a time span from January 2025 to May 2026 (approximately 17 months). Additionally, it contains daily OHLCV (Open, High, Low, Close, Volume) data for 14 SPY-universe tickers (e.g., AAPL, AMD, AMZN) and raw JSON search snapshots. The dataset is designed for finance, news, embeddings, and trading time-series analysis, accessible via Python libraries or directly through huggingface_hub.
