Financial World Archive - 1902-1930

Complete pre-1930 archive of Financial World, an early 20th-century financial newspaper exposing investment fraud, stock manipulation, and market speculation. **121,354 rows** of clean, structured text documenting the rise of modern American finance. **What this data does for your model:** - Your model learns authentic early 20th‑century investigative financial journalism, exposing frauds and warning investors. - Your model retrieves original reporting on bucket shops, stock swindlers, fake mining companies, and "guaranteed return" schemes. - Your model trains on the language of market manipulation, learning to recognize red flags in investment promotions. - Your model understands the watchdog voice of financial reporting before SEC regulation, when journalists were the primary check on fraud. **What's inside:** - Exposés of fraudulent investment schemes and stock promoters - The Texas oil boom, Beaumont, Spindle Top, J.P. Morgan vs. Rockefeller - Mining fraud - Thunder Mountain, Colorado promoters - Railroad finance and industrial consolidation - Early consumer protection journalism **Perfect for:** - LLM fine-tuning on financial history and journalism - Fraud detection AI training (historical patterns) - Economic history and market research - Investigative journalism and media studies **Format:** Snowflake-native JSONL with columns: ISSUE, TITLE, AUTHOR, TYPE, TEXT. Fully cleaned, bias-audited, and ready for AI training. *From the 1902 Texas oil boom to the eve of the Great Depression - financial journalism that warned investors, now ready for AI.* ***Cortex Agent Prompts:*** 1. Analyze the historical evolution of global financial markets, investment strategies, and economic reporting as documented within this archive to support the training of domain-specific AI models focused on the history of international finance. 2. Extract longitudinal trends in market volatility, corporate fiscal performance, and regulatory developments from this corpus to assist in the creation of RAG applications for financial research and comparative economic analysis. 3. Evaluate the expert discourse on economic policy, banking practices, and fiscal management within this dataset to provide a foundational baseline for NLP tasks concerning the progression of global financial standards throughout the 20th century. <p><br/></p>

原始信息汇总

Financial World Archive - 1902-1930 数据集概述

数据集基本信息

• 数据集名称: Financial World Archive - 1902-1930
• 副标题: Early 20th-Century Financial Journalism - 1902 to 1930
• 提供商: Devin Media Corp.
• 数据规模: 121,354 行
• 数据格式: Snowflake-native JSONL
• 数据列: ISSUE, TITLE, AUTHOR, TYPE, TEXT, INGESTION_DATE
• 数据描述: 完整的前1930年《金融世界》档案，这是一份20世纪初的金融报纸，揭露投资欺诈、股票操纵和市场投机。包含干净、结构化的文本，记录了现代美国金融的崛起。

数据内容详情

主要内容涵盖

• 欺诈性投资计划和股票推广的曝光
• 德克萨斯石油繁荣、博蒙特、斯宾德尔托普、J.P.摩根与洛克菲勒
• 采矿欺诈 - 雷山、科罗拉多州推广者
• 铁路金融和工业整合
• 早期消费者保护新闻业

适用场景

• 金融历史和新闻业的LLM微调
• 欺诈检测AI训练（历史模式）
• 经济史和市场研究
• 调查性新闻和媒体研究

数据结构与质量

• 数据表: FW_CORPUS
• 数据列说明:
  • ISSUE: Varchar 类型
  • TITLE: Varchar 类型
  • AUTHOR: Varchar 类型
  • TYPE: Varchar 类型
  • TEXT: Varchar 类型
  • INGESTION_DATE: Timestamp_NTZ 类型
• 数据质量: 经过完全清洗、偏见审计，可直接用于AI训练。

商业需求应用

机器学习

在121,000多行经过整理的20世纪初金融文本上训练、微调和部署机器学习模型。适用于欺诈检测模式识别、金融术语提取和历史市场分析。

真实世界数据

利用历史记录的投资计划、股票操纵和市场泡沫作为研究和分析的真实世界数据。该档案记录了早期美国金融的阴暗面。

生命科学商业化

通过记录投资欺诈、市场监管和消费者保护演变过程的整理历史新闻，支持金融研究。

使用示例

搜索欺诈警告

sql SELECT ISSUE, TITLE FROM FW_CORPUS WHERE TYPE = article AND TEXT ILIKE %fraud% OR TEXT ILIKE %swindler% OR TEXT ILIKE %promoter% LIMIT 10;

按类型统计行数

sql SELECT TYPE, COUNT(*) FROM FW_CORPUS GROUP BY TYPE;

搜索德克萨斯石油繁荣

sql SELECT TITLE, ISSUE FROM FW_CORPUS WHERE TYPE = article AND TEXT ILIKE %texas% OR TEXT ILIKE %oil% OR TEXT ILIKE %beaumont% LIMIT 10;

试用与定价

• 试用: 提供7天限时试用，可完整访问1902–1930年所有121,354行数据。
• 定价: 获取完整访问权限需联系提供商获取许可证。

数据维护与覆盖

• 更新频率: 每年
• 地理覆盖范围: 美国
• 云区域可用性 (AWS): 加拿大（中部）、美国东部（弗吉尼亚北部）、美国东部（俄亥俄）、美国西部（俄勒冈）等。

提供商信息

• 提供商: Devin Media Corp.
• 专业领域: 为AI训练提供优质历史数据，提供全面、来源可追溯、经过偏见审计、1930年以前的出版物和档案，经过专业清洗和结构化处理，适用于机器学习应用。
• 数据集特点: 1930年以前，经验证的公共领域/无版权；经过专业OCR和深度清洗；来源可追溯且经过偏见审计；格式为JSONL，适合AI使用；通过安全API交付（无文件下载）。
• 联系方式:
  • 销售: [e-mail]
  • 支持: [e-mail]