Introducing Groundsource: Turning news reports into data with Gemini

2026-03-12 · Source: The latest research from Google · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

Google Research has introduced Groundsource, a new scalable methodology that uses the Gemini Large Language Model to convert unstructured global news reports into structured, historical data. The initial open-access Groundsource dataset focuses on urban flash floods, comprising 2.6 million records across more than 150 countries from 2000 to the present. This initiative addresses the critical scarcity of high-quality historical data for hydro-meteorological hazards, which traditionally lack standardized global sensor networks. Groundsource processes news in 80 languages, standardizes it to English via Cloud Translation API, and employs Gemini for classification, temporal reasoning, and spatial precision, achieving 82% practical accuracy in location and timing. This expanded dataset significantly enhances the ability to provide near-global urban flash flood forecasts up to 24 hours in advance, now being rolled out in Google's Flood Hub.

Key takeaway

For AI Scientists developing predictive models for natural disasters, Groundsource demonstrates a powerful approach to overcome historical data scarcity. You should consider integrating large language models like Gemini into your data pipeline to extract structured event data from unstructured news and reports. This method can significantly expand your training datasets, improving model accuracy and enabling more timely, localized forecasts for hazards like flash floods, and potentially other events lacking traditional sensor networks.

Key insights

Groundsource leverages Gemini to transform unstructured global news into structured historical data for disaster forecasting.

Principles

• Unstructured news is a rich source for historical event data.
• LLMs can systematically extract ground truth from diverse text.
• Data scarcity hampers accurate global-scale AI model training.

Method

Groundsource analyzes news reports, isolates primary text in 80 languages, translates to English, then uses Gemini for classification of events, temporal anchoring, and spatial mapping to Google Maps Platform polygons.

In practice

• Use Gemini for precise event classification and temporal reasoning.
• Apply Google Maps Platform for granular spatial grounding.
• Translate diverse language sources for global data coverage.

Topics

• Groundsource
• Gemini LLM
• Flash Flood Forecasting
• Natural Language Processing
• Geospatial Data

Best for: AI Scientist, Research Scientist, Software Engineer, AI Data Scientist

Related on AIssential

• Creating Custom Event Data Without Dictionaries: A Bag-of-Tricks — Modern annotation systems like Prodigy streamline custom event data generation, overcoming past tedious manual efforts.
• How Google DeepMind is researching the next Frontier of AI for Gemini — Raia Hadsell, VP of Research — DeepMind is advancing Frontier AI through omnimodal embeddings, superior weather prediction, and interactive world generation.
• Predicting Risk in Content Launches: How Data-Driven Insights can Transform Launch Planning — Netflix's boosted tree models significantly enhance content launch schedule accuracy by predicting media asset delivery dates from dynamic production data.
• Protecting cities with AI-driven flash flood forecasting — AI-driven flash flood forecasting using news data significantly enhances urban climate resilience globally.
• Google’s LangExtract Explained: Why Source Grounding Makes AI Extraction Trustworthy — LangExtract uses LLMs with source grounding to reliably extract verifiable structured data from unstructured text.
• Build a smart financial assistant with LlamaParse and Gemini 3.1 — Combining LlamaParse with Gemini models enables robust, accurate extraction and summarization of complex unstructured documents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The latest research from Google.