Data Commons: Democratizing Access to AI Data

Data Commons is a game-changer in the world of artificial intelligence. It's a platform that makes high-quality, diverse, and inclusive data accessible to everyone.

By providing a centralized repository of data, Data Commons aims to democratize access to AI data, allowing researchers, developers, and organizations to tap into a vast pool of information. This is especially crucial for underrepresented communities, who often lack access to quality data.

Data Commons is not just a repository, but also a community-driven effort to ensure that data is used responsibly and ethically. By fostering collaboration and knowledge-sharing, Data Commons helps to create a more inclusive and equitable AI ecosystem.

A unique perspective: Google Sheet Highlight - and in Cells

Data Commons is a collaborative platform where researchers, policymakers, and developers can share and access open data. It's like a big library where you can find and use data from various sources.

Data Commons was founded by the University of California, Berkeley, and is now a project of the California Institute for Quantitative Biosciences. This means they have a strong foundation in research and development.

The platform aggregates data from various sources, including the US Census Bureau, the National Institutes of Health, and the National Science Foundation. This makes it easier to find and compare data across different fields and disciplines.

If this caught your attention, see: How to Find Google Accounts

A data commons is a shared repository of data that's been made available for anyone to use, often for research or innovation purposes. This is in contrast to traditional data systems, which are often isolated and inaccessible to others.

The goal of a data commons is to provide a collaborative environment where researchers, developers, and other stakeholders can access, share, and build upon existing data sets.

By doing so, data commons can accelerate innovation, improve research outcomes, and foster a culture of sharing and collaboration.

Data commons can be built on top of existing data platforms, such as databases or data warehouses, and can be tailored to meet the specific needs of a particular community or industry.

A good example of a data commons is the National Institutes of Health's (NIH) data commons, which provides access to a vast repository of biomedical data for researchers to use.

Recommended read: How Often Does Google Crawl a Site

Building a data commons is crucial because it allows for the sharing and reuse of data across different organizations and communities, promoting collaboration and innovation.

Data commons enable the creation of a single, unified dataset that can be accessed and utilized by multiple stakeholders, reducing the need for redundant data collection and analysis.

By leveraging a shared data commons, organizations can avoid the costs and inefficiencies associated with duplicating data collection efforts.

A data commons can also facilitate the development of new insights and discoveries by combining data from diverse sources, as seen in the example of the Human Connectome Project, which integrated data from multiple institutions to advance our understanding of brain function.

This leads to a more efficient use of resources and a faster pace of innovation, ultimately benefiting society as a whole.

The data commons can also provide a framework for addressing issues related to data quality, governance, and security, ensuring that the data being shared is accurate, reliable, and protected.

By establishing clear guidelines and standards for data sharing, a data commons can promote trust and confidence among stakeholders, enabling the free flow of data and ideas.

Related reading: Point Multiple Places on Google Maps

Data Commons is a powerful tool that places a strong emphasis on statistical data, which is a departure from the typical focus of linked data and knowledge graph initiatives.

It includes a wide range of data categories, such as geographical, demographic, weather, and real estate data, making it a valuable resource for researchers and analysts.

Data Commons represents data as semantic triples, each with its own provenance, which allows for a high level of accuracy and reliability.

The platform centers on entity-oriented integration of statistical observations from public datasets, making it easy to compare and contrast different data points.

Data Commons supports a subset of the W3C SPARQL query language, but its APIs also include tools like a Pandas dataframe interface that are specifically designed for data science, statistics, and data visualization.

Data Commons is an integrative platform, meaning it doesn't host datasets itself, but rather consolidates information from multiple datasets into a single data graph, making it easy to access and analyze large amounts of data.

Here are some of the key data categories included in Data Commons:

Data Commons is built on a graph data-model, which can be accessed through a browser interface and several APIs. This allows users to easily load and interact with the data.

The graph is also expanded through loading data, typically in CSV and MCF-based templates. This flexibility makes it easy to add new data to the graph.

Data Commons is integrated with Google, allowing users to access the graph through natural language queries in Google Search. This makes it easy to find and use the data without needing to learn complex queries.

The data vocabulary used to define the graph is based on Schema.org, which provides a standardized way of describing data. This ensures that the data is consistent and easy to understand.

Here are some key technologies and partners involved in Data Commons:

Retrieval Augmented Generation (RAG) is a powerful approach that retrieves relevant information from Data Commons before the LLM generates text, providing it with a factual foundation for its response. This approach is only possible because of Gemini 1.5 Pro's long context window, which allows appending the user query with extensive Data Commons data.

The average input length of data returned from broad queries is a staggering 38,000 tokens, with a maximum input length of 348,000 tokens. This emphasizes the need for a robust architecture like RAG to handle large amounts of data.

Here's how RAG works:

RAG generates fine-grained natural language questions answered by DC, which are then provided in the prompt to produce the final response. This results in more accurate and transparent answers, with a link to the source data and metadata in Data Commons for verification.

Data Commons is built on a graph data-model, which can be accessed through a browser interface and several APIs. This makes it easy to integrate with other systems and applications.

The graph is expanded through loading data, typically in CSV and MCF-based templates. This allows users to add new data to the graph and make it more comprehensive.

The data vocabulary used to define the graph is based on Schema.org, which is a widely-used standard for data markup. Specifically, the StatisticalPopulation and Observation terms were proposed to Schema.org to support datacommons-like use cases.

Intriguing read: Extract Sheets from Google Sheet Based on Criteria

The project's software is available on GitHub under the Apache 2 license, making it open-source and free to use. This is a great advantage for developers who want to contribute to or build upon the project.

Data Commons is a foundation for factual AI, containing over 250 billion global data points across hundreds of thousands of statistical variables, sourced from trusted organizations.

This broad and openly available repository provides a rich foundation for building more grounded and reliable AI, covering a wide range of topics from economics and climate change to health and demographics.

Data Commons is Google's publicly available knowledge graph, which continues to expand its global coverage, making data AI-ready.

DataGemma, a significant step forward in grounded AI, fine-tunes Gemma 2 to identify statistics within its responses and annotate them with a call to Data Commons, including a relevant query and the model's initial answer for comparison.

For more insights, see: Openshift Data Foundation

By leveraging Data Commons, DataGemma can double-check its work against a trusted source, increasing accuracy and transparency in its responses.

Data Commons provides ways for third parties to set up their own Data Commons instances, making it a flexible and extensible foundation for AI accuracy.

Here are some key features of Data Commons:

Data Commons has a robust set of repositories that make its technology and architecture accessible to the public. The Data Commons website is built on TypeScript and is licensed under Apache 2.0.

The website has had 28 commits in the past year, with 107 changes made to the code. The data repository is built with HTML and has also had 70 commits in the past year, with 123 changes made to the code.

The schema repository is built with Python and has had 19 commits in the past year, with 29 changes made to the code. The mixer repository, which provides the translator engine and API interface to access the Data Commons graph, is built with Go and has had 16 commits in the past year.

Discover more: How to Track Changes in Google Spreadsheet

Here is a breakdown of the repositories and their activity:

The Data Commons repositories are regularly updated, with the most recent updates occurring on October 1, 2025, and September 30, 2025.

Curating and harmonizing data is a crucial step in creating a successful data commons. It's time-consuming, expensive, and labor-intensive, but it's essential to produce data products of broad interest to the community.

The value of data commons lies in centralizing this work, so it can be done once instead of many times by each group that needs the data. This approach makes it easier to reuse and build upon existing data.

Curation and harmonization involve making data accessible and usable, which can be a challenge, especially when working with diverse and complex datasets. It's easy to overlook this step and expect the data to be ready for use, but that's not how it works.

By curating and harmonizing data, you can ensure that it's accurate, complete, and consistent, making it more valuable and useful to the community. This process can be done once, and then the data can be reused and built upon, saving time and resources in the long run.

Consider reading: Once in a Blue Moon Google Easter Egg

Data Commons has had a significant real-world impact, thanks in part to its partnership with The ONE Campaign, a global organization working to create investments for economic opportunities and healthier lives in Africa.

The ONE Campaign contributed substantially to Data Commons, advocating for and proposing the design of the platform. They even coded the client library to make Data Commons' insights available to data scientists and analysts.

This collaboration perfectly exemplifies Data Commons' goal of encouraging community contributions and enabling innovative uses of the platform.

Related reading: Que Es Google One Y Para Que Sirve

Our partnership with ONE.org is a prime example of how Data Commons is making a real-world impact. The ONE Campaign, a global organization, was instrumental in shaping this milestone.

The ONE Campaign advocated for and proposed the design of Data Commons to make its rich insights available to data scientists and analysts. This collaboration perfectly exemplifies Data Commons' goal of encouraging community contributions and enabling innovative uses.

ONE.org also coded the client library to integrate Data Commons with Python analytical tools and libraries, expanding its ecosystem.

Here's an interesting read: Nexus One

Gemma is a family of lightweight, state-of-the-art, open models built from the same research and technology used to create our Gemini models.

DataGemma connects LLMs to Data Commons' real-world data, harnessing the knowledge of Data Commons to enhance LLM factuality and reasoning.

By leveraging innovative retrieval techniques, DataGemma helps LLMs access and incorporate into their responses data sourced from trusted institutions.

These institutions include governmental and intergovernmental organizations and NGOs, mitigating the risk of hallucinations and improving the trustworthiness of their outputs.

DataGemma utilizes the natural language interface of Data Commons to ask questions, instead of needing knowledge of the specific data schema or API of the underlying datasets.

We use two different approaches, Retrieval Interleaved Generation (RIG) and Retrieval Augmented Generation (RAG), to train the LLM to know when to ask.

You might like: Important Interview Questions

The NCI Genomic Data Commons is a prime example of a successful data commons, containing over 2.9 PB of curated cancer genomics data from over 60 projects.

It's accessed by over 50,000 unique researchers each month, which is roughly the same number of members in the American Association of Cancer Research. Over 1.5 PB of data are accessed every month.

The data from different projects is curated with respect to a single data model, making it easier for researchers to analyze and integrate the data. This is a stark contrast to traditional data repositories, where data from different projects is often analyzed by different groups using different pipelines.

The GDC makes it easy for researchers to access its data and make new research discoveries with much less effort than if they were to analyze the raw data themselves. This is a major reason for its popularity.

A successful data commons is one that targets a specific user community and addresses their specific research challenges. The NCI Genomic Data Commons is a partnership between data scientists and disciplinary scientists with research challenges in cancer genomics.

The GDC has facilitated over 100 high-impact publications, demonstrating its impact on the cancer genomics research community.

Discover more: Get Link to Specific Sheet in Google Sheets

We're at the forefront of a revolution in AI, and you're invited to join the journey. Data Commons is an open-source platform that's already making a real-world impact, thanks to its partnership with ONE.org.

ONE.org advocated for and contributed to Data Commons' design and development, making its rich insights available to data scientists and analysts worldwide. This collaboration is a great example of how community contributions can shape the future of AI.

We're not just talking about a new tool or technology – we're talking about a fundamental shift in how AI is developed and used. By grounding Large Language Models (LLMs) in real-world data from Data Commons, we can unlock new possibilities for AI and create a future where information is not only intelligent but also grounded in facts and evidence.

Our research paper goes into more detail on the research behind DataGemma, and we encourage you to dive in and explore it. We're also excited to see how researchers will extend this work beyond our specific implementation with Data Commons.

See what others are reading: Google Spreadsheet Share Only One Tab

Data Commons itself provides ways for third parties to set up their own instances, making it easy to replicate and build upon our work. We're looking forward to further research and exploration in this space.

To get started, you can download the DataGemma models from Hugging Face or Kaggle (RIG, RAG). Try our quickstart notebooks for both the RIG and RAG approaches to get a hands-on introduction to using DataGemma and exploring its capabilities.

Here's a summary of the data sources, datasets, and dataset characteristics available through Data Commons: