{"id":249001,"date":"2025-07-10T04:52:45","date_gmt":"2025-07-09T18:52:45","guid":{"rendered":"https:\/\/science.nasa.gov\/science-research\/artificial-intelligence-metadata-tagging\/"},"modified":"2025-07-10T04:52:45","modified_gmt":"2025-07-09T18:52:45","slug":"smarter-searching-nasa-ai-makes-science-data-easier-to-find","status":"publish","type":"post","link":"https:\/\/www.vibewire.com.au\/?p=249001","title":{"rendered":"Smarter Searching: NASA AI Makes Science Data Easier to Find"},"content":{"rendered":"
\n
\n
\n

6 min read<\/p>\n

Smarter Searching: NASA AI Makes Science Data Easier to Find<\/h1>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\"Image<\/a><\/figure>
\n
Image snapshot taken from NASA Worldview of NASA\u2019s Global Precipitation Measurement (GPM) mission on March 15, 2025 showing heavy rain across the southeastern U.S. with an overlay of the GCMD Keyword Recommender for Earth Science, Atmosphere, Precipitation, Droplet Size.<\/div>\n
NASA Worldview<\/div>\n<\/figcaption><\/div>\n<\/div>\n<\/div>\n

Imagine shopping for a new pair of running shoes online. If each seller described them differently\u2014one calling them \u201csneakers,\u201d another \u201ctrainers,\u201d and someone else \u201cfootwear for exercise\u201d\u2014you\u2019d quickly feel lost in a sea of mismatched terminology. Fortunately, most online stores use standardized categories and filters, so you can click through a simple path: Women\u2019s > Shoes > Running Shoes<\/em>\u2014and quickly find what you need.<\/p>\n

Now, scale that problem to scientific research. Instead of sneakers, think \u201caerosol optical depth\u201d or \u201csea surface temperature.\u201d Instead of a handful of retailers, it is thousands of researchers, instruments, and data providers. Without a common language for describing data, finding relevant Earth science datasets would be like trying to locate a needle in a haystack, blindfolded.<\/p>\n

That\u2019s why NASA created the Global Change Master Directory (GCMD)<\/a>, a standardized vocabulary that helps scientists tag their datasets in a consistent and searchable way. But as science evolves, so does the challenge of keeping metadata organized and discoverable.\u00a0<\/p>\n

To meet that challenge, NASA\u2019s Office of Data Science and Informatics (ODSI) at the agency\u2019s Marshall Space Flight Center (MSFC) in Huntsville, Alabama, developed the GCMD Keyword Recommender (GKR)<\/a>: a smart tool designed to help data providers and curators assign the right keywords, automatically.<\/p>\n

Smarter Tagging, Accelerated Discovery<\/h2>\n

The upgraded GKR model isn\u2019t just a technical improvement; it\u2019s a leap forward in how we organize and access scientific knowledge. By automatically recommending precise, standardized keywords, the model reduces the burden on human curators while ensuring metadata quality remains high. This makes it easier for researchers, students, and the public to find exactly the datasets they need.<\/p>\n

It also sets the stage for broader applications. The techniques used in GKR, like applying focal loss to rare-label classification problems and adapting pre-trained transformers to specialized domains, can benefit fields well beyond Earth science.<\/p>\n

Metadata Matchmaker<\/h2>\n

The newly upgraded GKR model tackles a massive challenge in information science known as extreme multi-label classification. That\u2019s a mouthful, but the concept is straightforward: Instead of predicting just one label, the model must choose many, sometimes dozens, from a set of thousands. Each dataset may need to be tagged with multiple, nuanced descriptors pulled from a controlled vocabulary.<\/p>\n

Think of it like trying to identify all the animals in a photograph. If there\u2019s just a dog, it\u2019s easy. But if there\u2019s a dog, a bird, a raccoon hiding behind a bush, and a unicorn that only shows up in 0.1% of your training photos, the task becomes far more difficult. That\u2019s what GKR is up against: tagging complex datasets with precision, even when examples of some keywords are scarce.<\/p>\n

And the problem is only growing. The new version of GKR <\/a>now considers more than 3,200 keywords, up from about 430 in its earlier iteration. That\u2019s a sevenfold increase in vocabulary complexity, and a major leap in what the model needs to learn and predict.<\/p>\n

To handle this scale, the GKR team didn\u2019t just add more data; they built a more capable model from the ground up. At the heart of the upgrade is INDUS<\/a>, an advanced language model trained on a staggering 66 billion words drawn from scientific literature across disciplines\u2014Earth science, biological sciences, astronomy, and more.<\/p>\n

\n
\n
\n
\"Image<\/a><\/figure>
\n
NASA ODSI\u2019s GCMD Keyword Recommender AI model automatically tags scientific datasets with the help of INDUS, a large language model trained on NASA scientific publications across the disciplines of astrophysics, biological and physical sciences, Earth science, heliophysics, and planetary science.<\/div>\n
NASA<\/div>\n<\/figcaption><\/div>\n<\/div>\n<\/div>\n

\u201cWe\u2019re at the frontier of cutting-edge artificial intelligence and machine learning for science,\u201d said Sajil Awale, a member of the NASA ODSI AI team at MSFC. \u201cThis problem domain is interesting, and challenging, because it\u2019s an extreme classification problem where the model needs to differentiate even very similar keywords\/tags based on small variations of context. It\u2019s exciting to see how we have leveraged INDUS to build this GKR model because it is designed and trained for scientific domains. There are opportunities to improve INDUS for future uses.\u201d<\/p>\n

This means that the new GKR isn\u2019t just guessing based on word similarities; it understands the context in which keywords appear. It\u2019s the difference between a model knowing that \u201cprecipitation\u201d might relate to weather versus recognizing when it means a climate variable in satellite data.<\/p>\n

And while the older model was trained on only 2,000 metadata records, the new version had access to a much richer dataset of more than 43,000 records from NASA\u2019s Common Metadata Repository<\/a>. That increased exposure helps the model make more accurate predictions.<\/p>\n

The Common Metadata Repository is the backend behind the following data search and discovery services:<\/p>\n