Earth Observation · nasa-impact
EO-via-NLP
A dataset pairing Earth observation data and text to support natural language interfaces and entity recognition.
What EO-via-NLP is
EO-via-NLP is an Earth science dataset developed by researchers at NASA-IMPACT. It pairs Earth observation imagery with natural-language descriptions to facilitate the development and evaluation of multimodal machine learning models designed for remote sensing understanding tasks.
The dataset addresses the challenge of making Earth observation data and computational workflows accessible via natural language. It provides structured examples intended for researchers developing agentic artificial intelligence systems, task-oriented models, and conversational interfaces for geoscience software.
How it was made
The dataset consists of English-language records provided in JSON format, designed to support named entity recognition (the automated identification of domain-specific entities in text) as well as multimodal grounding. All selections and annotations within the collection were compiled and curated manually by human annotators.
Key facts
- Language
- English
- Format
- JSON
- Primary task
- Named Entity Recognition (NER) in Earth science
- Domain
- Earth observation and climate science
What you can use it for
- Earth Science Named Entity Recognition
- Training and evaluating natural language processing models to identify and extract specialised geoscience concepts and entities from text.
- Natural Language Workflow Interfaces
- Building and assessing conversational interfaces that allow scientists to interact with Earth observation software and scientific workflows.
- Agentic AI System Research
- Benchmarking autonomous and task-oriented artificial intelligence agents operating within geoscience computational environments.
- Multimodal Remote Sensing Benchmarking
- Evaluating models that connect natural-language descriptions with remote sensing imagery and Earth observation data.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from nasa-impact/EO-via-NLP at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 107 rows of default/train
json_mapping
string_text- min
- 57
- median
- 66
- mean
- 65.96
- max
- 79
user_query
string_text- min
- 33
- median
- 61
- mean
- 62.59
- max
- 104
Strengths and limits
Good at
- All annotations and dataset selections were manually created and curated by human contributors.
- Focuses specifically on the intersection of natural language processing and Earth observation workflows.
- Supports both text-based entity extraction and multimodal benchmark evaluation.
Watch out for
- Available only in the English language.
Questions
Can I look inside EO-via-NLP without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is EO-via-NLP under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
What primary task does this dataset target?
It is primarily designed for named entity recognition (NER) in Earth science contexts, alongside benchmarking multimodal and agentic systems.
How was the dataset curated?
All selections and annotations were created directly by human contributors rather than automated extraction tools.
What formats and modalities are included?
The records are distributed in JSON format, containing textual natural-language descriptions paired with Earth observation data.
How to cite it
Ellis, M., Gurung, I., Ramasubramanian, M., & Ramachandran, R. (2025). Toward Open Earth Science as Fast and Accessible as Natural Language. arXiv preprint arXiv:2505.15690.
Written from EO-via-NLP's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.