#168 · Primary category: NLP Tools & Text Processing

SceneGraphParser

nlp

A python toolkit for parsing captions (in natural language) into scene graphs (as symbolic representations).

Project last updated:01/23/24

GitHub Stars

593

Forks

53

Contributors

5

License

MIT

Why we included this project

SceneGraphParser turns a plain English sentence into a scene graph: noun phrases become nodes, with their determiners and adjectives kept as modifiers, and the relations between them become edges. The output comes back as ordinary Python dicts and lists, so you can feed it straight into your own code without a conversion layer. The parser applies a set of human-written rules to spaCy's dependency trees, which makes it transparent and easy to debug, though the project itself warns that corner cases are still being collected. It grew out of a CVPR 2019 paper on unified visual-semantic embeddings.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category