Skip to content

Natural Language Inference problems specialized for spatial semantics & reasoning

License

Notifications You must be signed in to change notification settings

kovvalsky/SpaceNLI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SpaceNLI

Natural Language Inference problems specialized for spatial semantics & reasoning.

The NLI problem patterns and automatically generated NLI sample problems are under dataset/.

Generating data

The dataset of NLI sample problems is generated from the patterns (using the mini world with selection restrictions and a toy grammar) with the following command:

python/generate_nli.py --sr config/selection_restriction.yaml --gr python/base_grammar.fcfg dataset/problem_patterns.xml -v --out dataset/160x200.json --out-max-seed 200

Note that changes to the selection restriction file can result in a different sample dataset as the fixed number of problems (e.g., 200) per pattern are randomly drawn from the all possible samples per pattern.

Paper

Abzianidze, L., Zwarts, J., Winter, Y. (2023). SpaceNLI: Evaluating the Consistency of Predicting Inferences In Space. In Proceedings of the 4th Natural Logic Meets Machine Learning Workshop (NALOMA IV).

About

Natural Language Inference problems specialized for spatial semantics & reasoning

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages