A library of common NLP / compositionality metrics.
Implementations:
- RSA
- Topographic similarity
- Message distinctness
- Perplexity
- Jaccard Similarity
- Uniform and unigram chimeras
- Language Entropy
- TRE
- Pseudo-TRE
- Language Comparaison (edit distance / perfect message matches)
Tests Needed:
- RSA
- Topographic similarity
- Message distinctness
- Perplexity
- Jaccard Similarity
- Uniform and unigram chimeras
- Language Entropy
- TRE
- Pseudo-TRE
- Language Comparaison (edit distance / perfect message matches)
Potential features:
- Pytorch/Numpy tensor/ndarray agnostic
- Docs + Examples
- logging
- pip support
- more ?