fowler.corpora
is software to create vector space models for distributional
semantics.
It is possible to instantiate a vector space from
- Brown corpus
- British National Corpus
- ukWaC and WaCkypedia
The weighting schemes include:
- PMI
- PPMI
- nITTF
The implemented experiments are:
- Word similarity
- Sentence similarity
- Documentation update: installation instructions, similarity experiment quick start.
- Correlation and Eucliedean similarities are computed.
- PMI variants and parameters.
- Frobenious operators.
- Word2vec space import.