Generate words from semantic neighborhoods

[jbrew]

A. Flat distribution

1. Given a GloVe word vector, identify the k closest words in GloVe space.
2. Return new words that are phonetically similar to this list of k words.

B. Normal distribution

1. Given a GloVe word vector, identify the k closest words in GloVe space and their distance.
2. Return new words that are phonetically similar to this list of k words with a bias towards closer words, according to a normal distribution centered on the original word vector, with a second parameter for the standard deviation of the distribution.