26medias / context-aware-markov-chains

Markov Chain combined with word vector embedding (word2vec) and part-of-speech tagging, for context-aware text generation. License: MIT
98 stars 9 forks source link

Synonym alternative phrasing #3

Open cphoover opened 7 years ago

cphoover commented 7 years ago

Would love to be able to use synonyms from WordNet

So same algorithm but try with variances of Each word with multiple synonyms

So if I provide a seed input. "Merry christmas"

It might give all kinds of output

"Jolly Dec 25 to you and yours"

Just brainstorming

cphoover commented 7 years ago

So this could generate different phrasing through

n-gram calculate cartesian product w/ synonyms (see if it exists in graph if so you have a successful alternative phrasing) n-gram - 1word calculate cartesian product w/ synonyms finish rest of sentence with markov chain algorithm described n-gram - 2word... etc ... repeat.

26medias commented 7 years ago

I was planning to use a pre-trained word2vec to find alternative words and subjects while staying in context. I stopped working on this before I had a chance to play with it.

I had used wordnet early in the project to find synonyms, but the results were not great in practice. My approach was more simple than the one you proposed. I'd be curious to see the output. Not sure when I'll have the time to try this...