You can find the current version of this code here.
Based on WVTM, a contribution to 2016 NaNoGenMo.
This repository contains the code and data necessary to generate Enivrez-vous_modulé, Baudelaire's prose poem modified with analogies for each word in the original text based on the word pair bénir / maudire and on word vectors derived from 117 texts written by Honoré de Balzac.
The following command
./transformText.py bénir maudire Balzac Baudelaire_Enivrez-vous.md
will produce this text.
You can perform a comparable generated text with a word vector model based on 69 texts written by George Sand:
./transformText.py bénir maudire Sand Baudelaire_Enivrez-vous.md
(In this repository the only options for word vector models are "Balzac" and "Sand": anything else will produce an error.)
Using gensim to build a word2vec model based on a corpus of French texts , the code takes a pair of words (e.g. "homme" and "femme") and a text as parameters to generate a new text. Each word in the original text is replaced by a word that is "most similar" to it according to the word pair. For instance, if "roi" is a word in the original text, it would be replaced by analogy:
>>> model.most_similar(positive=['femme', 'roi'], negative=['homme'], topn=1)
[(u'reine', 0.8085041046142578)]
If the word vector model is unable to complete an analogy, the word from the original text does not change in the new text.
Handling verb conjugations and adjective agreements computationally in French is tricky but the code produces a mostly readable text needing grammatical polishing (a good exercise for students): the corrigé files are examples. The code can generate a new text from any text against any pair of words.