A creativity-support tool for poetry.
Low Phreak (Low Frequency Phonaskos) discourages the poet from using common, high-frequency words. It recommends alternatives---semantically-related low-frequency words.
Try live version in this ☞Google Colab notebook☜.
This repo houses data that this notebook uses (via wget).
The basis of Low Phreak is a word2vec model trained on pos-tagged sentences (e.g. ["the_DT", "dog_NN",...]) extracted from Project Gutenberg text and Amazon product reviews.
See also:
prepping_data_for_training.ipynb: tokenizing and pos-tagging sentences, counting tokens to identify low-frequency ones.trainword2vecmodel.ipynb(Google Colab): training the word2vec model on sentences from the previous notebook.