pic2prose 0.0.2

Enables real-world data collection, bridges the gap between OCR and NLP, enabling you to convert text from any image to ready to use nlp data structures.

pic2prose

A package that can take in images and build a corpus and produce nlp datastructures for direct use in experimentation and model training.

Take any image with text, use p2p to generate NLP datastructures ready for use in fine-tuning LLM's, generating embeddings, sentiment classification, etc.

Installation

pip install pic2prose

Open up your favorite editor, import, and build a robust corpus.

from pic2prose.structures import *

# initialize the object
# may take longer if you're not using a GPU
corpus = Corp(image_path="ex1.png")

# generate co-occurrence matrix
corpus.get_co_occurrence_matrix()

# generate tf-idf matrix
corpus.get_tfidf_matrix()

# one-hot encodings
corpus.one_hot_encode()

Coming Soon

Support for building corpi from URL