sumukha

Project sumukha - This is the official repository for building domain adapted embeddings for robust NLP tasks

---

Motivation

Generic Embeddings for text representation may not be accurate in every domain of interest, this project is to leverage knowledge from the domain.

Table of Contents

• Description
• How to Use
• Author Info

---

Description

sumukha uses the state of the art techniques like standard preprocessing techniques on texts, fasttext models to train embeddings, scikit-learn decomposition libraries for further processing.

How to Use

• Run pip install . to install all the requirements.
• Run sumukha -r ./ preprocess --input dataset_path --output preprocess_path
• Run sumukha -r ./ train --input preprocess_path --output trained_results_path
• Run sumukha -r ./ encode --input preprocess_path --gen_emb_path general_embeddings_path --dom_emb_path domain_embeddings_path

Back to The Top