misha345a / E-commerce_Reviews_Classifier

Text augmentation, deep learning, and aspect-based sentiment analysis.

E-commerce Reviews Classification

Dataset

The dataset used is from https://www.kaggle.com/nicapotato/womens-ecommerce-clothing-reviews.

Notebooks

This repo contains several notebooks:

• EDA contains visualizations and analysis gathered from the dataset.
• TF-IDF_Random_Forest.ipynb includes a RF model trained on TF-IDF and optimized with grid-search.
• Create_Balanced_Dataset.ipynb cleans and balances the dataset using two different methods: resampling and text augmentation.
• LSTM_Models_And_Attack.ipynb trains an LSTM neural network and evaluates its performance after a textual adversarial attack.
• ASBA.ipynb demonstrates aspect-based sentiment prediction to find relevant business insights.

Published Article

https://towardsdatascience.com/nlp-project-with-augmentation-attacks-aspect-based-sentiment-analysis-3342510c90e7

Attack on a Trained LSTM Model!