Sentiment-Analysis

Building a text classification model using BERT and pytorch for sentiment analysis.

The model is trained and built for a corpus of Google Play Reviews. The dataset has been scraped using the google_play_scraper library. The models can be found on my google drive, and are not uploaded here due to space constraints.

Directory Structure

.
│   README.md
|   LICENSE
│   dataset-scraper.ipynb
│   text-preprocessing.ipynb
│   text-classification-bert.ipynb
│
└───data
│   │   apps.csv
│   │   reviews.csv
│   │
│   
└───models
    │   model.bin
    │   model_base_cased_state_842.bin
    │   model.pth

Usage

• The dataset-scraper.ipynb notebook has all the code required to scrape the reviews dataset
• The text-classification-bert.ipynb is the final notebook that has the code required to run the classification model.

Performance

Below are the performance metrics of the classification model

precision	recall	f1-score	support
negative	0.83	0.80	0.81
neutral	0.75	0.75	0.75
positive	0.85	0.88	0.86
accuracy	-	-	0.81
macro avg	0.81	0.81	0.81
weighted avg	0.81	0.81	0.81