SentimentClassifier

Overview

SentimentClassifier is a lightweight neural network tool designed to perform sentiment analysis on textual data. It highlights the sentiment of input texts using a pretrained model from HuggingFace. The project includes preprocessing, processing, and postprocessing steps to deliver human-presentable results.

Table of Contents

• Installation
• Usage
  • Using Makefile
  • Using Docker
• Examples
• Testing
• Project Structure

Installation

Requirements

• Make
• Docker
• Python 3.9+

Usage

Cloning repository

git clone https://github.com/AntonKorznikov/SentimentClassifierProject.git

Using Makefile

1. Install Dependencies

make prereqs

2. Build Executables

make build

3. Clean Executables

make clean

4. Run Tests

make test

5. Run Project

make run_all

Using Docker

1. Build Docker Image

docker build -t sentimentclassifier:latest .

2. Run Docker Container

docker run -v $(pwd)/examples/input_raw:/input_raw -v $(pwd)/examples/output:/output sentimentclassifier:latest

Examples

Sample input files are provided in the examples/input_raw/ directory. After running the preprocessing, processing, and postprocessing steps, the results can be found in the examples/output/ directory.

example of input_raw/sample_input.txt :

I love this product!
This service is terrible.
The quality is just okay.
Absolutely fantastic experience!
Not worth the investment at all.

example of output/sample_input.txt :

Sentiment: POSITIVE, Score: 0.9998855590820312
Sentiment: NEGATIVE, Score: 0.999573290348053
Sentiment: POSITIVE, Score: 0.9998542070388794
Sentiment: POSITIVE, Score: 0.9998812675476074
Sentiment: NEGATIVE, Score: 0.999803364276886

Testing

Run all tests using the Makefile:

make test

Project Structure

• preprocessing/: Contains scripts and tests for data preprocessing.
• processing/: Contains scripts and tests for running the sentiment analysis model.
• postprocessing/: Contains scripts and tests for processing the model's output.
• examples/: Contains example input and output files.