LSTM Text Classification using PyTorch for App Reviews

Overview

Long Short-Term Memory (LSTM) is a type of recurrent neural network (RNN) architecture used in deep learning. LSTMs are specifically designed to handle long-term dependencies in data, making them well-suited for tasks involving text data, speech, and time series. In this project, we build an LSTM model to classify app reviews on a scale of 1 to 5 based on user feedback using PyTorch.

---

Aim

To build text classifier to classify app reviews on a scale of 1 to 5 using LSTM.

---

Data Description

The dataset consists of app reviews and corresponding ratings. The "score" column contains ratings in the range of 1 to 5, and the "content" column contains the review text.

---

Tech Stack

• Language: Python
• Libraries: pandas, TensorFlow, matplotlib, scikit-learn, NLTK, NumPy, PyTorch

---

Approach

Data Preprocessing

1. Lowercasing text, removing punctuation, and eliminating links.
2. Balancing classes.
3. Tokenizing the text.
4. Scaling the data.

Model

• Training an LSTM model in PyTorch.

Model Evaluation

• Evaluating the model on test data.

---

Modular Code Overview

1. Input: Contains the data used for analysis, including:

   • [List of data files]

2. ML_Pipeline: This folder contains functions distributed across multiple Python files, each appropriately named for its functionality. These functions are called from the Engine.py file.

3. Notebook: Contains the Jupyter Notebook file of the project.

4. Engine.py: The main script that orchestrates the different parts of the project by calling functions from the ML Pipeline.

5. Readme.md: Instructions for running the code and additional information about the project.

6. requirements.txt: Lists all the required libraries and their versions for easy installation using pip.

---