BERT-Powered News Text Classification

BERT, a cutting-edge transformer model, has proven its prowess in complex NLP tasks, such as NLU, NLG, Sentiment Analysis, and Text Classification. Its exceptional performance often surpasses human capabilities.

---

Objective

Conduct text classification on the AG News dataset using the state-of-the-art transformer model, BERT.

---

Data Description

The AG News dataset (AG’s News Corpus) is a subset of AG's corpus, constructed by combining titles and description fields from articles in the four major classes: "World," "Sports," "Business," and "Sci/Tech" of AG’s Corpus.

---

Installation

1. Install the required packages specified in the requirements.txt file.

   For Anaconda:

   conda create --name <yourenvname>
   conda activate <yourenvname>
   pip install -r requirements.txt

   For Python Interpreter:

   pip install -r requirements.txt

2. The entire repository is modularized into distinct sections, each handling specific tasks. Start by navigating to the src folder.

   Within src:

   • ML_Pipeline: Contains modules with function declarations for various Machine Learning tasks.
   • engine.py: The core of the project, orchestrating all function calls.

3. Run/Debug the engine.py file, and the necessary steps will be automated according to the defined logic.

4. Input datasets are stored in the input folder.

5. Predictions and models are saved in the output folder.

---