bert-text-classificaiton-arxiv

AI or not AI? Classifying ArXiv articles with BERT

Installation

Prerequisites

• Python ≥ 3.6

Provision a Virtual Environment

Create and activate a virtual environment (conda)

conda create --name py36_bert-arxiv python=3.6
source activate py36_bert-arxiv

If pip is configured in your conda environment, install dependencies from within the project root directory

pip install -r requirements.txt

Get ArXiv dataset

The dataset used in this repository should be downloaded from Kaggle

Create a folder data from within the project root directory. Place the downloaded file arxivData.json in the data folder.

Feature Extraction code

Now that the environment is setup and the dataset is available, you can run the code using the following command:

python feature_extraction.py 

This will by default use the arxivData.json file as input and generate in the same data folder the X,y training and test files:

model training

Use the jupyter notebook run_model_keras to train the model. This is easier to visualise the results we get.