About

This is my solution for the Kaggle Real or Not? NLP with disaster tweets getting started competition.

Models

RNN Model

BERT Model

Guide

Dependency Installation

pip install -r requirements.txt

Running program

All commands are executed using

python main.py

and using the flags

Preprocessing data

Preprocessing carries out some cleanup of the data. Run using:

python main.py --mode=preprocess

Training standard models

Training BERT models

First run the BERT preprocessing

python main.py --mode=preprocess_bert

This will generate the training data in a python shelve file created at ./tmp/bert_data.

Next run the training program

python main.py --mode=train_bert