Indonesian News Title Category CLassifier

This is my final project for Information Retrieval Course

Want to give it a try? Try it now on

Dataset

Indonesian News Title by Ibrahim on Kaggle. This dataset contains 91,017 Indonesian News Title caregorized into 9 labels ('finance', 'food', 'health', 'hot', 'inet', 'news', 'oto', 'sport', 'travel').

Example

Date	Url	Title	Category
02/26/2020	https://finance.detik.com/berita-ekonomi-bisnis/d-4916114/kemnaker-awasi-tka-di-meikarta	Kemnaker Awasi TKA di Meikarta	finance
02/27/2020	https://sport.detik.com/detiktv/d-4916359/kangen-rooney-lihat-nih-gol-gol-kerennya	Kangen Rooney? Lihat Nih Gol-gol Kerennya	sport
04/22/2020	https://travel.detik.com/travel-news/d-4985787/jadikan-rumah-serasa-tempat-travelling-ini-tipsnya	Jadikan Rumah Serasa Tempat Travelling, Ini Tipsnya	travel

Result

Model	Test Accuracy	Train Loss	Validation Loss
Word2Vec + LSTM	89.16%	0.4327	0.4741
Word2Vec + CNN 1D	88.53%	0.4324	0.5803

Project Structure

├── datasets
│   ├── preprocessed
│   │   ├── data.pkl
│   └── indonesian-news-title.csv
├── models
│   ├── cnn.py
│   ├── lstm.py
│   ├── inference.py
│   └── w2v_matrix.pkl
├── utils
│   ├── preprocessor.py
│   └── tokenizer.pkl
├── app.py
├── train_cnn.py
├── train_lstm.py
├── requirements.txt
├── README.md
└── .gitignore

Streamlit Demo

streamlit run app.py

Author

Rizky Adi

Last update: 22 November 2022