Multi-label Movie Genres Classification From Original Movie Overview

Members

This project is a part of the course Natural Language Processing at the University of Information Technology

No	Student ID	Full name	Email
1	23520179	Phùng Minh Chí	[email protected]
2	23520183	Nguyễn Hữu Minh Chiến	[email protected]
3	23521467	Lê Ngọc Phương Thảo	[email protected]

Course Information

• Course Natural Language Processing
• Course code: CS221
• Class code: CS221.P22
• Semester: HK2 (2024 - 2025)
• Instructor: TS Nguyễn Trọng Chỉnh

Instruction

1. Clone the repository:

   git clone hhttps://github.com/chisphung/CS221-GenresPrediction-from-Overview

2. Install dependencies:

   pip install -r requirements.txt

Data preprocessing:

To preprocess the dataset, run the following command:

python tools/preprocess.py 

You can also download the preprocessed dataset with the following command:

python tools/download.py

Model training:

To train the BERT models, run the following command:

python - m tools.train <pretrained_model_name> <dataset_path>

Replace <pretrained_model_name> with the name of the pretrained model you want to use (e.g., bert-base-uncased) and <dataset_path> with the path to your dataset.

Evaluation:

To evaluate the model, run the following command:

python -m src.evaluate

Modify the target list path and weights path to match your setup

Pretrained model:

To save your time, we are current support 3 pretrained models:

• bert-base-uncased trained on preprocessded + undersampled dataset
• distilled-bert-base-uncased trained on preprocessded dataset
• bert-base-cased trained on raw + undersampled dataset

You can download them from the following links:

• bert-base-uncased
• distilled-bert-base-uncased
• bert-base-cased

After downloading, you can place them in the weights folder.

Prediction:

To make a single prediction using the trained model, run the following command:

python -m src.main

Deployment:

To deploy the model using streamlit, run the following command:

streamlit run src/app.py