This is the code for the paper:
Felipe Viegas, Washington Cunha, Christian Gomes, Antônio Pereira, Leonardo Rocha and Marcos Goncalves. CluHTM - Semantic Hierarchical Topic Modeling based on CluWords (ACL2020)
Python3 requirements:
- numpy
- scipy
- gensim
- pandas
- scikit-learn
- joblib
- matplotlib
- prettytable
- nimfa
- pillow
- seaborn
Word Embedding exploited:
- FastText (Pre-trained Word Embedding)
Once installed, setup the paths in the startup file :
main.py
Build docker container:
docker build -t cluhtm <project_path>
Run docker container:
docker run --rm --name cluthm -v <project_path>:/cluhtm -i -t cluhtm /bin/bash
To run the code:
- Store the dataset in
./datasets/textual_folds/example.txt. Where example.txt contains the data samples; one sample per line - Store the fasttext embeddings in
./datasets/wiki-news-300d-1M.vec python3 main.py -d <dataset>
For more information about building and running a docker container, see: https://docs.docker.com/
Conference paper:
title={CluHTM - Semantic Hierarchical Topic Modeling based on CluWords},
author={Viegas, Felipe and Cunha, Washington and Gomes, Christian and Pereira Antonio and Rocha, Leonardo and Gonçalves, Marcos André},
booktitle={The 58th Annual Meeting of the Association for Computational Linguistics (ACL ’20)},
year={2020},
organization={ACL}
}```