Topic Modeling Tool for Persian Short Texts

The tool for topic modeling provided by the Data Science Innovation Center extracts topics from digitized Persian texts and compares their performance in short texts using a variety of topic modeling techniques.

Visit the website to view the description in Persian.

Installation

We recommend Python 3.6 or higher, gensim 4.2 or higher.

Install from sources

You can also clone the latest version from the repository and install it directly from the source code:

git clone https://github.com/DSInCenter/topicmodel.git
cd topicmodel
pip install -r requirements.txt

Getting Started

To get started, you can see the demo of GSDMM's algorithm in this link:

• GSDMM demo

These examples demonstrate how to clone and execute a model on Google Colab:

• Run NMF model on Google colab
• Run LDA model on Google colab

LDA demonstration:

First, import Dataset Class from Dataset.py and import LDA model from LDA.py:

from tools.Dataset import Dataset
from LDA import LDA

Create Objects from Dataset and LDA Classes and Traing The Model:

lda = LDA(num_topics=11, iterations=5)
dataset = Dataset('Dataset', 'utf-8')
lda_result = lda.train_model(dataset, hyperparams=None, top_words=10)
print(lda_result)

Citing & Authors

If you find this repository helpful, feel free to cite this work :

@article{karimi2023comparative,
  title={Comparative Analysis of Topic Modeling Algorithms for Short Texts in Persian Tweets},
  author={Karimi, Amir Hossein and Akbari, Masoud and Akbari, Mohammad},
  year={2023}
}

Don't hesitate to send us an e-mail or report an issue, if something is broken (and it shouldn't be) or if you have further questions.