Do you know what your words mean to anybody?
Fuzzy Octo Waffle is a simple machine learning system that utilizes pre-trained BiLSTM neural network models, trained on both English and self-created Ukrainian datasets, to perform emotion analysis.
This repository, created for a bachelor's diploma project, provides a straightforward ML application in two main versions:
- Google Colaboratory Version: Leverages Google Colab for training, validating, and using emotion analysis models, utilizing Google’s resources.
- Local Version: Utilizes Gradio for the interface. This version allows the usage of pre-trained models but does not support training new models.
- Pre-trained BiLSTM Models: Access pre-trained models for emotion analysis in both English and Ukrainian.
- Dual Deployment: Choose between Google Colab for full model training and local deployment for quick model inference.
- User-friendly Interface: Easy-to-use interface with Gradio for local deployment.
- Download and Extract: Download this repository as an archive and extract it to your desired location on your PC.
- Install Dependencies: Navigate to the extracted folder and run the
Install.batfile to install all required libraries. - Launch Application: After successful installation, run
launch_main.batand wait for the Gradio interface to launch athttp://127.0.0.1:7860/.
- Download Repository: Download this repository as an archive and extract it to your desired location on your PC.
- Upload Files to Google Drive: Upload the three files from the
colabfolder to your Google Drive. - Follow Instructions: Open the uploaded files and follow the instructions provided for usage.
- English Model: Trained on the Emotions Dataset for NLP from Kaggle.
- Ukrainian Model: Self-created dataset consisting of 200k+ labeled words categorized as "negative" or "positive". Feel free to use this dataset and credit me in your project.
I'm open to questions, requests, propositions, and contributions. If you have any suggestions or improvements, feel free to create a pull request or open an issue.
This project is licensed under the MIT License. See the LICENSE file for more details.