This is a final-year project done by a group of Bishal Gaire, Bishal Rijal, Dilip Gautam and Saurav Sharma. This project classifies whether a question is sincere or insincere using deep learning. It uses dataset provided by Kaggle.
Project Info:
| License | LoC |
|---|---|
 |
Install
To install a project Poetry should be installed for package management. You can learn about poetry by visiting their Github or Website.
Running Locally
This project is built using Google Colab free service so to run locally initially kaggle.json file is required. The steps for installing and authenticating a kaggle CLI can be found here.
All Jupyter Notebook files have two initial cells which are used for uploading a kaggle.json file on Google Colab so those initial two cells should be ignored while running locally.
Also, you are not required to run cells that installs a package using the pip command over Google Collab where packages are not preinstalled. All packages can be preinstalled using a poetry locally.
Similarly, a Python file is also present which is built from a jupyter notebook by removing unnecessary code cells from it but kaggle dataset needs to be downloaded early
You need trained model file and tokenizer file to run a Flask app or which can be generated running a Python file named train_and_save_model.py for running and training ANN and saving the best model. Initially models folder needs to be created otherwise it may fail to save a model altogether
Running Over Google Colab
To run Jupyter Notebook over Google Colab you need to have a kaggle.json file locally. While using Google Collab all cells are required.
To open the notebook in Google Colab you can replace out https://github.com address with https://colab.research.google.com/github or simply install chrome extension or firefox extension.
You can find out more information about opening a GitHub link in Google Collab and other functionality here
Other Models
Other models which are tested are also present in different-models branch if you need to see models then you can see them in that branch and run as required. Only the Jupyter Notebook is available for those models.
Images
Images such as word cloud, bar and other diagrams are present in the images directory to visualize data length & and distribution.
Research Paper
We have also published our research paper on the project if you need to read our research paper and learn about the project then you can read out research paper present in this repo.