Skip to content

This is our capstone project for Human Language Technologies course.

Notifications You must be signed in to change notification settings

dhakalmahima188/Question_Generation

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Question and MCQ Generation

Project Overview:

Our project automate the process of generating questions from text passages.Here, we have finetuned T5-small model for question generation. and used the same finetuned model for MCQ generation.

Acces our training notebooks and models

  • Access our Colab notebook here.
  • Acces our Question Generation Model on Hugging Face here
  • Training notebook on Kaggle [End to End Generation] here

View our Demo:

Demo Video

Run

  • Clone the repository : git clone git@github.com:Nadika18/Question_Generation.git
  • Install the requirements : pip install -r requirements.txt
  • Run python3 manage.py runserver in the terminal
  • Open the browser and go to http://localhost:8000/
  • Submit the question at /
  • Get the generated question at /submit-question

System diagram for finetuning T5-small model for Question Generation

Question Generation

System diagram for MCQ generation workflow

MCQ Generation

Outputs:

1. Question Generation:

Question Generation

Question Generation

2. MCQ Generation:

Question Generation

Question Generation

Question Generation without Answers : End to End Generation

TrainingArguments

Parameter Value
evaluation_strategy epoch
learning_rate 2e-5
per_device_train_batch_size 8
per_device_eval_batch_size 8
num_train_epochs 3
weight_decay 0.01
save_strategy epoch
disable_tqdm False
gradient_accumulation_steps 2

Note : The batch size and acclumulation steps were decreased during training due to memory constraints.

alt text alt text

About

This is our capstone project for Human Language Technologies course.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 98.7%
  • Other 1.3%