DistilBART-QGen-3-3 Open Source Model - Freely Generate Corresponding Questions Based on Text Paragraphs and Answers

Distilbart Qgen 3 3

Developed by gpssohi

This model is a BART variant fine-tuned on the SQuAD dataset, specifically designed to generate corresponding questions based on text passages and answers.

Question Answering System

Transformers

EnglishOpen Source License:Apache-2.0 #Question Generation #Text Summarization Distillation #Educational Assistance

Downloads 21

Release Time : 3/2/2022

Model Overview

This is a text-to-text generation model capable of generating natural language questions based on given text passages and answers. The model is compressed into a smaller scale through distillation techniques, making it suitable for practical deployment.

Model Features

Distillation Compression

Distilled from a 6-6 layer BART model to a 3-3 layer structure, reducing model size while maintaining performance.

End-to-End Question Generation

Directly generates fluent and natural questions based on text passages and answers.

Fine-tuned on SQuAD

Fine-tuned on a high-quality QA dataset to ensure the relevance of generated questions.

Model Capabilities

Text Understanding

Question Generation

Text Summarization (basic capability)

Use Cases

Educational Technology

Automatic Question Bank Generation

Automatically generates exercises and exam questions based on textbook content.

Significantly reduces teachers' workload in question preparation.

Dialogue Systems

Dialogue Guidance

Helps chatbots proactively ask relevant questions to guide conversations.

Improves dialogue fluency and user engagement.

🚀 Question Generation Model

This model is designed for question generation and summarization tasks, offering high - performance solutions for various NLP scenarios.

🚀 Quick Start

This model checkpoint is obtained by first fine - tuning the sshleifer/distilbart - cnn - 6 - 6 summarization checkpoint on the SQuAD dataset. After this, the 6 - 6 fine - tuned model is distilled down to a 3 - 3 model which gives us the final checkpoint. GitHub Link for training scripts.

💻 Usage Examples

Basic Usage

The input format is as follows: [answer] <s> [passage]. The model will predict the question that corresponds to the answer from the passage.

📚 Documentation

Plot

Distillation Run

Dataset

The goal of Question Generation is to generate a valid and fluent question according to a given passage and the target answer. Hence, the input to the model will be a passage context and an answer, and the output / target will be the question for the given answer. Question Generation can be used in many scenarios, such as automatic tutoring systems, improving the performance of Question Answering models and enabling chat - bots to lead a conversation. The final dataset is created by taking the union of the following Question Answering Datasets. The dataset must have the following three columns: context, question, answer.

SQuAD

Stanford Question Answering Dataset (SQuAD) is a reading comprehension dataset, consisting of questions posed by crowd - workers on a set of Wikipedia articles, where the answer to every question is a segment of text, or span, from the corresponding reading passage, or the question might be unanswerable. We use the SQuAD 1.1 variant which does not have unanswerable questions. So, every question will have a corresponding answer and vice - versa.

Preprocessing

The first step is to remove questions which don't have answers. After that, we split the train set into Train and Eval sets and treat the dev set as the test set.

Stats

Original Dataset

Split	Num Docs	Num Contexts	Ques w/ Ans	Ques w/o Ans	Num Unique Ans
Train	442	19035	86821	43498	86821
Dev	35	1204	5928	5945	10279

After Preprocessing

Split	Num Rows	Context	Answer	Question
Train	80995	653,120,20	43,3,1	40,10,1
Eval	5826	445,123,67	28,3,1	29,10,3
Test	10297	629,129,25	29,4,1	31,10,3

The numbers in the columns indicate max, avg, min number of words.

📄 License

This project is licensed under the Apache 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご