Downstream Transformer Generation of Question-Answer Pairs with Preprocessing and Postprocessing Pipelines

TL;DR

This paper introduces TP3, a pipeline that fine-tunes pretrained transformers to generate high-quality question-answer pairs from articles, utilizing preprocessing and postprocessing steps to improve output relevance and quality.

Contribution

The paper presents a novel end-to-end system combining fine-tuning, preprocessing, and postprocessing for question-answer pair generation from text.

Findings

High-quality QAPs generated on Gaokao-EN dataset

Effective use of T5 models with SQuAD for fine-tuning

Pipeline improves relevance and filtering of QAPs

Abstract

We present a system called TP3 to perform a downstream task of transformers on generating question-answer pairs (QAPs) from a given article. TP3 first finetunes pretrained transformers on QAP datasets, then uses a preprocessing pipeline to select appropriate answers, feeds the relevant sentences and the answer to the finetuned transformer to generate candidate QAPs, and finally uses a postprocessing pipeline to filter inadequate QAPs. In particular, using pretrained T5 models as transformers and the SQuAD dataset as the finetruning dataset, we show that TP3 generates satisfactory number of QAPs with high qualities on the Gaokao-EN dataset.