Techniques to Improve Q&A Accuracy with Transformer-based models on Large Complex Documents

TL;DR

This paper evaluates different text processing techniques and their combinations to enhance the accuracy of transformer-based Q&A systems on large, complex documents, identifying the most effective methods for improved performance.

Contribution

It systematically analyzes and identifies the best combination of text simplification and encoding techniques that significantly improve transformer-based Q&A accuracy.

Findings

Optimal technique combination improves accuracy statistically

Simplified text leads to more relevant responses

Certain encodings enhance transformer performance

Abstract

This paper discusses the effectiveness of various text processing techniques, their combinations, and encodings to achieve a reduction of complexity and size in a given text corpus. The simplified text corpus is sent to BERT (or similar transformer based models) for question and answering and can produce more relevant responses to user queries. This paper takes a scientific approach to determine the benefits and effectiveness of various techniques and concludes a best-fit combination that produces a statistically significant improvement in accuracy.