Conference Article

·2022

A Business Workflow For Providing Open-Domain Question Answering Reader Systems on The Wikipedia Dataset

Dilan Bakır YTU , Mehmet S. Aktaş YTU

2022 IEEE International Conference on Big Data (Big Data)

Abstract

In a variety of sectors, we observe the emerging need for responding to user questions in a fast and efficient manner. We argue that addressing this need by developing question answering reader system applications will lead to several benefits: a) the density of call centers is reduced, and b) time is saved by getting answers through the application instead of going to the company itself. Examples of these applications, such as search engines help users find answers to their questions on documents containing important information, such as legal documents. These applications can be applied in digital banking, electronic commerce, and legal documents. In this study, we investigate the design of a business workflow that can provide answers to questions through documents containing important information, such as legal documents. In this study, we examine open-domain reader systems and propose a business workflow for open-domain reader systems. We are implementing a prototype application on the dataset to investigate the usability of the proposed business workflow. We discuss the prototype’s implementation details and share its evaluation results. The results show that the T5-based model provides better results in open-domain reader systems.

Keywords

Workflow Computer science Usability Domain (mathematical analysis) Variety (cybernetics) World Wide Web Question answering Open domain Workflow engine Workflow technology Data science Information retrieval Database Human–computer interaction Artificial intelligence

Subject Areas

Topic Modeling ·Artificial Intelligence ·Physical Sciences
Wikis in Education and Collaboration ·Communication ·Social Sciences
Multimodal Machine Learning Applications ·Computer Vision and Pattern Recognition ·Physical Sciences