Data is now as valuable as land or labour, and a whole industry exists to type it into systems by hand. This research proposed an algorithm that automates much of that repetitive work.
The problem
The data-entry industry employs many people to look at sources, collect information and store it for business analysis. The work is tedious and repetitive. It stifles creativity and leaves room for mistakes.
What the research did
We developed an algorithm to automate data entry using the computing and communication infrastructure that data-entry companies already have. We tested it and compared it with practices at several data-entry focused companies.
What we found
The algorithm automated data entry with above 97% accuracy, more than 15 times faster than the existing speed, with no extra cost beyond existing infrastructure. It outperformed the manual approach in the companies we compared.
Who benefits
- Data-entry and outsourcing companies, including many in Bangladesh and South Asia
- Businesses that need clean data for analytics
- Workers, who can move from repetitive typing to higher-value tasks such as checking and analysis
Limits and what comes next
Automation changes jobs. Responsible adoption means retraining staff for quality control and analysis roles. Results also depend on the type of source documents.
The team
I led this research with Mohammad Abu Yousuf (Jahangirnagar University), Partha Chakraborty (Comilla University) and Md. Safaet Hossain (City University). It was presented at ICONCS 2020 and published by Springer.
Frequently asked questions
Can data entry be automated?
Much of it can. This algorithm reached over 97% accuracy at more than 15 times the manual speed.
Does it require new equipment?
No. It runs on infrastructure data-entry companies already have.
What happens to data-entry workers?
Ideally they move into checking, quality control and analysis, which automation still needs.