AI-based Optical Character Recognition

Business Challenge

Our client, a growing professional services organization, needed an OCR solution that provided a full alphanumeric recognition of printed or handwritten characters at electronic speed by simply scanning documents. Considering the volume of business content that the organization generated or received every day, the processing of complex or unknown formats and highly variable documents were difficult to capture with traditional systems and manual interfaces were time-consuming. Data from applications, forms, reports, office documents, images had to be clearly understood and the information had to be reliably extracted, even from poor-quality digital images and unstructured formats. If the documents were poorly managed, not digitized, or disconnected from critical business processes, it would impact the business’s ability to deliver exceptional customer service; it slows down important processes, increases security risk, and negatively impacts revenue. Conversely, controlling content precisely can significantly improve analytics strategy by gaining insight and business value from dark or unstructured data sources. With over hundreds of document and image files, manual indexing of huge amounts of data had its own challenges, including:

• A tedious and cumbersome process
• Expensive in terms of money and resources
• Involvement of the third party to manually index data
• The probability of errors increased with human intervention
• Customer claim request time increased due to manual work

The Client Requirements

Develop AI-based Optical Character Recognition (OCR) application to perform on PDF/Images for extracting the text (digits) from the defined region of interest(ROI).

Algorithms Used

• AI/ML model to detect the specified ROI
• ROI image enhancement
• Text cleaning
• Data dump in CSV/JSON

Business Transformation

Implementation of AI and ML

Understanding the requirements, ThinkPalm leveraged artificial intelligence and machine learning technology to develop APIs that helped recognize text, images, documents, and PDFs. API will be used by the end customer to integrate with their application, therefore transforming it into a digital counterpart. The OCR solution made use of innovative AI and ML tools to reduce the time, effort and cost of information extraction projects helping businesses achieve unprecedented efficiency. This helped the company staff to focus on strategic activities and not on monotonous, time-consuming and repetitive tasks.

Data dump in CSV/JSON

Thanks to the intelligent documentation process, the application offers automated accurate, consistent information access and retrieval. As per the client requirement, the meaningful data can be extracted into a structured format, such as CSV or JSON and could be further analyzed to derive actionable insight. The result can be shared across the organization, which enables efficient, consistent and accurate content collaboration and decision-making across every department. These results can be quickly downloaded into various formats as mentioned below, thus saving plenty of time and costs spent on manual document processing. Moreover, it helps mitigate the risk of human error. The result is improved compliance, reduced time employees spend on rework and prevention from related losses for you and your customers.

Text cleaning

ThinkPalm made sure that the client maintained an accurate encryption of data and leveraged a rich set of data quality, profiling, cleansing, and monitoring capabilities. Through text cleansing, ThinkPalm ensured that the information is of high quality and provided the ability to quickly understand the resultant document. The solution enabled the client to cleanse text quality continuously to reduce the proliferation of incorrect or inconsistent data. This automation empowered the team to manage data assets effectively and respond faster to business objectives with trusted data. The high-quality and easily understandable text ensured that data quality is preserved to meet all their business goals across a business-ready data pipeline. By implementing a text cleaning feature into the OCR solution, the client can enhance data integrity, cleanse data and monitor data quality on an ongoing basis, turning their data into trusted information.

Tools Used

Technologies

• Deep Learning
• Faster RCNN
• Tesseract

About Client:

MDIU Co, Ltd. is Japan’s leading technical service provider that mainly implements IT, RPA and AI technology across the board. They are one of the few companies that focus on promoting digitalization with ICT technology. Their mission is to enrich the lives of millions of people by utilizing technology both in Japan and overseas and to provide a place for a wide variety of business opportunities and encounters. MDIU creates business opportunities with next-generation business platforms specializing in BtoB through the utilization of linked data of people and companies. Other than offering services MDIU also addresses social issues by creating new opportunities through education on advanced technologies such as AI in emerging countries such as Africa and Asia.