Arrk Engineered a two-tier process to harvest bulk data extraction from PDFs - resulting in the reduction of costs incurred by the client while improving the efficiency of the process.

Customer

Our client is an award-winning eLearning company that offers corporate training solutions to help improve workforce performance while driving business on a global level. With over 25+ years of experience in building meaningful learning solutions, the brand has helped transform people into leaders of tomorrow with a 360-degree approach to training, leadership, procurement, and consulting.

Award Winning Company

Corporate Training Solutions

360 Degree Approach

Problem Statement

Our client was faced with multiple challenges in their quest to gather tabular data from PDF files that contained numerous financial details. One of the biggest problems was to extract information accurately for multiple channels within a single file. This meant that to avoid any discrepancies in the extraction of data, a sophisticated extraction tool needed to be used. The current in-house tool uses Python, Tesseract, Spacy, and Transformer to gather bulk data from PDFs.

Another issue that our client faced is that while the in-house PDF system is robust, it cannot extract tabular data from certain PDFs. This incomplete data extraction hinders the ability to not only gather insights but also impact the entire decision-making process. Our client wanted a solution to balance the cost-effectiveness while ensuring that the data extracted was accurate. However, a cost-intensive solution for every single PDF file did not seem practical. So, to strike the right balance between cost and efficiency was the need of the hour.

Information Extraction

Robust System

Data Efficiency

LLM and ChatGPT

Solution Development

To address the challenges, we at Arrk designed a two-tiered solution that combined the in-house tool with AI-driven technology.

For the in-house PDF crawling system, we implemented an algorithm that is capable of handling multiple financial instruments and tables within a single PDF file. The system was made to crawl the PDFs periodically to extract the data and prioritize accuracy above all else. By using the in-house solution, we allowed our client to achieve bulk extraction of data without any additional costs.

To act as a safe gate where the in-house solution did not work, the PDF was automatically routed to the AIMLbased Amazon Textract solution. The tool is driven by advanced machine learning to work as a fail-safe mechanism where all tabular data can be extracted accurately from PDFs. But, this platform would only be used when the in-house tool failed. Optimizing the use of Amazon Textract meant that the client’s expenses were minimized and were now cost-effective.

AI Driven Tool

Bulk Data Extraction

Cost Effective

Outcomes

Collaborative results with our client helped yield the following results:

  1. Combining the use of the in-house crawling system and Amazon Textract ensured that the most relevant data was extracted from the PDF.
  2. Prioritizing the in-house tool resulted in significant savings as the AI tool was only used for challenging situations.
  3. The two-tiered approach led to an overall efficient bulk data extraction process and maintained the balance between cost and performance.

Seamless Collaboration

Optimal Prioritization

Two Tiered Approach

More of Our Case Studies

See all case studies
See all case studies