Optical Character Recognition (OCR) Feasibility Challenge

Key Information

Register
Submit
The challenge is finished.

Challenge Overview

Welcome to the Optical Character Recognition (OCR) Feasibility Challenge. As part of this challenge, we are looking for ideas / suggestions to read handwritten forms and store the information in a digital format - This is an Idea Generation Challenge.

 

Project Overview

Our client is an American health insurance company that provides insurance products and financial services throughout the United States.

This project is to find a solution to address our client’s challenge of automating the handwriting and text recognition on scanned documents with 95%+ accuracy.  Current volume is approximately 5,000/month, however with an accurate solution may increase.

In our discovery, the OCR solution, Captricity, utilizes a combination of machine and human validation and emphasizes, ‘machine learning’; each time a form is reviewed, the solution’s recognition capability is higher.  

Your recommendations may include your own experience within this space, from evaluations you can perform with hands-on testing, and where hands-on testing cannot be performed, solution research and comparisons.

 

Challenge Details

Our client currently receives a significant number of their customer documents physically rather than online. They would like to automate reading the data that is handwritten or typewritten by their customers. In this context, you will have to propose ideas and suggestions on using Optical Character Recognition to read the physical copies and model them into digital ones.

Kindly note the following:

  1. We are not restricted to a tech stack / programming language. However, we would like to make use of web based services that are RESTful.

  2. All the documents that need to be scanned are handwritten or typewritten.  Note: Feel free to complete several claim forms for your testing.

    1. Blank Forms, Sample handwritten Claim Forms, and typewritten payroll deduction forms (PDR) are attached.

  3. You only need to focus on OCR for this challenge - the hardware behind it is not a matter of concern for this challenge.

  4. The documents will be in English.

  5. The documents may or may not have a specific template. All of it could be handwritten or some of it too.

  6. Only the text in the documents is of importance, their location / coordinates are not.

  7. You need to propose multiple solutions and rank them (and justify your rankings with your % accuracy findings).

  8. Captricity should be included in your review. You can download a trial from the respective websites. We have an internal resource that can answer any questions you may have. Please contact us via forum to get the resource names and contact information.

  9. Your solutions should be secure and HIPAA Compliant.

 

Deliverables

  1. Videos / Samples to demonstrate and validate your process. The videos need not have to be created by you and if they are, make sure they are unlisted and hosted on Youtube. Do not submit any video files. Host them on Youtube as an unlisted video and submit only the links to view them.

  2. A Pros and Cons Comparison chart with competency / % accuracy, and speed.

  3. Sample Outputs from any testing/review.

  4. A Word Document that proposes your findings, ideas, suggestions, and experience including technical overview (platform, speed, how solution is accessed) that meet the requirements of this challenge.

  5. For each solution proposed, please define if this knowledge is from previous experience, hands on testing or simply research/evaluations.

 



Final Submission Guidelines

No code submission required for this contest.  Kindly check out the deliverables section. Submit a single zip file containing your submission and upload it to the Online Review tool.

ELIGIBLE EVENTS:

2016 TopCoder(R) Open

REVIEW STYLE:

Final Review:

Community Review Board

Approval:

User Sign-Off

SHARE:

ID: 30052125