Document capture with advanced machine learning

Parascript has introduced a data location, extraction and verification software solution that deploys template-less, neural network-based document extraction.

Parascript says it has ‘productised’ it’s machine learning platform to support custom-developed recognition projects with much quicker turnaround than traditional rules-based approaches. The result is significantly faster production with more reliable and refined results.

“Machine learning offers a whole new set of opportunities for organisations across many industries to more precisely streamline their operations and deliver rapid, accurate data to their clients,” said Greg Council, Vice President of Marketing and Product Management.

Traditional recognition and capture solutions often successfully use business rules to process information. These rules place parameters around how information should be entered, increasing the accuracy of data recognised by software and reducing the amount of manual data entry that has been required. Unfortunately, rules are only valid when they are comprehensive, and these rules can only be comprehensive when the document types and their variability are well understood.

“Rules are brittle to change, that’s why implementing machine learning allows for so much more accurate results over time because it gracefully handles a dynamic environment without manually creating a whole new set of rules every time you have a new document type or image added to the system,” said Council.

“Even the best-run organisations struggle to maintain and improve data quality. Variations in documents and their data constantly occur. Using truth data or the set of data on sample forms that represents what the actual recognition answer should be provides the software with the knowledge of what the right results should look like so that it can automatically develop new rules.”