Automatically classify files and route them to the proper folder or recipient using our classification software. With this technology, you can import your company’s taxonomy, train the system to accurately identify specific file types, and quickly organize and classify scanned and electronic files – both structured and unstructured.

Of course, understanding transactional documents is a lot more complex than just page-based classification. For example, the relevance of a document relates to its version. Is this tax document the most current version, or is there a newer one? Is this closing disclosure the final version or some intermediate version? Our advanced classification methods include:

  • Auto-versioning of documents – Determines the most recent version of each document class
  • Document-splitting – Decides where documents begin and end
  • Hierarchical taxonomy support – Supports classes, sub-classes, and super-classes of the same document, allowing the system to automatically align itself with a unique company taxonomy

Increase processing

To enable real-time loan decision-making, our software focuses on real-time performance. Real-time speeds vary by use case, but typically range from seconds to just a few minutes. Users can take control over processing speed using a few key levers. Our solution runs in the cloud and increases in speed as computing power increases, allowing end-users to control how fast loan decisions are made.

Increase processing

The Trapeze solution has the highest accuracy rates in the industry. Instead of relying on templates or solutions that can only understand structured forms and data, Trapeze understands documents and analyzes them using flexible document-parsing technology. If a document varies either over time or because it’s generated using a non-standard software package, our parser has no problem understanding and analyzing the document contents.

In addition, we believe in the power of redundancy. SoftWorks AI solutions typically include a set of different methods to solve the same problem. For example, for critical documents, we often employ multiple OCR streams coupled with machine learning and computer vision to deliver the highest accuracy rates. Should one document analysis method fail, our recognition solution recovers from it.

Enhance control with
exception validation

Our Trapeze software minimizes any need for human validation. The system can automatically calibrate passthrough and validation thresholds that require minimal human review in order to meet your SLA accuracy requirements. Files requiring manual review and validation appear in our CVIEW Validation interface, where users can approve or modify suggested classifications. Once the files are validated, the software automatically routes them to their appropriate locations.


The automated data classification software helps you:

  • Reduce labor costs – Classify documents automatically to minimize human intervention and time sorting files.
  • Classify various documents and complex files with ease – Process structured, semi-structured, and unstructured files and support scanned images, digital files, and complex hybrid documents.
  • Boost accuracy – Rely on our mortgage-built Trapeze solution that comes pre-configured with knowledge of more than 675 forms and 6,200 data fields. In addition, using our machine learning capabilities, the solution can train on an unlimited number of new forms and variations, and it can be configured to extract additional data fields.
  • Dynamic scalability – Handle shifts in loan volume dynamically and with ease. Your company cannot afford to lose market opportunity when there’s a strong increase in home lending. With Trapeze’s classification functionality, you can achieve accuracy rates of more than 99.5%, which is more accurate than human classification. You also can scale in real-time to meet market opportunity.
  • Real-time loan feedback – Deliver real-time understanding and status on a loan application. Trapeze supports real-time feedback, helping you answer such questions as:
    • Has a client uploaded all the documents required for the loan to go to underwriting?
    • Have all the documents been received for underwriting to perform a verification of income check on the borrowers?
    • Are the submitted documents, even those captured by mobile devices, of sufficient quality for the loan application or does the applicant need to resubmit?

Trapeze classification functionality can make these decisions in real-time and provide necessary feedback to borrowers, loan officers, and underwriters.

Our solution development is rapidly moving toward document classification requiring no humans in the loop (HITL). For most of our in-production systems, classification accuracy rates for passthrough documents are more than 99.5%. These rates are higher than single human key rates. Typically no human validation is required on these passthrough documents, which usually comprise 85% to 95% of the mortgage document workflow.

Our technology enables your mortgage workflow to move one step closer to touchless automation.