Many technology options are available today to automate mortgage document processing tasks. Some solutions are well marketed with great claims of mortgage ‘knowledge’ and an ability to provide tremendous results. In some cases, providers use offshore labor rather than technology, or some combination of the above. When conducting due diligence, understanding the differences in these approaches is key to supporting expansion and scalability requirements.
Three technology approaches help determine the right solution for automating mortgage processing documents:
1. Zonal OCR
The zonal Optical Character Recognition (OCR) approach looks in specific locations, or ‘zones’, on a page for relevant text. The benefits of this system include:
- Minimizing the OCR processing time since only configured zones are processed by the system.
- Works well on documents with a static layout such as tax forms or a 4506-T form.
However, this process is administratively heavy, as variations in document layout require distinct zonal templates. Many times the relevant text is in a highly volatile location, making it difficult to find.
2. Full Page OCR
This approach makes a full-page OCR ‘read’ of every page of every document, much the same as a human being. Ideally, each page is read in less than one second and the content is processed through a set of rules to determine the document type of each page. While this may seem to be the obvious way to approach the task of indexing the diverse documents found in the mortgage industry, most technology providers are unable to deliver the speed necessary to successfully scale with this approach. The benefits of this approach include:
- Works on all mortgage document variations, even pay stubs with millions of variations.
- Ability to index document versions which may have never been seen before by the system assuming they are lexically similar (same words and phrases found throughout similar learned examples).
- Ability to accurately distinguish between leading and following pages, eliminating the need for adding document separator sheets.
- Ability to “discover” data in a manner similar to a human being, using words and phrases across the entire document to find key data elements for extraction.
- High-speed OCR allows for almost infinite scalability with a relatively small hardware footprint.
3. Visual Classification (a.k.a ‘Fingerprinting’)
This legacy approach has recently been remarketed for use in the mortgage industry. While it does have the advantage of sub-second speed, it is not an OCR solution. Instead, an image analysis (non-text based) approach is used to identify documents and page types.
Visual classification attempts to differentiate between document types A and B largely by examining the distribution of ink on samples of each known document. Similar to thumbprint analysis, the graphical signature of each document type is learned and remembered. It has the following advantages:
- Processing speed.
- Works well on documents with a static layout such as tax forms or a 4506-T form.
- Training time, as it is relatively simple for the system to learn a small range of document variations from image signature analysis.
However, this system also has certain drawbacks, which include:
- The layout-specific configurations needed for each document variation can take a long time to set up if the number of document variations/types is high.
- These layout-specific configurations need to change if the layout of a document changes.
- The graphical signature approach tends to be less reliable with more than one hundred document variations/types to compare, affecting accuracy in some cases.
- Image processing time tends to be linear relative to the number of document variations/types.
- This approach fails to leverage the rich text present in a mortgage file to detect document boundaries for multiple page documents, while also lacking the ability to extract data from the documents once indexed.