 | Document Recognition |
Generation II of the CursiVision system brings an entirely new dimension to document recognition and the power and flexibility in how it works. To illustrate the idea, we'll use the particular type of enterprise for which we developed this new capability.
One of the key markets for CursiVision is the income tax preparation industry. From smaller stand alone tax preparers all the way up to large enterprise companies that might have thousands of franchisees
Indeed, we used actual commercial tax preparation software as our development environment.
A key challenge for these companies is the shear variety of documents that differ in so many ways. In addition to this variety, there are also two other significant issues we knew we had to solve:
- The sequencing of different documents, or forms, combined together in arbitrary ways in any given instance for processing.
- Different instances of any particular document - even a single page from the document - could have different requirements for signature capture.
Our committed goal was to satisfy these two issues in a powerful and flexible way, and, did we ever!
At the end of our development, we had setup document recognition for nearly 900 different documents, which included personal and business IRS tax forms for all 50 US states, and also for variations of all of these forms, for example, the same personal tax form, but with different filing status selections. Married filing jointly gathers two signatures, where filing single only gathers one. We developed the powerful technology to handle such variations and it is working very well.
Different forms in any order
First, some background...
Consider the prior generation of document recognition technology in CursiVision. Though powerful, it was limited in that it only looked for one particular "profile" when it recieved a document from the virtual printer.
To clarify "recognition", note that the user, when first printing a document to the virtual printer, would be presented with a user interface they would use to select text on some page of the document that uniquely identifies it. For example, the word "Invoice" at some exact location on some exact page.
In subsequent printing of that document, when that text is found on that page - the "profile" is selected and automated processing begins.
To clarify "profile", this is the definition of signing areas and processing steps exercised on a document by the system once it is recognized. Since the system picked the profile from those available when the software recognized the document, essentially only one profile would be found and used.
The new recognition strategy...
Now, when a document is printed to the virtual printer, the system inspects the document for every page !
A particular document can be setup to be recognized by content on multiple pages in which case those pages must have the specified content to be recognized (though the order of the pages does not matter). However, even after a document is recognized and the automated process begins, the system continues to inspect all pages for additional recognition opportunities, and thus, any additional "profiles" may be selected (which will also be excercised in series after the first one is done).
This is how you can now print any series of documents, indeed some or all pages of a document, to the CursiVision virual printer, and all of the different sets of signature capture and processing (profiles) will be excercised on that entire document in a totally automated way.
Another capability of the new recognition strategy: the system allows you to define "optional" signature capture areas when it finds specific text in a specific location. Consider, for example, every type of document you might print in your organization may or may not need to gather a particular signature, say from the Big Cheese. Now, you can gather this signature only when that particular location is found on any document irrespective of whether that document is tied to some other document type.
A final detail on recognition strategy: CursiVision uses both text recognition and image recognition while looking for candidate profiles for a given document. This helps assure accuraccy in selecting the appropriate processing steps to use for a document.
BTW: Did you know that the signer can be anywhere in the building with the EnVisioNate Phablet Signature pad, and that they will see the content surrounding the requested signature area on the pad (i.e. Android phone or tablet) while signing?
The same form or page with differing signature gathering requirements
First, some background...
In the prior generation of CursiVision, document recognition sometimes required multiple "profile" definitions for the same document which might differ only slightly in variations of it.
A specific example is the IRS form 1040. It is easy to setup recognition for a profile to handle this form for all instances, but that doesn't handle the situation where different steps might be required for variations of the form.
A clear example is when the filing is for Married Filing Jointly, or Filing singly. In the former case, you'd want to gather two signatures, primary filer and spouse, in the latter, only one. This usually meant having two signature areas "just in case", and having the single signer skip one. This also prevented configuring different steps for the form. For example, maybe you want to store the primary and spouse's SSN in a spreadsheet, did you know you can do that ?
The new strategy...
CursiVision now has the concept of "variables" that it can use when inspecting documents to find the correct processing profiles. Or, as in the previous section, the correct "set" of profiles!
Essentially, you can now define different profiles, or if you want to think of them as "sub-profiles", from the same document with different data!
To use the IRS 1040 example, there's a checkmark in one of the "Filing status" selections on the first page. You can actually define a profile now that can inspect text in the document and do different things based on whether or not if finds that text in that location.
In particular, if there's a checkmark in the MFJ status field, two signatures are captured, otherwise 1. The power of this feature cannot be overstated. It is, for example, what enabled us to develop nearly 900 different profiles from a set of 4 or so documents from 50 states, which would be about 200 profiles without this feature, but is actually more than 4 times that with this capability!
But the old strategy still works
Even after all that - you still have the very simple recognition strategy that may be quite sufficient for the vast majority, if not all, of the documents you are likely to need processing for