Handwritten forms are the bane of a paperless project.
Sure you can scan them, but in most cases OCR does not work on handwritten text[1], and being able to extract and do something with the information on the form has been difficult if not impossible.
At least, this is what I thought
until the service Captricity was brought to my attention. They do something that I thought was unicorn and jetpack territory to this point: allow you to extract handwritten information from paper forms.
Here’s how it works: you upload a sample of your form to Captricity, then on their site you tell it where the fields that you want to capture are.
Once you’ve done this, you upload your filled out forms, and Captricity provides you the data in CSV format for you to take away and do something with.
I decided to give it a try while I sit here for a 2 hour delay at YVR airport using some vehicle repair forms from Mazda.
Captricity says their service works best for structured data like this, and not so much for big OCR jobs like letters etc.
Upload a blank form
When you create a new job, they ask you to upload a blank template. They say that if you don’t have a blank template, you can use the one with the least amount of writing.
I didn’t personally try that, as thanks to some Google-fu I was able to find a blank copy of my Mazda form.
Define fields
After you’ve uploaded your form, they show it on screen. You can then use your mouse to draw the fields over top of the document. It is super slick. If you want, you can then give the field a name and type.
Upload your filled out forms
Once you’ve defined your fields, you are then ready to upload your documents. You can either scan them with a scanner, take a picture with your camera, or import them from a number of services like Dropbox, Box, or Google Drive.
Your job will then start. They normally charge .20/page, but right now for new customers you get 25 pages free to try it out. You just need to answer a few questions about how you might use it.
According to Captricity, this is not just done by OCR, but the data is reviewed by humans three times.
View your data
After a while, you’ll get an e-mail saying that your job is ready. When you go in to view it, the data from the forms will be shown in table format. If you click on any of the cells, you will see an image of the underlying source data.
You can then edit any incorrect information.
Speaking of incorrect information, I was impressed with what Captricity was able to pick up, but it wasn’t 100% accurate.
For example, my wife’s last name is “Yeung”, but the data returned by Captricity was “Yeunc”. At first I thought this was a problem, but when I actually looked at what the jackals at Mazda wrote, I could see how it would look like “Yeunc”. Can’t blame Captricity for that one.
The good news is that the system learns. If you correct the first few, that correction will become the “Gold Standard” and then you can re-run it and it should be correct going forward.
Export Data
Once you have your data looking right, you can export to CSV, Google Docs, or Constant Contact.
They also have an API, so if you were into this sort of thing, you could build it right in to another application or workflow.
All in all, from my very limited test, I thought Captricity did a great job, and they seem to be gaining a lot of traction. I will be keeping my eye on them and if I ever have a job where I need a bunch of paper forms digitized, I would definitely give it a look.
-
At least the OCR that most of us normal humans have access to. ↩