Preferred data extraction using advanced field prompting
Dear Customers,
We hope you’re well
A quick background
Intelligent Character Recognition (ICR) comes as part of Zia’s optical recognition capability called Zia Vision. When we introduced it last April, the data extraction was training-based and was applicable for standard templates.
Later that year, we added one other extraction technique called field prompting where you are only required to choose the fields you would like to extract in the input image. And Zia, using Vision Language Model (VLM), was able to extract data from images in any kind of orientation, on-the-go.
Today, we are expanding the scope of the field prompting technique.
Advanced field prompting
About the enhancement:
As mentioned above, the field prompting extraction involves choosing of desired fields to extract in the ICR rules (where you configure preferences for ICR extraction for a module). To this, we add another layer called advanced field prompting.
If you have similar fields in your CRM layout or if you foresee similar kinds of values that can be used in the input image, then ICR might extract one for the other resulting in inaccurate data association.
To address this is why we built this advance field prompting technique. Here, you select the field that might presumably carry one or more similar values in the input image, an provide more character context as prompts. And, the best part—it can be presented in natural language.
One classic example we can use is the address fields.
In CRM, you have various fields to store billing address, company HQ address, and shipping address. Likewise, a customer can present several input images like invoices, business cards, handwritten bills, receipts, brochures, ID cards—each with one or more addresses printed on them. In such cases, specific prompts on the values’ characteristics will help Zia identify, localize extract, and associate a value that is more accurate on your standards.
While advanced field prompting is built to extract accurate values, the scope extends to all those use cases that have to do with data transformation. Here’s an example!
Data format transformation: The below image shows how Zia understands the specifics on the chosen fields and rendered the desired format in the extraction page.
There—by instructing what you want to extract from a source image, ICR can deliver accurate extraction, leaving more time for processing instead of spending time in recording and refining input data.
Release plan: This is released for the applicable DCs—US, EU, and IN
Resource: Help document
That’s all for today. We will see you soon with more interesting updates to ICR in Zoho CRM.
Kind regards
Saranya Balasubramanian