AI document parsing

Document parsing – is one of the stories we dealt with a lot in prev projects. But it usually had to do more with OCR rather than AI application directly. We tediously marked and labeled samples, then trained the model and had it work fine with a predefined set of documents. However, the outcome got still unstructured in most cases. Parse the document is one story, and having it correctly assigned with attributes and details – is another one. The second part – was usually a manual thing to do.

Now with recent AI NLP capabilities (ChatGPT, Alpaca), we can harness the full process – from incoming documents as input through its automatic recognition, classification, assigning the correct attributes, and uploading and linking with business entities in CRM, ERP, PIM systems.. wow!…

(though, yet not so easy in real life! 🙂 )

Together with our partner-client multichannel connector Holocene Services , we work on product handling automation, which lets users free up so much time eliminating tons of manual operations.

Imagine you have Shopify online-store or PIM system, and you want start selling on marketplaces like Amazon, Noon, Facebook, or any other. Other times, you would need to collect all the items in Excel and manually match tens of attributes of hundreds of products between systems.

Even worse, imagine you’ve got just PDF tech specifications of your products encountering hundreds of items in your catalog. Hours and hours of work to prepare it once and keep it updated all the way down 🙂

That’s indeed something that has prevented companies from entering other markets for years (real stories)..

Back to one of Holocene case scenarios – you are a seller and you’ve got just PDF product specifications. And you want start selling on Amazon.

What do we need to do? Yes, parse files, extract meaningful data, assign attributes based on data extracted, identify and pick the best fit Amazon category, identify and pick the right product listing template – and finally… create the listing!