Document forms fuel almost every process in every company. AI promises to solve a lot of problems and the overhead the paper format creates, but it first needs massive volumes of data and that’s once of the most common obstacle for any AI initiative.
Document form problems
Those who seek to undertake document AI projects would often face either or all of these common problems:
- Lack of data - AI needs data and lot’s of it for training and testing AI models. However, having the right variety and amount of real world documents to do training and testing could be a challenge.
- Lack of structure - even if one has enough data, odds are these documents would lack proper labeling, needed for creating AIs.
- Privacy - real world documents often hold personal information which could make it difficult/impossible to use them beyond their original purpose, considering GDPR and the other data privacy laws.
We created Fabricator AI to solve the document data problem and help you launch your document AI projects fast. Fabricator AI allows you to create an infinite volume of documents, using 100% artificially created personal data and providing all sort of configuration options to help you mimic your real world document forms.
How it works
Creating a set of fake documents consists of two steps and depending how many template you have, it could take you form a few minutes to a few hours.
1. Configure your document templates
First you need to setup the template forms you want to use. To do that, you'll need blank form templates as PDF that Fabricator AI will use as base for your fake documents.
Uploading your templates, you can identify them with different document types that later will be listed in the labeling manifest CSV.
The Fabricator will allow you to define in detail the type of data that needs to go into every field. You can choose between different options that will allow you to fine tune everything in order to match a real document.
You can also choose to make certain fields and portions of a document optional and ergo certain items in your document set will have these fields blank.
Once you have your templates set up, you can use reuse them and create as many set variations as you like, without having to reconfigure the templates again (given that there are no changes to the templates needed).
2. Create a set
Once you have your form templates set up, you can use any or all of them to create a set with populated versions of them. Configuring your document set options, you need to select which templates you want and how many fake versions of them you want. Along with that, you can choose from a few more options like font style, noise, colour. These will allow you to mimic some of the problems real documents have and make your set more authentic.
Once you launch the new set creation, the Fabricator will create your fake documents and give you an S3 bucket where you can download your fake document, along with a CSV label file. Depending on your set volume, it could take from a few minutes to an hour for all items to be created.