You can mold Impira’s flexible platform to automate and accelerate your workflow. Here are a few handy tips and tricks that you can use to get the most out of Impira’s capabilities.
1. Use collections to group files together with common data
Collections are the core functionality that you can use to organize your files and extract the data you want. The main idea of the collection is that each file within that collection should have common data that you wish to extract. For example, if you need to extract data from application forms as well as from paystubs, you should add them to separate collections because the sets of information that you wish to extract will be different across those two types of forms.
2. Creating multiple collections
Impira’s system is built to handle scanned and imaged documents. However, if you have a variety of documents that are laid out differently on paper, you’ll get the best results if you put those different documents into separate collections, even if the information that you want to extract is the same.
3. Add more than five files to a collection
Impira’s AutoML learns from the very first manual extraction that you do. For you to see predicted extractions, you need at least two files: one from which you’ve manually extracted data and one on which AutoML can make a predicted extraction. In general, we recommend that you add at least five files to a collection and manually extract all the fields from at least a few of the files to see the best initial results.
4. Turn all confidence scores to green
While you can start seeing AutoML results from the very first extraction, the more manual extractions and verifications that you do, the more the AutoML learns and the more accurate it gets. Impira visually displays how confident the AutoML prediction is for each extraction. That allows you to quickly see the predictions for which the AutoML is confident and those which you should review first.
We recommend “getting to green” and verifying or correcting all of the predictions with the orange indicator to confirm they are correct. In addition to making sure that those values are correct, you’ll be teaching the system with each interaction. You might even see some of the orange indicators switch to green as you’re verifying other predictions.
5. Add your own notes with a manual field
In addition to fields powered by AutoML, you can also create manual fields to track more metadata for your records. Some examples include: using a manual checkbox field to verify whether a document has been reviewed by a QA worker, a date field to track by when something needs to be reviewed, or a text field for additional notes.