Hi @veerish uppin​ ,

I'd recommend to keep the training set < 150 documents, post training upload the Testing Set via Task Bot for checking the LI outcome and there will be no restriction for this. Make sure each page you upload once the LI is set to Prod will consume License count.

As of now we have trained the LI model with 80 files. The ​accuracy is not great. So thought of adding more file for the learning instance to increase the efficiency.

​

Any suggestions on this ? ​

Hi @veerish uppin​ ,

Categorize the Training set as below,

1. Have 10 - 15 sample invoices for Each format
2. Test the LI with different set of invoices which is not used in training.
3. Hope you would have done Pre Processing before uploading the invoices to LI
4. If you are not getting results with a particular OCR Ex: ABBY try to do POC with other OCR's available in IQBOT.
5. If above things are not in favor, connect with your PEM or CSM for getting additional help from AA.

@ChanduMohammad S​ Yes I am pre processing the documents with python before giving it to IQ Bot.

​

Also, wanted to check if the Learning Instance learn from execution, say like ​there are 20 invoices out of which 5 went to validation center for human validation.

​

Will the learning instance learn when we do the manual validation ?