Decipher-unable to read the data from similar invoice
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
12-07-21 01:13 PM
Hi,
I am working on Decipher. I have trained the decipher for 3-4 invoices. For those invoices it is able to extract the data. whereas no data is extracted for invoices with similar templet apart from trained invoices. What could be the possible issue. Please Let me Know.
Thank you in advance.
------------------------------
Nistha Agarwal
------------------------------
I am working on Decipher. I have trained the decipher for 3-4 invoices. For those invoices it is able to extract the data. whereas no data is extracted for invoices with similar templet apart from trained invoices. What could be the possible issue. Please Let me Know.
Thank you in advance.
------------------------------
Nistha Agarwal
------------------------------
6 REPLIES 6
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
14-07-21 03:59 AM
Hi Nishtha,
A good starting point would be to first understand all the different formats you'll be working with, and then to collect at least 10 documents for each type.
All this makes it easier for Decipher to learn from a wider set of data and adjust it's model accordingly.
Another thing that helps is if you set formulas or regex patterns for your fields in the DFD to help Decipher understand the nature of the fields on the forms.
Here a step by step guide for you to go through - http://bpdocs.blueprism.com/decipher/user-guide/getting-started.htm?tocpath=User%20guides%7CGetting%20started%7C_____0
Also, a couple points to note:
1. Any changes to the structure of the DFD will reset the ML model and hence it may be the reason you're not receiving desired results.
2. Training Decipher repeatedly on the same set of documents is not considered best practice. The more variations you can add to this batch the better.
Regards
Harpreet
------------------------------
Harpreet Kaur Product Consultant
------------------------------
A good starting point would be to first understand all the different formats you'll be working with, and then to collect at least 10 documents for each type.
All this makes it easier for Decipher to learn from a wider set of data and adjust it's model accordingly.
Another thing that helps is if you set formulas or regex patterns for your fields in the DFD to help Decipher understand the nature of the fields on the forms.
Here a step by step guide for you to go through - http://bpdocs.blueprism.com/decipher/user-guide/getting-started.htm?tocpath=User%20guides%7CGetting%20started%7C_____0
Also, a couple points to note:
1. Any changes to the structure of the DFD will reset the ML model and hence it may be the reason you're not receiving desired results.
2. Training Decipher repeatedly on the same set of documents is not considered best practice. The more variations you can add to this batch the better.
Regards
Harpreet
------------------------------
Harpreet Kaur Product Consultant
------------------------------
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
19-07-21 09:10 AM
Hi @NisthaAgarwal,
Have you manage to resolve this issue?
Thanks
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Have you manage to resolve this issue?
Thanks
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Ben Lyons
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
19-07-21 10:26 AM
Hi @Ben.Lyons1
I am trying but still not able to resolve this issue.can you please suggest me how i can resolve it?
Thank you in advance
------------------------------
Nistha Agarwal
------------------------------
I am trying but still not able to resolve this issue.can you please suggest me how i can resolve it?
Thank you in advance
------------------------------
Nistha Agarwal
------------------------------
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
19-07-21 10:28 AM
Happy to help, I'll just need a little more info.
Please can you share screenshots of your DFD and Document Type?
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Please can you share screenshots of your DFD and Document Type?
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Ben Lyons
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
19-07-21 10:52 AM
Hi @Ben.Lyons1
PFA for DFD and document type.
------------------------------
Nistha Agarwal
------------------------------
PFA for DFD and document type.
------------------------------
Nistha Agarwal
------------------------------
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
19-07-21 01:04 PM
Hi Nistha,
Thanks for uploading those.
Your DFD looks good and nothing obvious to change.
However, I would recommend disabling the ML model until you have successfully trained the "Training Data". You should be able to get very good results without the additional ML model being created, this might only be wanted once you have progressed into production.
You have also set the primary and secondary languages the same. Firstly, these only need to be specified if they are different from the Batch Type language. The language in document type will override the language in batch type, but can be left blank if the batch type language is correct.
Secondly I would only specify a secondary language if it's different from the primary.
Once you've made these changes you will need to upload new documents, or restart your existing batches at the capture stage.
Let me know how you get on.
Thanks
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Thanks for uploading those.
Your DFD looks good and nothing obvious to change.
However, I would recommend disabling the ML model until you have successfully trained the "Training Data". You should be able to get very good results without the additional ML model being created, this might only be wanted once you have progressed into production.
You have also set the primary and secondary languages the same. Firstly, these only need to be specified if they are different from the Batch Type language. The language in document type will override the language in batch type, but can be left blank if the batch type language is correct.
Secondly I would only specify a secondary language if it's different from the primary.
Once you've made these changes you will need to upload new documents, or restart your existing batches at the capture stage.
Let me know how you get on.
Thanks
------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------
Ben Lyons
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
Principal Product Specialist - Decipher
SS&C Blue Prism
UK based
