Decipher Functionality Queries

[HowardBlake]

Hi - If anyone is able to help answer a few queries I have with Decipher it would be appreciated.

1. On the BP Product Roadmap a new vsn of Decipher is due for release in May - Decipher 1.2 I assume.  Is there an actual release date for this version and will it be available as a GA release?

2. Are there any plans to update the Decipher VBO Get Next Completed Batch action to include a parameter of Batch ID so that a specific completed batch can be retrieved from RabbitMQ queues?

3. If a DFD states a field is Mandatory and this field is NOT populated in a document then what does Decipher do with this document?  I am assuming that it would be flagged as requiring manual verification and the Decipher VBO action Has Batches To Verify would return True.  Are there any plans of having batch verification failures returned programmatically using a Decipher VBO action instead to avoid needing manual verification intervention?

4. I am building a demo process that uses the same document xx times with exactly the same data for each batch submission.  I have created a ML model and set it to enabled in the Document Type configuration page of Decipher UI.  I have also specified that the Model gets trained after every 50 documents.

I have run this document through Decipher approximately 100 times and each time I run my process the document is still flagged as requiring verification.  Why is this happening when I have the settings as mentioned above and the document has a green tick in the Verification window, all fields are captured correctly and I just press 'Submit'?

The confidence level for the documents processed also remains set to 1 in the Reports page - Batch History despite all the points above - any ideas why?

Many thanks

------------------------------
Howard Blake
RPA Development Lead
Sopra Steria
Europe/London
------------------------------

[Ben.Lyons1]

Hi Howard,

Thanks for getting in touch.

1. Version 1.2 is imminent, a communication will go out shortly to confirm when it's released. There's a presentation available at Blue Prism World of the new features.

2. I am guessing this is because you wish to segregate your batches/document types for different processes. If so you can set your RabbitMQ queue name depending on your process e.g. "UC001 - Process A". RabbitMQ will auto-create the queue if needed and Decipher will return it to the same queue on completion.

3. If a field has been flagged as required and the data cannot be found in the uploaded document, it will be held at the data verification stage. This will require manual processing, where the user can mark the correct region, manually write in the field or mark as an exception. That's a great idea and I don't know that it is currently planned for a release, please submit this idea by going to the Innovate button and Submit Idea.

4. A few things to address here. Much of the training is carried out using Decipher's native training mechanism and the additional ML model can be enabled once you're happy this is working as desired. The native training does use machine learning to update a model based on things such as keywords, data format, formulas and location.

I wouldn't recommend using exactly the same document multiple times, or at least not only 1 document, as this can mislead the training to believe the field values will always be the same.

To enable auto-verification, you go to Admin > Configuration > Automatically Skip User Modules > Data Verify > On. This will skip manual verification for all high confidence documents where all required data has been found. This is currently only available at a global level and not by Batch Type or other.

Let me know if you would like further clarification on any of those items.

Thanks

------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------

[HowardBlake]

Thanks again Ben for taking the time to respond.

I asked question 2 because I have a client use case that is rather niche and we intend to use Decipher in a less conventional way.  We would like to be able to retrieve a specific batch (& Doc Data) based on a Batch ID and not just rely on Decipher to return the Next Completed Batch from RabbitMQ as this may not relate to the Batch we require.  There are obviously ways I can work around this but I had hoped the Get Next Completed Batch may come with a Batch Id input param in the near future.  Based on your answer it sounds like this not currently planned so I will raise this as a potential new feature idea.

------------------------------
Howard Blake
RPA Development Lead
Sopra Steria
Europe/London
------------------------------

[Ben.Lyons1]

Hi Howard,

Thanks for the idea.

We'll be looking to update the VBO in the near future, so I'll see if we can get it included. This will not be linked to a major release, so we hope to get it out between releases.

Regards

------------------------------
Ben Lyons
Product Consultant
Blue Prism
UK
------------------------------

[Tejaskumar_Darji]

Point 4 - I've faced this issue. I too trained document like 100 times but still it's not getting trained at all. This did not happen with me in all documents but I've faced this while training couple of documents. It's like not getting trained at all. For rest of the times it picks up field automatically very next time once you map the field in data verification.

------------------------------
Thanks & Regards,
Tejaskumar Darji
Sr. RPA Consultant-Automation Developer
------------------------------