Decipher - Cons and Limitations

[NarendiranThan1]

Can anyone tell what are all the disadvantages in decipher? And What are all the constraints in uploading (document types, pages per document , number of documents per batch...etc).

------------------------------
Narendiran Thangadurai Blue Prism Developer
------------------------------

[Neel1]

Hello Narendiran,

Same questions of your has been answered already 4 days back. 

https://community.blueprism.com/communities/community-home/digestviewer/viewthread?GroupId=145&MessageKey=2040de41-11f8-4985-955f-588a3f851f16&CommunityKey=3743dbaa-6766-4a4d-b7ed-9a98b6b1dd01&tab=dig...




------------------------------
Neeraj Kumar
Technical Architect
------------------------------

[BenLyons · Accepted solution]

Hi Narendiran,

Decipher can work with PDFs, BMPs, JPGs and PNGs, so any other image type is not currently compatible. There's no limit on how many pages or documents per batch, however the performance may be affected if the infrastructure isn't designed to meet the respective needs.

If your documents have a lot of pages, you may want to test batches of different volumes to optimise performance for your local environment.

Decipher is optimised to work with images of 300dpi or greater, however it can work well with lower quality images in many circumstances.

If you have any specific requirements you would like to check, please let us know. Alternatively you can reach out to your Blue Prism account manager for more information.

Thanks

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

[NarendiranThan1]

Hi Neeraj,

 I was initially posted in Blue Prism Product community. That answer is suggesting me to post this question in this(Decipher) community to get more answers.

------------------------------
Narendiran Thangadurai Blue Prism Developer
------------------------------

[stepher]

Narendiran,

My background is in Master Data Management, and Audit and Control, with a Supply Chain focus.  So, to my mind, one absolute of any process, manual or automated, is to be able to trace and audit it from start to finish.  With that in mind...

It is still early days for me, but the single largest challenge is the difficulty of maintaining the connection between the input of the document being scanned and the output of the scanned data.  That is to say, if I submit Documents 1, 2 and 3, I feel it is important to be able to directly tie back Results A, B and C.  With Decipher, if all three Documents are submitted in a single Batch, it is impossible to maintain the connection with the Blue Prism tools currently offered.  I have raised this concern and the recommendation that I received from Blue Prism was to limit my Batch size to a single Document.  To their credit, this does work.  (I am calling this method Batch-1-Document-1, or B1D1, and looking to trademark it.)  But it seems like it adds a lot of unnecessary overhead.

Why does this happen?
Decipher assigns the document(s) within a Batch a GUID formatted value.  I do not have a problem with that.  The issue is that your Blue Prism processes will not see the Document GUID assignment until it is polling for completed Batches, and then it only returns the GUID assignments.  (The Batch ID is addressed when it is created, so not a problem.)  I would like see some qualifier also be returned with the Batch and Document IDs--either the file name/path or possibly an index value indicating where the Document was in Batch Create process.  This would allow for direct In/Out connections to be defined.

Other Solutions:
1.  You can query the Decipher SQL database, using the Batch ID, to gather the associated Document IDs.  As I understand it, the file name is stored in the database, but it is obfuscated in version 1.x.  I am told the obfuscation is removed in version 2.1.  We will be upgrading shortly, so I may be able to report back on that in a few days.  But I not really happy with this solution as well.  One of the selling points of Blue Prism is that it is supposed to be accessible by Subject Matter Experts, and not the sole domain of Developers and Coders.  This touches on a whole different topic of Custom-Code Objects, but this would be an improvement over the B1D1 approach.
2.  There is also a method of adding Metadata, which I have not explored, but it seems to presuppose that you have the Document ID in hand before adding in the external information.  Perhaps someone else could speak to this.

You asked specifically for the 'Cons and Limitations.'  I do not want this to leave the impression that I think the product is unusable.  It is challenging, but I think it is a significant step over the Surface Automation/OCR actions native to Blue Prism.  Again, still early for me, but Decipher feels more accurate and efficient.  

I also do not want this to seem like an ambush of the product or the development team.  Ben Lyons and I have spoken, at some length, on this topic.  I am fully aware of development cycles and reasonable expectations, so I will wait to see what comes of it.

Take Care,
Red

------------------------------
Robert "Red" Stephens
Application Developer, RPA
Sutter Health
Sacramento, CA
------------------------------

[ArikaJatwani]

Great stuff @BenLyons !!​

------------------------------
Arika Jatwani Process Manager
------------------------------

[stepher]

Update:

• Our Technical Team completed the update from v1.2 to v2.1 in nearly no time.
• There are a couple of things they want to refine, but the core functionality was accessible.
• The processes that I built in v1.2 transitioned and functioned in v2.1 without a problem.
• Someone else was able to do a "quick and dirty" report and I can confirm that File Name is not obfuscated.
  • There is a flag in the "Set Up" action to maintain the obfuscation, if desired.
  • The default value is "False" (do not obfuscate)
• Again, this is not natively available through the Decipher VBO [yet?], but for those with the skills and access, it should pose no problem.

Thanks,
Red

------------------------------
Robert "Red" Stephens
Application Developer, RPA
Sutter Health
Sacramento, CA
------------------------------

[BenLyons]

Hi Red,

Good to hear from you!

Ok, I totally hear you and this is certainly an area we're working on, but let me take you through a bit of the detail.

I appreciate in most circumstances you're uploading a "document", but that's not how Decipher sees it. Instead it's considered a file as it may contain multiple documents or multiple document types and in order for the classification engine to work, this has to be broken down. So the first part of the image processing is to split a file into pages.

If classification is disabled, then sure, these will be put back together as a single document. But if it's enabled, Decipher will refer to the classification model to decide how to proceed. At this time, the initial document processing doesn't differ between whether classification is enabled or not.

The next stages of OCR and capture both process pages 1 at a time, this is in line with how the training data is constructed and helps it work quicker than if it were processing entire documents. Following these stages, the document ID is created and allocated to the respective pages. So the ID's linking them can be traced as you mentioned via SQL, but as the document ID isn't created during the upload period it's not immediately available for tracking.

And to return to the multiple documents in a single file concept, you would then have the same file name across 2 or more documents. So you not only need to know what the file name was, but now you need how many of them there are.

This is one of the things which makes Decipher super flexible in its file management, but as a result you get the challenge outlined above.

We've got a lot of great things coming in this space, including a new accuracy report and we appreciate your interest.

Thanks for your feedback, I'll make sure it gets to the right people.

Regards


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

[lookman]

Hi Red.

Thank you for taking the time to provide us with excellent feedback. As Ben mentioned, this will go to the right people and your input will only help make Decipher a better product.

Cheers

jack

------------------------------
Jack Look
Sr Product Consultant
Blue Prism
------------------------------