How to train Decipher for a field which is most likely to be blank

[Tejaskumar_Darji]

Hello Team,

I have a field that can have value in some rare scenario otherwise it will be blank.

Currently, Decipher picks the value when the field has some value.

When there is no value for the field and only header is present Decipher starts picking a value from the adjacent field or bottom label/field whatever is nearby.

I drew the region manually multiple times but it is not saving any regional property in memory

How can I train to decipher to look for a specific region always?

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

[rizistt]

Hi @Tejaskumar_Darji,

Is the field that you're trying to capture is part of a table?

A. In case it is not, try these misc. parameters and mark the region where it appears:

1. FormFields=On
2. StrictPosition=On

We encountered the same issue but these misc. params made it work.

B. In case it is, try using Regex with the StrictRegex flag, or by including or excluding certain values to make sure it extracts the correct fields. By default, it shouldn't pick anything that doesn't appear under the specified column header (for tables of course) - at least what I noticed.

Let me know if that works.

------------------------------
Rizwan Tayyab
------------------------------

[BenLyons]

Hi,

I agree Strict Position will help lock down the location and this will be layout specific, so won't affect different document templates.

However, a note on the strict regex flag. Regex is always strict when used as validation, unless the soft validation flag is set. What this flag really means is, you only ever want it to match the regex value, even if it means it's in a different location. So this will override other training/DFD parameters if it finds a matching string elsewhere in the document, this would include overriding sample headers and may conflict with strict position.

If the region Decipher's incorrectly matching can be defined as another field, it may prevent it from incorrectly assigning it. You can then just set it as non-exportable, so it doesn't interfere with your process.

Thanks

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

[AhmedBasheer_M]

Hi , 

This is my Input Text 

"Can you send me the other invoices against our PO P00XXXXX? I have the invoice for

$5.20 inv# 2591XXXXX. I just need the 3 remaining invoices on the PO. Account #
356XXX
Order # 259130290002"

I need to get the "Account #" value which falls in the next line. When I use sample header strategy, I am not able to get the values because the value is not beside the header.

Kindly advise which Misc.Parameter I can use to extract this data. 

Best regards
Ahmed B.

​

------------------------------
Ahmed Basheer M Project Manager
RPA Developer
Capgemini India
Asia/Kolkata
------------------------------

[BenLyons]

Hi Ahmed,

This would technically fall into the category of "unstructured" text and would benefit from the NLP functionality available with Decipher.

That being said it may still be possible to extract the data without NLP. Is the account number in a specific format which could be determined using Regex?

You could then use Regex in the Format Expression field in the DFD configuration. Or alternatively, you can read the entire text block and use one the Regex formulas, such as REGEXSEARCH or REGEXSPLITGET. You can find the details for this on the formulas help page.

Do think this may work?

Thanks

------------------------------
Ben Lyons
Senior Product Specialist - Decipher
Blue Prism
UK based
------------------------------