Uipath document ocr extracting incorrect value with high confidence

Help Document Understanding
1

Hi everyone, I have a query related to OCR.
We are using UiPath document OCR in document understanding framework, we are extracting Amount, and the extracted value is incorrect, still the confidence score is 99%. This data has to entered into system, but there can be consequences to this kind of scenario. How we can handle these kind of scenarios given that we dont have any rule to validate value ?

Actual value = 25-
Extracted value = 25
Confidence score = 99%

Extraction method - ML extractor
Data type in AI centre - text

2

Hde you tried using any other extractor like form based extractor for that particular field.

3

Hi @Nirjara_Jain_IND ,

Could you let us know if the type is also the same in the Taxonomy field defined ?

Also, Try enabling the field as Multi-Line in the Data Labelling Session/Taxonomy field and Check whether the - gets extracted.

4

We have semi structured document, and documents have lot of variation.

5

In Taxonomy, Data type is also text. Where can I get this multiline facility in AI Centre and Taxonomy. I think Multivalue facility is only with simple field. I am trying to extract data from table.

6

use different extractor for that particular one field only.not for the entire document

7

Hi,

Can you try changing the taxonomy data type to number for this field. This way, it automatically removes special characters in the output.

You can check the output on Action center (click on the field and it shows the “Value” box that has the actual captured and cleansed value)

In addition, the output excel file also has formatted value under “line items - formatted”

Let know if this works…