Suggestable methods or options to extract unstructured data from email body

Help Robot
1

Suggestable methods or options to extract unstructured data from email body

tried to extract data from unstructured emails.

2
  1. Try Regex.
  2. Try String manipulation like split function and all.
  3. Check for Document understanding method/communication mining
3

I have a scenario where I need to extract two fields from an unstructured email body — Part Number and Quantity.

The challenge is that these fields are not consistently mentioned in the emails. For example,

  • Part Number may appear as: Part, P/N, Part#, Part No, etc.
  • Quantity may appear as: Qty, Quantity, Pcs, etc.

Given this variation in the field names, I want to know which approach or tool would be best suited for extracting this data — IXP, Agent Builder, or Gen AI.

Could you please help me choose the most suitable option for this scenario?

4

Hi, @Medikonda_Vijay

UiPath IXP is best for extracting Part Number and Quantity from unstructured emails with different names. It uses AI to understand varied terms, fast and no coding needed. Agent Builder builds AI workflows but depends on tools like IXP for extraction. Generative AI is flexible but needs more setup and lacks built-in controls.

So, for easy, accurate, and reliable email data extraction, go with UiPath IXP.

5

We tried with IXP, but it could not recognize the fields while training, please share the steps, if any

6

@Medikonda_Vijay

To improve UiPath IXP field recognition:

Add diverse sample emails showing all field name variations.

Define clear field names and aliases like Part, P/N for Part Number.

Label fields completely and consistently during training.Write simple, specific prompts for each field.

Use human validation to correct extractions and retrain.

Group related fields logically to improve accuracy.

Use custom regex only for simple, fixed patterns.

However, continuous tuning and validation are important to achieve the best results in any unstructured data scenario.

7

@Medikonda_Vijay,

Try using GenAI activity Named Entity Recognition

8

@Medikonda_Vijay,

Do not create duplicate topics.

9

@Medikonda_Vijay

Welcome to the community

Regex or string manipulation if easy and more similar layouts

else go wtih AI activities and use a good prompt to get the values

cheers

10

Hello @Medikonda_Vijay

Do you have a few examples of how these mails could look like and which information you would like to extract?
Then we can help you find the best suited solution for your use cases.

Regards
Soren