NLU Model (Re)-Training

Shaksham
Kilo Contributor

‎02-10-2022 10:12 PM

Hello All,

I hope you are doing well.

I have a question regarding (re)-training the NLU model. Currently, I am following these steps for model building/training:-

  1. Create a 'Data Source'. 
  2. Upload the dataset in the required format.
  3. Create a transform map.
  4. Load all records.
  5. Run transform.
  6. Model the now is visible under 'Natural Language Understanding --> NLU Models'.
  7. Train the model, and publish.

Now, suppose I have to update my dataset (either new intents have been added or I have more data). How do I go about doing it without deleting everything and redoing the process? (This is what I am doing at the moment. However, it is breaking the flow and the mapping between topics and intents.)

Your inputs are much appreciated 🙂

Thanks

Shaksham

0 Helpfuls
  • 742 Views
2 REPLIES 2

Prakash36
ServiceNow Employee
ServiceNow Employee

‎02-10-2022 10:58 PM

Hi.. What is the content in your datasource ? 

If the changes specific to intents or utterances. you can go to the specific table & upload the data. Then you can train the nlu model.

0 Helpfuls

Shaksham
Kilo Contributor
In response to Prakash36

‎02-10-2022 11:05 PM

I upload the file having the following columns: Model, Intent, Utterance, Language

 

Example:

Version 1: 10 intents - 5 utterance each (Model 1)

Version 2: 12 intents - 10 utterance each (Model 2)

 

I want to understand how can I update my model (Model 1) when my data is changed from Version 1 to Version 2 without deleting anything.

0 Helpfuls