---
sourceDocument: Australia Enable AI
sourceDocumentLink: https://servicenow-prod.fluidtopics.net/r/pt-BR/intelligent-experiences

 Release :

    - australia

ft:locale :

    - pt-BR

ft:publication_title :

    - Australia Enable AI

ft:clusterId :

    - platai

bundleId :

    - platai

workflow :

    - Platform


---

# Create a test set

# Create a test set {#ariaid-title1}

* Versão de lançamento: Australia
* 
* Atualizado 12 de mar. de 2026
* 
* ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 2 min. de leitura

To create or add to an NLU test set, you can upload a file
of test utterances matched with correct intents. Use the test set to assess the performance
of your model.

## Antes de Iniciar

* Ensure that the NLU Workbench - Core plugin, NLU Workbench plugin, NLU Workbench - Advanced Features plugin and Predictive Intelligence plugin are all installed and activated
* You can use test sets with NLU models for Virtual Agent and AI Search.
* Role required: nlu_editor, nlu_admin, or admin. The editor must be assigned to the model.
{#import-batch-test-set__ul_hbx_rzc_wrb}

## Por Que e Quando Desempenhar Esta Tarefa

Your CSV or XLSX (Excel Workbook) file should contain a table that pairs your test utterances with the intents that you expect for them. Your file can contain up to 10,000 utterances. Ensure that the file has columns titled "Utterance" and "Expected intent".  
Nota:  
For test sets in languages other than English, you must add the glide.import.csv.charset system property with the value UTF-8. See [Import sets
properties](https://www.servicenow.com/docs/access?context=r_ImportSetsProperties&version=australia&pubname=australia-integrate-applications&ft:locale=en-US).

For the most accurate test results, include utterances that the model is likely to
encounter from your users. Ensure that you have test utterances covering all the
intents in your model.  
Aim to include about 10% of test utterances with no expected intents. Including utterances with no expected intent helps assess your model's ability to detect irrelevant utterances that should not have any intent predicted.  
Nota:  
To indicate that a test utterance in your file has no expected intent, the value for "Expected intent" should be empty.
{#import-batch-test-set__table_bdr_zc2_g4b__entry__2}

| Utterance | Expected intent |
|-|-|
| Let me have a burger | Order |
| I want to pay | Payment |
| Get me something sweet | Order |
| Is the restaurant open |   |
| Something wrong with my payment | Order, Payment |
| Total cost | Payment |
[Tabela 1. Example test set table]

{#import-batch-test-set__table_bdr_zc2_g4b}

## Procedimento

1. Navigate to AllNLU WorkbenchNLU Advanced FeaturesMulti-model Batch Testing.
2. Click Test sets
3. Click Create test set.  
4. Choose a name for the test set.
5. Choose a language.
6. Click Select file and choose a CSV or XLSX (Excel Workbook) file.
7. Click Create.  
   Your test set appears in the list.

## O que Fazer Depois

Use the test set to run a test on your models. To learn how, see [Test your model](https://servicenow-prod.fluidtopics.net/AxsqSC_egQCQEAVOd3GDdg "Test your Natural Language Understanding (NLU) model against its default test set. Testing helps determine how your model is performing with the current content.") or [Run a multi-model batch test](https://servicenow-prod.fluidtopics.net/tK4pOTS2tNxrpnLOQusjCw "Test multiple Natural Language Understanding (NLU) models against a test set. Evaluate the quality of your models and refine them to improve intent prediction.").  
After you have created a test set, you can add more utterances to it. When viewing a test set, click Import utterances.Figura 1. Test set page with import utterances button  
Select a CSV or XLSX (Excel Workbook) file with your additional utterances to import. Figura 2. Import utterances window

Click Import. The system adds your utterances to the test set.
After importing, rerun any tests that use the test set.