Intents: Training phrases
Add or modify user utterances (training phrases) to an intent in your NLX Dialog Studio workspace
What's the Training phrases tab?
Training phrases are the samples of what users might say to a bot to trigger a task that the intent satisfies. Your chosen NLP becomes trained on these samples and matches the user's utterance to a training phrase.
For an OrderRoomService intent, training phrases such as 'Order something to eat,’ 'I want room service’ or even ‘I’m hungry’ can let the bot know the user wants OrderRoomService.
The minimum number of training phrases recommended is 15. This provides a strong variety of ways in which users may construct their request. However, intents that have much larger samples of training phrases over others may cause your NLP engine to favor those intents and match more utterances to them.
NLP engines also help address variations, misspellings, and other unexpected conversational elements.
Access
To access this feature, click the Training phrases tab of your intent:
Add Training phrases
Click + Add new training phrase
Enter SHIFT + RETURN/ENTER to bring up another field. Min. of 15 phrases is recommended
Make changes by typing into the phrase field. Remove a phrase by clicking the remove icon
Click Save
Leverage slots in training phrases by attaching them to your intent first.
Saved changes do not take effect with bots that are already deployed. To experience changes to training phrases with bots in production, create a new build and redeploy.
How to craft a good training phrase
Predicting what users might say to a bot in order for the right intent to be triggered is a challenging task. Users may say a lot or say very little when explaining to a bot what they want to accomplish.
Take a deep dive on organizing intents and developing training phrases with our favorite best practices.
Length
To start, keep training phrases short. NLPs are able to identify filler words , misspellings, uncapitalized words, and politeness from user utterances, so no need to worry about creating multiple variations to account for them:
✔️place an order for room service
✔️would like to order room service
❌Can I please place an order for room service for my family this evening?
Leverage Slots
Slots help you capture dynamic info that determine the parameters to a user's request or choice. Small/medium/large or yes/no are common examples. Introducing slots into training phrases can reduce creating different intents that would have similar training phrases or prevent the need for drafting multiple variations of a single training phrase:
✔️would like a {size} pizza
✔️look up {accounttype} account balance
❌look up savings account balance; look up checking account balance, etc.
Experiencing issues with incorrect utterances matching to an intent or correct utterances getting unknown detection? Be sure to even the playing field across intents tied to your bot by having a consistent sample size between them. Eliminate unused training phrases or add to a sample to keep them all balanced.
Generate using AI
Want to enhance your Training phrases pool in record time? Use the Generate using AI feature:
Create a sample set of at least 5 training phrases to allow the AI to learn and expand on phrases with similar characteristics
Click Save
Select Generate using AI to create phrases in batches of five
Modify before or after adding the phrases to your pool
Repeat the process as needed (recommended pool size: 15-30)
If you have slots you'd like GenAI to reference in its training phrases, include them in your sample set first.
Upload/download training phrases
Upload: Allows for a .json, .csv, or .txt file to be ingested with training phrase data
Download: Prompts existing training phrase data to be downloaded to a .csv or .json file for editing. Tap the .csv/.json button beside the Download link to swap formats. Filename format: <IntentName>-Training.json
Last updated