FAQ
Frequently asked questions related to NLX
Bots
What is required to create a build and deploy my bot?
At a minimum, an NLP engine must be integrated at the workspace level, one intent must be attached to your bot, and at least one communication channel matching the intent(s) attached must be created for your bot.
Can I design and test bot conversations before selecting an NLP integration?
Designing a conversation flow and setting up resources to be used within a flow may be done without an NLP in your workspace. However, to test conversation flows, an NLP must be integrated, as each build and deployment tests how the NLP processes and understands a user.
How do I publish my bot and make it live to users?
Once you create a build that is successful (indicated by a green Built status), you may click the Deploy button in the production column of the Deployment tab. This will officially deploy your bot to be available to users.
How do I take a bot offline?
Select the Details link that appears beside the current deployment. From there, locate the Delete deployment option found at the bottom of the pop-up. This will take the bot out of a deployed state. You may re-deploy the same build or create a new build and deploy to make your bot available again.
My build failed. What do I do?
NLX runs through all conversation flows of your bot to confirm it will work properly before you deploy it to the world. For a failed build, select the Details link beside the failed build status to view a report of detected errors to address.
Intents
Should I create a new flow or create a secondary channel of the same flow?
If a flow is to be identical when in use by another communication channel, you may simply add a secondary channel to the same flow. Any changes made to the flow will impact all channels associated with that flow.
If considerable differences in steps and messaging need to be designed for another communication channel, create a new flow within the intent. Changes made only impact the channel(s) established for that flow.
How do I translate the messaging of my intent flow and not just training phrases?
As languages are attached to your intent through its Languages tab, you'll notice you have the option to manage translations (manual or auto). This translation process, however, only applies to any training phrases belonging to your intent. To translate messaging in one or more flows of your intent, hover over the flow card and select the globe icon. From this menu, you'll be able to initiate translating your flow to each supporting language of your intent.
What's the difference between multiple messages and message variations?
Multiple messages on the same node are displayed/read off in succession and in the order they're arranged. This primarily allows you to break up large paragraphs of text in chat channels so they appear as individual chat bubbles.
Message variations provide a bank of message options for your bot to select randomly when a user hits that node in the conversation. This is especially useful if the same node has a chance to be reached twice in a single session.
Can I playback my voice bot's messaging to hear how it would sound to users?
When an intent flow has a voice channel selected for its primary or secondary channel, a play icon will display next to any bot messages added to a node. Though a different bot persona may be selected outside of NLX, you may hear what your bot will generally sound like reading off the messages you've designed.
How do I change the inflection or get advanced control on how a bot sounds?
When drafting messages for voice channels, you can use SSML tags (Speech Synthesis Markup Language) supported by your NLP. Inserting SSML into your messages can control volume, pitch, pauses, emphases, breathing, and more.
For Amazon Lex, check their supported SSML guide here.
For Google's Dialogflow, check their supported SSML guide here.
Why do I need to give an attached slot a name?
Whether custom or native, Slots are available for use across any intent flow once attached, and they can be attached multiple times to the same intent (e.g., an NLX.Country slot attached twice for a step that collects an international departure location and another for an arrival location). Because AI models (NLP or LLM) reference these names when resolving the parameter or routing users according to their preferences, the slot names must be unique and distinguished from others. Therefore, the name you give your slot is the name that will appear when referencing it in your flow.
Why can't I rename my intent?
Intents are resources that may be attached to several bots in a workspace and can break or disrupt deployments if renamed. Instead, you may clone your intent to provide a new name, if needed.
Slots
Can I reuse a custom slot across multiple intents?
Absolutely. In order to use a slot with any intent in your workspace, it must first be attached to the intent. From there, you may decide how it's used in training phrases and/or the intent's flow as well as determine the path that continues forward from each selected value in a User choice node.
Should I use a built-in slot type or custom slot?
Several built-in slots are provided in NLX that capture common values that would otherwise be painstaking to recreate: list of countries, time, dates, common first names, etc.
Create custom slots if built-in types don't exist and when you want to provide fairly fixed choices that users select from or would normally use to clarify a preference: Yes/No, reason for cancellation, t-shirt size, etc.
For values that may change regularly or are massive in scale (e.g., showtimes, available hotel rooms, grocery store inventory, etc.) create a Data request instead.
Why do I need to attach a slot to an intent?
Since slots can be used in training phrases or within an intent flow when users make selections on User choice nodes, the slot's values then become part of the data that trains the NLP to understand and process the users requests.
API integrations
When should I use a static response implementation for Actions or Data requests?
We recommend using a static JSON response in the early stages of designing your bot conversation or while an API may still be in development. This prevents any blockers and allows you to test the functionality and traversal through your intent flows.
Where can I add the payload for fields set up in my Action's Request model?
This can be done in the canvas builder when adding your action to an intent flow. You'll first need to reference your action by placing an Action node on the canvas. Select the action you want to use from those in your workspace and enter values in each payload field using the action node's side panel. Payload fields are auto-generated from the request model you created.
Why is a Lifecycle hook attached to a bot rather than used in an intent flow?
Since the three lifecycle events are pre-defined (start, end, and escalation), lifecycle hooks don't need to be added to any of the multiple flows that could be attached to a bot in the way that data requests or actions do. The bot is able to define when these lifecycle moments occur for simple one-time setup of your webhook, which can be used for triggering surveys, profile lookups, etc.
Voice+
What happens when a Voice+ SMS is accessed by a messaging app that's not a mobile device?
Voice+ experiences continue to work as designed, regardless of the device used to open and navigate through the SMS link. A bot assistant will continue to be synced and will read aloud all steps through the user's mobile device as they view their weblink on their desktop, phone, smart watch, or tablet.
Last updated