Models

Models are the frameworks that tell EVE what to listen for and how to respond. A model defines a hierarchy of Topics, Sub Topics, and Keywords — when a participant mentions a keyword, EVE probes on the corresponding sub-topic. Models also carry the probe and prompt wording EVE uses when those keywords are triggered.

Models are used in Guided AI and Model Only conversations.

---

Models in EVE Builder

The Models section on the EVE Builder home lists all the models you have access to, organised across three areas:

Section	Description
Models	Your personal models — created by you or shared with you
Account Library Models	Models that have been promoted to the account library, available to all users in the account
Global Models	HumanListening’s shared model library, available to all accounts

---

Creating a model

You can create a model directly from the Models section, or a new model can be generated as part of setting up a Guided AI or Model Only conversation.

Create a model from the Models section

1. Click the CREATE button at the top of the app home screen. The button label reflects the last project type you created — click the dropdown arrow and select Create Model.
2. Enter a Model Name and click OK.
3. The model editor opens — build out your topics, sub-topics, keywords, and probe/prompt wording.

Generate a model during conversation setup

When configuring a Guided AI or Model Only conversation, you can choose to:

• Generate full Model — AI generates a complete model from your context statement
• Use Generated Topics to start building a new Model — AI generates topics only; you complete the rest

In both cases you’ll be prompted to enter a Model Name. The model is then available in the model editor during setup, and saved to your Models section when the conversation is published or saved as a draft.

---

The model editor

The model editor is where you build and refine your model’s structure and the wording EVE uses during conversations.

Model structure

Level	Description
Topic	The broad category (e.g. “Customer Service”)
Sub Topic	A specific aspect within the topic (e.g. “Wait Times”)
Keyword	Words or phrases that, when mentioned by a participant, trigger a probe on the parent sub-topic (e.g. “queue”, “wait”, “hold”)

Generating content in the model editor

The model editor includes AI generation tools for building out the model:

• Generate Sub Topics — generates sub-topics for a selected topic based on the topic name and conversation context. Once generated, the button changes to Extend to add more
• Generate Keywords — generates keywords for a selected sub-topic. Once generated, the button changes to Extend to add more

You can also type topics, sub-topics, and keywords manually at any point.

Probe and prompt wording

Each sub-topic has its own probe and prompt wording — what EVE says when a participant triggers that sub-topic, or when EVE prompts on an unmentioned sub-topic:

Field	When used
Probe wording	When a participant mentions a keyword associated with this sub-topic
Prompt wording	When EVE prompts the participant to share their thoughts on this sub-topic (because they haven’t mentioned it yet)

Write probe and prompt wording in a natural, conversational style — these are the actual words participants will read during the conversation.

Adding and editing keywords

To add a keyword to a sub-topic, click the Add Keyword button on that sub-topic row. To edit an existing keyword, click the keyword in the model editor. Both actions open the keyword dialog.

The keyword dialog

The keyword dialog has three parts:

i. Type your keyword or phrase

• Enter the word or phrase in the Keyword field
• Check Include Plurals to automatically match the plural form of the keyword alongside the singular

ii. Choose the tagging rule Select how EVE should match the keyword against participant responses. See Tagging rules below.

iii. Fine-tune your tagging rule Specify associated words that should influence whether a match is made. These fields are available for all tagging rules except All of these words, in any order:

Field	What it does
Include with associated words	Also match when these words appear on either side of your keyword
Exclude with associated words	Do not match when these words appear on either side of your keyword
Include with in-sentence words	Also match when these words appear in the same sentence as your keyword
Exclude with in-sentence words	Do not match when these words appear in the same sentence as your keyword

Avoid duplicating keywords

Avoid using the same keyword or phrase in more than one sub-topic — this leads to double-tagging. If you add a keyword that already exists elsewhere in the model, the dialog will warn you before saving.

Special characters

The following characters cannot be used in keywords: \ . * + ? ( ) [ ] | ^ $ { }

---

Tagging rules

The tagging rule controls how EVE matches a keyword against participant responses. Choose the rule that best fits how the word or phrase is likely to appear in text.

Tagging rule	What it matches
Exact word or phrase	Comments that contain this exact word or phrase
Starting with	Comments that contain words starting with these letters (e.g. “deci” matches “decide”, “decision”, “decisive”)
Ending with	Comments that contain words ending with these letters (e.g. “fair” matches “fair” and “unfair”)
Containing	Comments that include words containing these letters anywhere within them (e.g. “confirm” matches “confirmed”, “confirms”, “unconfirmed”)
Any of these words	Comments that contain any one of the words entered
All of these words, in any order	Comments that contain all of the words entered, regardless of order
All of these words, in this order, in a sentence	Comments that contain all of the words entered, in that exact order, within a single sentence
Do not contain	Comments that do not contain this exact word or phrase
Custom (regular expression)	Comments that match a custom regular expression pattern

Which rule should I use?

Exact word or phrase is the right choice for most keywords — it’s precise and avoids false matches. Use Any of these words when you want a single keyword entry to cover several synonyms. Use Starting with or Containing when you need to catch word variations (e.g. “satisf” to catch “satisfied”, “satisfying”, “satisfaction”). Use Do not contain to exclude responses where a word appears in a context you don’t want to tag.

Do not contain behaviour

When a participant response matches a Do not contain keyword in a sub-topic, it won’t be tagged against any other keyword in that same sub-topic — even if it also matches other keywords there.

---

Managing models

Edit a model

1. Click the model name in the content table to open it, or hover over the row and click the view icon.
2. Make your changes in the model editor.
3. Save your changes.

Republish after changes

If you edit a model that’s linked to published conversations, those conversations won’t reflect the changes until you republish them.

Duplicate a model

1. Hover over the model row and click the more icon (⋮), then select Duplicate.
2. Enter a name for the new model and click OK.

Rename a model

1. Hover over the model row and click the more icon (⋮), then select Rename.
2. Enter the new name and click OK.

Delete a model

Check for linked conversations

Deleting a model that is linked to published conversations will affect those conversations. Check the model’s usage before deleting.

1. Hover over the model row and click the more icon (⋮), then select Delete.
2. Click YES to confirm.

Share a model

1. Hover over the model row and click the Share icon.
2. In the Share dialog, enter the email address(es) of the people you want to share with.
3. Click the Edit icon and select the required Role from the permissions drop-down.
4. (Optional) Check Notify people and add a personalised message.
5. Click Save.

View model usage

To see which conversations are using a model, hover over the model row and click the more icon (⋮), then select Model Usage.

---

Account library models

The Account Library is a shared space where models can be made available to all users in the account — useful for standard frameworks that multiple teams use.

Promote a model to the account library

1. Hover over the model row and click the more icon (⋮), then select Promote.

The model moves to the Account Library Models section and becomes available to all account users.

Demote a model from the account library

1. Hover over the model row and click the more icon (⋮), then select Demote.

The model moves back to the personal Models section.

---

Frequently asked questions

What’s the difference between Topics, Sub Topics, and Keywords?

Think of them as a hierarchy. A Topic is a broad area of interest (e.g. “Price”). A Sub Topic is a specific aspect within that topic (e.g. “Value for Money”). Keywords are the words or phrases that trigger probing on that sub-topic — if a participant says “expensive”, “cheap”, or “worth it”, EVE probes on Value for Money. Topics and Sub Topics organise your model; Keywords are what EVE actually listens for.

Can I use the same model in multiple conversations?

Yes — a model can be linked to any number of conversations you have access to. Changes to the model don’t automatically update linked conversations; you’ll need to republish each conversation after making model changes.

Can I use a model from a different sub-account?

Only if it has been shared with you or promoted to the Account Library. You can only link models you own or that have been shared with you.

What’s the difference between an Account Library Model and a personal model?

Personal models are only visible to you and people you’ve explicitly shared them with. Account Library Models are available to all users in the account without needing individual sharing. Promoting a model to the library is useful for standard organisational frameworks that multiple teams rely on.

What does “Model Usage” show?

Model Usage shows you which conversation projects are currently linked to the model. Useful before editing or deleting a model, to understand what impact your changes might have.

If I edit a model, do I need to update every linked conversation?

Yes. Editing a model doesn’t automatically update published conversations that use it. You’ll need to republish each linked conversation for the changes to take effect in live surveys. The survey itself will then automatically pick up the latest published version of the conversation.