## Objectives

The objectives of this section are:

- Becoming familiar with the different ways of language understanding in Teneo
- Knowing what Match Requirements are and how to use them to understand the user intent
- Knowing what it means that an input receives _annotations_, among others class annotations
- Knowing how to manage the Machine Learning classes in a solution

Objectives - The Power of Machine Learning - Fundamentals

Objectives

Teneo uses a hybrid model when understanding language. This means that in Teneo you can mix machine learning algorithms and linguistic conditions for Natural Language Understanding (NLU). This approach gives you both the power of machine learning and the precision and linguistic scalability that linguistic conditions provide.

![Build - Language Understanding: ASO228-Hybrid-static](//images.ctfassets.net/2vbtnveccz5s/6aZgf0LOhOMQJ5aqOkabt0/8fe981c48c6b3c5385a271b53aeab5a1/ASO228-Hybrid-static.png)

Language Understanding in Teneo  - Learning Journey

Language Understanding in Teneo

Teneo's model for understanding language provides us with multiple ways to recognize an intent using [Match](/studio/flows/concepts/triggers) in triggers.

![3 matches](//images.ctfassets.net/2vbtnveccz5s/1YLiXdkSyoSaIixSRlVf0Z/0992bc12e23a9d337c837aed37835827/3_matches.png?classes=w60)

!!! More information on Match can be found [here](/studio/flows/concepts/triggers#match-requirements)

## Class Match

Classes use machine learning to determine what an input is about. You train a class by providing training phrases (called 'learning examples' in Teneo) in the [Class Manager](/getting-started/class-manager) that are as close to real user inputs as possible. You can either create your own Class or let Teneo generate one for you as seen [here](/getting-started/your-first-flow#populate-the-trigger). Training and test examples for classes can also be generated using Teneo's copilot features, as seen [here](article/getting-started/class-manager?client=studioWeb#generate-examples). We recommend starting with 10-20 example inputs per class.

## TLML Syntax Match

Teneo Linguistic Modeling Language (TLML) Syntax Matches use a [language condition](/studio/language-understanding/concepts/language-conditions) to understand what users are saying. The TLML Syntax Match is expressed using Teneo's [condition syntax](/studio/reference/condition-syntax). TLML Syntax Matches look at words used, their order, synonyms, and other linguistic properties to understand what an input is about. A language condition to cover inputs like 'I would like to order an espresso' could look like this, for example:

```tlml
%HOW_DO_I_BUY.PHR & %COFFEE.NN.SYN
```

You can either hand-craft a language condition yourself, or let Teneo automatically [create it for you](/article/build/flows/how-to/use-conditional-match-requirements#generate-a-condition) from a set of examples, after which you can edit the condition.

## Language Object or Entity Match

Just like the TLML Syntax Match, the Language Object and [Entity](/article/build/flows/concepts/entities) Match use a single [Language Object](/studio/language-understanding/concepts/language-objects) as conditions.  
An example of each can be found here:

*   [Language Object](/studio/conversations/how-to/create-a-custom-order-group#create-a-new-flow-and-add-a-match-requirement-for-a-language-object)
*   [Entity](/getting-started/build-a-slot-filling-flow#configure-the-transition)

## Which Match should I use?

The possibility to combine all of the Match Requirements provides us with the power of machine learning and the precision of condition syntax, but it can be difficult to determine which trigger to use for which purpose. The following are suggestions that can help us make the best of the intent.

!!! Note that Match Requirements can be used in both Triggers and Transitions.

### Classes only

The class matches are the 'go-to' use for Triggers when creating a flow. When the class is performing as intended, the flow is triggered correctly.  
If you find that the class is a bit too greedy, i.e. triggering on too many inputs, you can require a higher *confidence* by adding a context restriction __Class Confidence Medium__ or __Class Confidence High__. This is demonstrated in the page on how to [combine match requirements](/studio/language-understanding/how-to/combine-match-requirements).

!!! A flow can have multiple class triggers. This allows us to attach [listeners](/studio/conversations/concepts/listeners) to each trigger and apply different confidence levels.

### Classes and TLML Syntax

When the Class match is performing well, but we also have a language condition that we trust more (specifically using Entities or rich synonyms), we may choose to add both a class and a corresponding match requirement to our flow. In this case, it could be one [Entity](/article/build/flows/concepts/entities) that might be valuable to cover, or an advanced set of words using a [Condition](/studio/language-understanding/concepts/language-conditions) in addition to adding training examples to your class.

The trigger, in this case, should be stored in an order group above the order group containing the trigger with the Class Match Requirement. This way, the Conditional Match Requirement trigger will be evaluated first. If an input happens to not be caught by the Condition created by the user trigger, the trigger with the Class Match Requirement might catch it. Having both a trigger with a Conditional Match Requirement and a trigger with a Class Match Requirement also has the advantage of allowing us to easily improve the performance of the flow by adding more training data to the trigger (without updating the language condition of the other trigger).

If necessary, we can increase the *confidence* of the class by applying the pre-defined [context restrictions](/studio/conversations/concepts/context#predefined-context-restrictions) 'Class Confidence Medium' or 'Class Confidence High'. We can do this if the class trigger turns out to be too greedy.

!!! More information on creating a new context restriction can be found [here](/studio/scripting/how-to/define-a-scripted-context).

### TLML Syntax only

You can choose to not use a class at all for certain intents. This is a good approach when the classes created is not performing well enough, it is stealing inputs, or it is confusing the solution model, for example. A language condition is much more reliable in such cases. Examples are [follow-up](/getting-started/deal-with-follow-ups) triggers, or triggers that should trigger when inputs consist of just a [particular keyword](/studio/conversations/how-to/create-a-custom-order-group#create-a-new-flow-and-add-a-match-requirement-for-a-language-object) (or synonyms thereof).

If the language condition of the Match is very precise, matching on very specific inputs, we will most likely want to add it to an [order group](/studio/conversations/how-to/create-a-custom-order-group#adjust-the-group-ordering) above the other triggers' groups. This will prevent a less precise trigger from accidentally catching the input from this trigger, which is more precise.

If the trigger is fairly broad, on the other hand, and we want other triggers to be evaluated first, we can add it to an order group below the precise trigger's group. This will prevent the broad language condition from inadvertently catching input that could have been caught by a more precise trigger. More information on ordering can be found [here](/studio/conversations/concepts/ordering).

!!! Note that TLML Syntax Match triggers cannot be improved using the [Optimization](/article/optimize/refine-your-bot/how-to/improve-your-class-triggers) panel, or by just adding example inputs to the trigger. To change the behavior of a trigger with a TLML Syntax Match, you need to modify its [language condition](/studio/language-understanding/concepts/language-objects).

### Combining Matches
This in essence is a combination of the Matches available to an intent trigger. The trigger uses a Class Match on top of having a Condition/Entity/Language Object Match, plus some TLML Syntax constraint. We use this when the trigger is performing OK but we specifically need to constrain the current class with some conditional part.

For example, a trigger 'I love you' includes a Class Match 'I_LOVE_YOU' with 20 training examples. Adding a Condition allows us to force the intent containing the word 'you' and not trigger when using a negation like 'I **don't** love you' using the following condition on top of the Class Match:

```tlml
%YOU.REC & !%NEGATION.REC
```

!!! Combining Matches is also supported in transitions.

Since these triggers make use of an intent trained by a class, we can still improve this class using the [Optimization Panel](/article/optimize#optimize-classes).


Build - Match Requirements

Match options

Annotations can be considered 'labels' that are attached to inputs. A word, word sequence, sentence, or full input can have multiple annotations that you can use in your language conditions. In most cases machine learned models are used for annotations, like Named Entity Recognizers (NER) or Classifiers. Teneo adds various such annotations out of the box. Just like [language objects](/studio/language-understanding/concepts/language-objects), annotations are building blocks for [language conditions](/studio/language-understanding/concepts/language-conditions). 

The main components of an annotation are:

* its name
* its (optional) annotation variables

The name reflects what is annotated. The annotation variables can contain extra information related to the annotation. For example, an annotation __USER_ASKS_ABOUT_OPENING_HOURS.TOP_INTENT__ may have a variable containing the confidence score. Annotation variables can be retrieved in the same way as [entity variables](/article/build/flows/concepts/entities#entity-variables).

## Class annotations
When Teneo receives an input, Teneo adds annotations for Class Match Requirements. The class with the highest confidence score will be annotated as the top intent, for example: __WHERE_DO_YOU_HAVE_STORES.TOP_INTENT__. For classes that have a confidence score that is close to the top intent, additional annotations can be added, like __DO_YOU_HAVE_A_STORE_IN_CITY.INTENT__.

When a class match requirement is generated in a flow, a class label is automatically created for that trigger. But you can also specify it yourself in the [Class Manager](/getting-started/class-manager) window.

![class name](//images.ctfassets.net/2vbtnveccz5s/4yKpCJRKTfQ0qoXfgRzVKJ/460ff31e80badd75e75f29a021d1b07c/class_name.png)

## Inspecting annotations in Tryout
If you want to see how an input is annotated, enter the input in Tryout and open the 'Advanced' window.. You will find all annotations in the [Input](/article/optimize/refine-your-bot/how-to/how-to-use-try-out#try-out-advanced_input-and-intent-classification) section under 'Annotations'. Hovering over an annotation will display additional information that is stored in annotation variables.

![Build - Annotations: 2-inspecting annotations](//images.ctfassets.net/2vbtnveccz5s/3tsjIHoqpOhKEJIxhl7WR8/7fb21092792e286c3164b4942bdbf14d/2-inspecting_annotations.PNG)

Class Annotations - Learning Journey

Introduction to annotations

Teneo offers a wide range of Match requirements to add to your trigger. One of them is the Machine Learning-based Class. We know from the previous page, [Your First Flow](/getting-started/your-first-flow#populate-the-trigger), how to generate a Class while working on a flow. This chapter will be about maintaining these classes.

Located in the ribbon bar is the 'Class Manager'. Clicking it will open the Class Manager in a new window that looks like this:

![class manager window](//images.ctfassets.net/2vbtnveccz5s/jlZmiyU0gRw9LCmnlEdPG/540f2e35c24c6349fcd2429929b572f9/class_manager_window.png)

The left section of the Class Manager is a list of the classes inside the solution. Selecting one will let the user see the training and test examples. To the left of that, the user can also see which flows the class is being 'Used by' and which test examples can be found in those flows.

!!! It's important to acknowledge that the training data is inside the Class Manager; changing it will fully affect the flow using the class, and it is only possible to change it from the Class Manager. The test data, however, can be changed from either the flow or the Class Manager.

## Add a training example

If we select the 'TELL_ME_ABOUT_YOUR_COFFEE' class, we can see that the solution doesn't have training data for the coffee type 'Cortado', even though it is a popular coffee type.  

If we want to add this coffee type inside the Class we'll need to do the following.

1.  Open the 'Class Manager' and select 'TELL_ME_ABOUT_YOUR_COFFEE' class.
2.  Click on 'Edit' in the ribbon bar and add the training data `Do you serve Cortado`.
3.  Hit 'Save'.

Now you'll see the before and after we added training data for the coffee type 'Cortado'.  

![1-add training example](//videos.ctfassets.net/2vbtnveccz5s/6rNYDrWQFZhGK0ewEihvZQ/12f41d3fdcbcab52e582d37e9276afad/1-add_training_example.mp4)

!!! When you go back to the Tryout window you'll see a small text saying, 'Model training in progress', this indicates that the changes you've made are saved and the platform is currently re-training the classes in the machine learning model. This will change to 'Model update completed' when it's done.

## Add a test example
It is important to note that training examples can only be added from the Class Manager page once the Class has been created. It is not possible to add training examples from the flow editor page. This will only add 'Test Data' that will be used when [testing your bot](/article/build/test-your-bot/run-automated-tests) and not 'Training Data'.

![test data](//images.ctfassets.net/2vbtnveccz5s/5PAuwRoyGe3pnDR3NjTXsA/0d31b3b9621bf4894f910a957f29c2f1/test_data.png)

Test data can also be viewed and added in the Class Manager itself. Let's try that now.

1.  Open the 'Class Manager' and select the 'TELL_ME_ABOUT_YOUR_COFFEE' class.
2.  Click on 'Edit' in the ribbon bar and click on the 'Test Examples' tab.
3.  Add the test data `Do you serve espresso?`.
4.  Hit 'Save'.

![class-manager-test](//videos.ctfassets.net/2vbtnveccz5s/46kMR0JlMSI8lO3djrATQD/3a47c87f1db9841805e8900489f2bb3d/class-manager-test.mp4)

## Generate examples
Teneo's copilot features allow you to quickly generate new training and test examples with the help of Generative AI. These can be generated based on existing examples or based on the class description.

!!! In order to enable copilot features, make sure to first [define settings to use Generative AI](/documentation/latest/home/teneo-7.4/teneo-studio#teneo-studio_generative-ai_studio-account-page).

### Training examples
Let's generate some training examples based on existing examples.
1. Open the Class Manager and select the 'TELL_ME_ABOUT_YOUR_COFFEE' class.
2. Click on 'Edit', followed by the 'Generate' button.
3. Select whether you want to generate examples based on existing examples or based on the description, and how many examples you'd like to generate.
4. Click on 'Generate' when ready to generate.

The new examples are displayed in a preview window where you can review them first. If you are happy with the examples, you add them by clicking 'Apply'. Then they are added to the list of examples with a yellow highlighting line to distinguish them from previously existing examples.

![generate training examples](//videos.ctfassets.net/2vbtnveccz5s/7ra6DUV9g7PzDLFQIXlqmr/f6126b42288edb02d083f40cebd0c015/1.2d_generate_training_examples7.5_desktop.mp4)

### Test examples
Just like training examples, test examples can be generated either based on existing examples or a description. Let's try the latter approach this time.

1. While still in Edit mode, switch to the Test examples tab.
2. Click on the Generate button and select the option to generate based on a description.
3. Enter a description, such as `The user asks questions about which types of coffee you serve`.
4. Click on 'Generate' when ready to generate.

Just as before, the newly generated examples are displayed in the preview window where you can review them first. Click 'Apply' to add them to the list of examples. The new examples will be marked with a yellow highlighting.

![generate test examples](//videos.ctfassets.net/2vbtnveccz5s/11GQdZDkfXkp9iZZiftkaj/f8d56ba41a85f4d727d91a6f1cdf68b6/1.2d_generate_test_examples7.5_desktop.mp4)

Understand the User Intent: The Power of Machine Learning

Objectives

Next page