With Conversational Language Understanding (CLU), you can train a model in one language and use it to predict intents and entities from utterances in another language. This powerful feature can save a huge amount of time and effort; instead of building separate projects for every language, you can handle multi-lingual datasets in one project. Your dataset doesn't have to be entirely in the same language but you should enable the multi-lingual option for your project in the project settings.

![teneoxclu](//images.ctfassets.net/2vbtnveccz5s/38FpMOQf0ILhxQaVMpjjco/cadba5b9570ab295a8a01e7e327989ee/teneoxclu.png?classes=w60)

Localizing a bot for different languages, markets, and/or cultures is a common use case. This is accomplished in Teneo through branching, where we have the option of creating a 'Master' version of a solution and its own 'Local' versions for each of the different languages we are interested in. In this way, Teneo allows us to create local solutions with all of the parts of the master solution that we want to include, making it an efficient way to create bots for different languages and markets. You can read more about localization [here](/article/build/language-understanding/concepts/localization).

In this guide, we will demonstrate how to localize the 'Longberry Baristas' solution, which is in English, to other languages with the help of CLU. More information on branching can be found in [Documentation](/documentation/latest/user-guide/solution/localization-setup).

By following the steps in this guide, you will complete the following:
1. Train a multilingual CLU model on English from within Teneo.
2. Create a localized Teneo solution in another language that uses the same CLU model.

Multilingual CLU - Introduction

## Log in to Teneo Studio

First, you need to log in to Studio. Studio will open in the tab 'My Work'. This gives you quick access to the things you’re working on. As you have not yet created anything, nothing will be displayed under this tab.

![login-teneo-73](//videos.ctfassets.net/2vbtnveccz5s/1owoWS2GH45skxYhGxSHXJ/fb9c17c4789e20bfea149a4ac62d4442/login-teneo-73.mp4)

Multilingual CLU - First steps

## Connect your CLU project to Teneo
We will now adjust the CLU settings in Teneo to allow access to CLU. This way, your solution will be able to use the CLU native intent classifier in multilingual projects instead of Teneo's native classifier and classes in the [Class Manager](/getting-started/class-manager).

1.  Navigate to the 'Account' section in the sidebar. Here you can see information about and settings regarding your CLU setup.
2.  Click on 'Edit' and enter your resource name and API key.
3.  Hit 'Save'.

![connect-clu-73](//videos.ctfassets.net/2vbtnveccz5s/6I0WeKP1bGMUW2gAIRlF8Y/f233af21bdf3fe9c88047bf14653e16e/connect-clu-73.mp4)

Multilingual CLU - Connect CLU to Teneo

## Open your solution
Now with CLU connected to our Studio account, it is time to test it out on one of our pre-built solutions, Longberry Baristas. 

1. Click on the 'Solutions' link in the sidebar to get into the solutions view.

In this view, you can see all the solutions available to you in this environment. A solution can be thought of as a container where all the bot's knowledge is stored.

As you can see, there's already a solution called 'Longberry Baristas' available to you. Each team gets this pre-built solution which can act as a reference solution when you're building your own bot. Let us take a look inside:

2. Select the solution. On the right, you can find meta-information about the solution, such as when it was built, when it was last updated, and when it was last published.
3. Click on 'Open' above the solutions field.

## Your first conversation
Being able to have a conversation with your bot is a crucial part of the development process to ensure that everything is working as intended. Luckily, you don't have to go far to test your bot in Teneo; let's use 'Tryout' to order a coffee! 

1.  To the right in the explorer window, you'll find Tryout.
2.  Chat away!  
Perhaps you could try a conversation like this:
*   `Hi!`
*   `I would like to order an espresso`
*   `John`
*   `Goodbye`

![talk-to-tryout](//videos.ctfassets.net/2vbtnveccz5s/31bI61vfOtBSsrqJ5y9Aqb/210392d7a7436159385279bdc96a14d3/talk-to-tryout.mp4)

At this stage, Tryout uses its own classifier, Learn, to detect intents. Next, let's explore how to use a CLU model instead.

Multilingual CLU - Your first conversation

At this point, your bot is already working, but it is now using Teneo's native classifier, Learn. Now that you have entered your CLU resource name and API key, you also have the option to train a CLU model and have the engine use that for classification instead.

## Open the solution
We will start by preparing the solution we want to localize. In this case, it is the 'Longberry Baristas' solution that's included in the environment you can request access at Teneo Developers by [requesting a Sandbox](/signup).

1. Go to the 'Solutions' section and select 'Longberry Baristas'.
2. Click on 'Open' above the Solutions field.

## Train the CLU model
The CLU model will be trained on the same training data as the native classifier. Our Longberry Baristas solution already has all of this training data ready (as you can see in the Class Manager), so let's get right into training the model.

1. Open the CLU Manager in the ribbon bar.
2. Click on 'Train'.
3. Wait for the model to finish training. This can take a while, but you do not need to keep this window open until it is done; if you want, you can close it and come back later.
4. Once the model is finished training, click on 'Assign' in the ribbon bar of the CLU Manager window. This will assign the model to the solution, ensuring that the engine will use this model for classification instead of Teneo's native classifier.

![train-clu-new](//videos.ctfassets.net/2vbtnveccz5s/4zoxnMi38XyCwVqAnO3ex4/315b79c07629b43b0bb5f278102c508a/train-clu-new.mp4)

Multilingual CLU - Train your CLU model

## Try it Out

Let us test how well our solution behaves in a multilingual enviroment, from Tryout! For this we will try to have a conversation with the bot using, English, German, Swedish, Turkish and Spanish in the same session. Let's start!

* Spanish - `Hola!` (Hello)
* English - `My name is John`
* Swedish - `Minns du vad jag heter?` (Do you remember my name?)
* Turkish - `İyi misin?` (Are you well?)
* German - `Auf wiedersehen` (Goodbye)

![localization](//videos.ctfassets.net/2vbtnveccz5s/3QxEKsYcnNVtt80nyzTupM/76d4bd40c3250dda07d5c1fa63d1037e/localization.mp4)

!!! See [Microsoft documentation](https://docs.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/language-support#multi-lingual-option) for more information on the CLU multilingual option.

Multilingual CLU - Try it out

## Include relevant content for branching
It is time to include the relevant content from Longberry Baristas for branching.

### Flows and Entities
We will start off with including all of the flows and entities. In Teneo, all dialogs are designed as flows, which contain the logic needed to resolve a specific question of the user.

Language objects are building blocks for language conditions. They capture words, synonyms, or various ways of expressing the same (partial) intent. They make it possible to efficiently write and re-use conditions. Entities are a special kind of language object consisting of one or more entries, which denote what the entity is matching and optionally also which features are assigned to this entry.

1. Start off by selecting the 'Flows' and 'Language Objects' folder.
2. Click on 'Include' in the 'Branching' section in the ribbon bar. The buttons in the 'Branching' section look like this:

![include content for branching](//images.ctfassets.net/2vbtnveccz5s/5OeSZw77gI9q59A7sMzXXx/b5807f68403fddc45144ea34c51855aa/include_content_for_branching.png)

A more detailed explanation of these steps can be found [here](https://documentation.artificial-solutions.com/latest/#/user-guide/solution/localization-setup).

### Classes
These flows use Class Match, which also means that they depend on Classes. Classes are built on training data and are used in the machine learning model of your solution. They are stored in the [Class Manager](/getting-started/class-manager).

1. Open the 'Class Manager' by selecting the 'Class Manager' button located in the 'Solution' section.
2. Select all the classes that do not have a 'TDR_' prefix. For example, 'ANOTHER_ONE_PLEASE', 'ORDER_A_COFFEE' and 'TELL_ME_ABOUT_YOUR_COFFEE'.
3. Finally, click on the 'Include' button in the 'Branching' section.

### Variables and listeners
Global variables and listeners are important to be included in branching. Global variables are used to store data and can be accessed from anywhere in the solution. Listeners help your bot pick up relevant pieces of information from user inputs, even when these have not been explicitly prompted for.

1. Click on the 'Solution' tab.
2. Click on 'Globals'.
3. Under 'Variables', select all of the listed Global Variables and click on the 'Include' button.
5. Under 'Listeners', select all of the listed Listeners and click on the 'Include' button.

### Integrations
Finally, make sure to include relevant integrations. Integrations are a reusable set of scripts that can be written once and used many times within flows throughout the solution. Most commonly, they are used to connect to external services.

1. In the solution backstage, click on 'Resources', followed by 'Integration'.
2. Select all listed integrations and click on the 'Include' button.
3. Finally, click on the 'Save All' button located on the sidebar in the solution backstage. This will make sure everything you selected to be included for branching inside the solution will be stored and saved.

![include content for branching](//videos.ctfassets.net/2vbtnveccz5s/5BbIVkFl0h9iDEUJQR7OBv/f4f4cfd7f73e284949ec138bd82d72ac/include_content_for_branching.mp4)

### Forget something?

Don't worry! It is possible to update your 'Master' solution if you forgot to include something before branching. Simply go back to the 'Master' __Longberry Baristas__ solution, and include it. Then return to your 'Local' solution, go to the Solution backstage, and click on the 'My Work' section. We can now see the changes from the 'Master' solution, where we can either select __Update All items__ to update everything or __Update Selected items__ to only update relevant parts. Once updated, the local solution will pull in the changes from the Master solution.

Multilingual CLU - Include content for branching

## Localizing Longberry Baristas
Now that we have prepared our solution, let’s branch our English Longberry Baristas bot to a different language. This process is fairly similar regardless of language; we will be using German for this example.

!!! While this guide is written with a German translation, feel free to pick another language and translate as necessary for the following steps.

1.  While in the Solution window, select the 'Longberry Baristas' solution.
2.  Click on 'Branch' above the solutions field.
3.  Make sure the 'Solution Language' for this first bot is German. This will be the mother tongue of your bot.
4.  Under 'Solution Content', click on 'Master and Template' and ensure 'Teneo Dialogue Resources German' is selected in the dropdown menu.
5.  Fill in a name for your solution, for example `Longberry Baristas - German (Deutsch)`.

!!! For languages without dialogue resources leave the __Master only__ selected under 'Solution Content'.

6.  Finally, click 'OK' and wait for the branching to complete.

![create german local soln](//videos.ctfassets.net/2vbtnveccz5s/2fQu7992UFDKCsCnFYCpBW/bd894aa04e0f3668dabd8809e1002b13/create_german_local_soln.mp4)

Multilingual CLU - Localizing Longberry Baristas

## Identify warnings and steps to take

Warnings in Tryout appear as yellow warning triangles. To see more detailed warning messages, open the Tryout window and navigate to the Messages section. More information on warnings can be found [here](/article/optimize/refine-your-bot/troubleshooting/warnings-and-errors-in-tryout).

![identify warnings](//images.ctfassets.net/2vbtnveccz5s/7Kjsxh2cUKVLpVI4ngX5mM/1cc192a85ebb83e8d1a061f847183474/identify_warnings.png)

## Localize TLML Syntax
At this point, you will be seeing errors like the ones in the image above because the Longberry Baristas flows from our Master solution contain references in their TLML Syntax to the English TLR and TDR. As the local solution is in German and contains the German TLR and TDR, these references are causing issues. To fix this, let's update flows with the correct references.

1. Open the 'User asks about coffees we offer' flow.
2. Select the trigger and scroll down to the Match section in the configuration panel.
3. Replace '%COFFEE.NN.SYN' with `%KAFFEE.NN.LEX`.
4. Replace '%DRINK.NN.LEX' with `%GETRÄNK.NN.LEX`.
5. Save the flow.

![localize tlml syntax](//videos.ctfassets.net/2vbtnveccz5s/558gg1IvcaSBJQ0NJamWnT/1eecf5ebd896ea84a55d90a7dcfdead2/localize_tlml_syntax.mp4)

Continue to localize TLML Syntax in the remaining flows taken from the Master solution. Make sure to keep an eye on the errors and warnings in Advanced Tryout to understand what issues still need solving.

## Import files
For the Longberry Baristas solution that we are using as an example to work, you will need to add a file to your new localized solution.

1. While inside the solution, click on Solution in the upper left corner.
2. Select Resources in the menu, followed by File.
3. Click on Add and add the file [business_hours.csv](//assets.ctfassets.net/2vbtnveccz5s/6cyowN6YR4bcHTMfHADeR6/4313d357c0613dc4c8d5c2d337761b66/business_hours.csv).
4. Set the published location to `/script_lib`.
5. Click on Save all to save the new files.

Multilingual CLU - Identify warnings

## Add local training data
At this point, we could train the CLU model and use it as a multilingual classifier. However, let's boost the model's performance even more by adding some training data in German for our German solution.

1. Open the Class Manager.
2. Select the 'TELL_ME_ABOUT_YOUR_COFFEE' class and click on 'Edit'.
3. Under the existing training examples for this class, paste the following ones in German:
```example-inputs
Können Sie mir über eure Kaffeesorten erzählen
Verkauft ihr espresso
Welche Getränke haben Sie
```
4. Hit 'Save'.

![add german training examples](//videos.ctfassets.net/2vbtnveccz5s/5vBLzZQ7I7xHR0IQF1fUrb/e46c8c1c56806b593c5135e419d68b64/add_german_training_examples.mp4)

Go ahead and do the same for the rest of the classes. You can add as few or as many training examples in the local language as you wish.

Multilingual CLU - Add local training data

## Retrain the model
As the classes in our localized solution are not identical to those in the Master solution, we will need to train a new CLU model.

1. Open the CLU Manager.
2. Click on 'Train'.
3. Once the model is done training, select it and click on 'Assign' to assign it for use in the solution.

## Try it Out
The bot will now use the newly trained CLU model to handle conversation in German. Let's give it a try by triggering the 'User wants to order a coffee' flow.

1.  Open the 'Tryout' window located in the 'Solution' section.
2.  Test the following conversation:
* `Servieren Sie Kaffee?`
* `Ein Espresso bitte`
* `John`
* `Danke`

![test with english replies](//videos.ctfassets.net/2vbtnveccz5s/E1MKP6o4sKX18OALmzPMN/7e2c43c60fe9b1ae6fd671e06999a9eb/test_with_english_replies.mp4)

The bot can now handle German inputs easily, but the Longberry Baristas flows taken from our Master solution are still in their original English. Let's try localizing some outputs in the next step to change this.

Multilingual CLU - Try out the localized bot

## Localize outputs
Let's try localizing some flow outputs to get the answers we want in German.
1. Open the 'User asks about coffees we offer' flow.
2. Replace the answer in the 'We offer a variety of handcrafted coffee drinks...' node with `Wir haben eine große Auswahl an handgefertigten Kaffeesorten, von Espresso bis Macchiato, Flat whites, und viele mehr.`
3. Open the 'User wants to order a coffee' flow.
4. Replace the answer in the 'Ask for name' node with `Ok, und wie war Ihre Name?`
5. Replace the answer in the 'Summarize order' node with `Vielen Dank für Ihre Bestelling, ${userNameForOrder}. Ihr Getränk ist in fünf Minuten fertig.`

![localize outputs](//videos.ctfassets.net/2vbtnveccz5s/1ft95op5DgTvJHV3MsMWAb/386a77ba0951cc0818f46cfb20d2896e/localize_outputs.mp4)

The example conversation from earlier should now have German responses.

![try out with german outputs](//videos.ctfassets.net/2vbtnveccz5s/kryBCPcDwB5DWFc9FsMai/c5fe2917717152511d6018781c1afddf/try_out_with_german_outputs.mp4)

Repeat this process for the remaining Longberry Baristas flows to get German answers for all use cases.

Multilingual CLU - Localize outputs

Localize with multilingual CLU

Localize your Teneo bot with multilingual Conversational Language Understanding.

Localize with multilingual CLU

Retrain the model

Try it Out

Next page