Hala Web Chat

Last Updated: 25.10.2019

Overview of the Chat

Hala chat is a communication Web Interface available in any modern browser. The end-users can use a chat interface for communication with Hala. All skills that you will build on the Hala Platform are available in Hala chat.

The chat is available in Hala Web Chat.

To login to the chat, you can use the same credentials as for Hala Platform.

Before staring the conversation with Hala you need to add the skills in Hala Platform.

Initial page

When you will the first-time login into the Hala chat, you will see the initial page, which includes the area for entering the text, quick-start buttons, and also a toolbar at the top of the chat.

1. Area for entering the text to start the conversation with Hala;

2. Quick start buttons. If you do not know how to start a conversation with Hala, you can use quick start buttons. By pressing the button, Hala will take this as an input text and start a dialog with you.

3. The toolbar includes the next buttons:

   1. Refresh - you can use this button to restart the dialog when you press it you will be prompted to the initial page.

   2. Microphone - by pressing this button you will be able to talk to Hala;

   3. Dynamic - by pressing this button you will get a response from Hala via voice;

Standard page

After entering the first message, you will be redirected from the initial page to the standard page that has simple chat functionality, where you can enter the text and get a response from the Hala.

Overview of the Chat test mode

Hala chat test mode was designed to help the developers to understand the behavior of the Hala backend logic, including the natural language processing and integration with the external applications.

The test mode is visible only for the users with the role - Developer. More information about the available roles here.

You can access the chat by using the next link: https://chat.hala.ai/

The test mode is the same as standard chat. Additionally, it includes detail information after processing each dialog node, which can help the developers to build the skills and dialog nodes in a better way.

The additional information include the next objects:

• Intent. This object includes information about the recognized Intent in the end-user input with the highest confidence. Hala will use this Intent to evaluate the conditions in the dialog nodes, and if the Intent is matched with node condition, then Hala will trigger the node that includes the matched condition.

• Entity. This object includes information about the recognized entities in the end-user input.

• Intents. This object includes information about the top five intents that were recognized in the user input. It can help you to understand the similarity between different Intents and improve them for better understanding (make the confidence higher) in the future.

• Context. The context includes information about the context variables that exist in your skill during the skill execution. The developer can create context variables. Additionally, any API response from the external application will be stored in the context variable

How it works

When you write a message to Hala, you will see an additional area (black background) that will describe each object. In other words, whenever you will write something, Hala will execute the internal logic and display the results below the message.

You can extend the "Show developer console" to see all the details.

Intent

For example, you have created the intent intent.get_weather.As a result, when you will write "current weather" Hala will determine the intent intent.get_weather, and in the test mode, you will see this information.

If the intent confidence is less than 0.2, Hala will not show any results in this section. It will mean that user input is irrelevant to the trained intents.

More information about the intents here.

Entity

For example, you have created the entity entity.city with few values: Tallinn and Riga. So when you write something like this, "Give me the weather in Tallinn," Hala will match word Tallinn with entity entity.city, because it includes the values Tallinn.

More information about the entities here.

Intents

In intents, you will see the top five intents (only with confidence more than 0.2) that recognized by Hala. For example, you have created two intents: intent.sap_block_user and intent.sap_unlcok_user. Those intents are very similar because the training phrases for them can differ only by the verb "Lock" and "Unlock". So here in this list, you will see the most similar intents with their confidence.

Context

The context section includes the context variables that you have created during the chat and also a response from the external API. For example, for the weather skill, at first, we need to store the information about the city name into the context variable context.city, and then we can make an API call to the external weather service to get the weather forecast for Tallinn.

As a result, the context context.city will look like this

And the API response will be available in the context variable context.weatherResponse

More information about the context here.