Sinch Community

Contextual entities are used to detect entities in user expressions by using the context of the sentence of that expression.   For example, in the expression: I want to book a train from Brussels to Amsterdam. The two city names mentioned here will be recognized as either @departure or @destination because of the context of the expression they're in.   When to use contextual entities Imagine you're a fruit seller who sells apples, pears, bananas and pineapples. Your chatbot can help people order fruit and ask for the price. Let's look at that pricing use case for apples. Some example expressions could be: How much does an apple cost? What is the price of an apple? Of course, these expressions could be asked for not just apples, but also for other fruits. Does that mean you have to copy each expression and replace the fruit? No, you can just use contextual entities! Create a contextual entity called 'fruit' and add the following values to it: apple banana pear pineapple Now you can simply create an expression with the contextual entity in it: How much does an @fruit cost? What is the price of the @fruit? Now your bot will understand each expression, even if the fruit changes!   When adding contextual entity variables, there's no need to add capitalised and non-capitalised words. All text is decapitalized by the NLP before any recognition happens, so 'api' won't be recognised differently from 'API'   Fuzzy matching Fuzzy matching allows you to recognise a slight variation of an contextual entity value (or its synonym) as the original value.   For example Brusselt will be automatically corrected to Brussels if fuzzy matching is enabled.   Create a contextual entity Let's see how we can use a contextual entity to determine the type of pizza that your customer is ordering.   To create a contextual entity: Go to the Entities page and click on Contextual Entities. To add new entity, click Add contextual entity. Give a name to your entity. When naming your entity, please do not use spaces. Fill the entity values with pizza types and you can also add synonyms for each value so the bot can understand if your customers say it in a different way.     4. Click Create entity.   Add contextuel entities to intents To add contextual entities in intents training expressions: Add a new intent to your bot. For this example, call it ‘Order pizza’. For this example let's assume that you already created a @pizza_type contextual entity. Add expressions with this entity you just created by typing '@' and a dropdown will appear, where you can find your created entity '@pizza_type'.     3. To improve the bot's understanding of the "order pizza" intent, you need to add or generate additional expressions using the @pizza entity. The more expressions you add, the better the bot will perform.      4. After adding enough expressions, do not forget to train your bot so it can learn the latest changes that you have just made.     4. Connect your intent to your flow. 5. Test your bot.     See also: [New bot builder] Detect entities in expressions [New bot builder] Recognize expressions with intents [New bot builder] Add an intent to your bot   ... View more

Adding intents to your bot allows your customers to be understood in natural language, without the need to click a button.    🎥You prefer to watch a quick hands-on tutorial? Check out our new bot builder video tutorial!   Create an intent  To add an intent to your chatbot: 1. Go to Intents page from the main navigation bar.     2. Click the + button. 3. Summarize your intent with a brief description. This step is optional.       4. Add expressions. This can be done in 3 ways: Either you write them in the input field and press 'enter' on your keyboard. Or you can click the + button on the left to add them. Or by generating expressions powered by LLM (Large Language Models). If you don’t have any expressions for your intent yet, the name and description of the intent will be used as the prompt to generate the expressions. Select the expressions that fit your intent and add them to your intent.     5. Train your bot after making changes to your intent.   Connect your intent to a flow   As mentioned before, an Intent can be used to trigger a specific flow.   In the previous example, the bot was already trained so that recognizes questions about opening hours. But it isn't connected to any answer flow yet.   To connect an intent to a flow: Go to Flows. Drag an Intent block onto the canvas, Open this block and search for your created intent. cClick Save.      5. Add Message block that will be sent after intent is triggered. 6. Select Add step and then Text.      7. Write the message then click Save.     8. Test your changes by clicking the play button.         See also: [New bot builder] Train your AI bot [New bot builder] Recognize expressions with intents [New bot builder] Blocks and steps ... View more

An intent is the overall goal or purpose of the user's expression. For example, if someone says "I want to order a pizza", the intent might be "Order pizza". Intents help the bot know what kind of conversation flow to trigger, like asking for menu preferences or payment details.   Expressions are example sentences or words that mean a specific intent: they're all the different ways a user can express their intention. It's crucial for an intent to contain diverse expressions so that the NLP can give more accurate results after training. You can set up your bot so that it detects entities inside of expressions.     As shown in the example below, the bot has already been trained with the intent to "order pizza". As you can see, the intent "Order pizza" has different expressions that convey the same goal, which is to order pizza. The entity mentioned in the example above is "pizza type".     Intents and entities are important for building an AI bot. By using intents and entities together, the bot can provide a more personnalized and interactive experience.   See also:  [New bot builder] Add an intent to your bot [New bot builder] Train your AI bot [New bot builder] Detect entities in expressions ... View more

As your chatbot cannot answer just any question, being able to transfer your customer to an agent is a key component of good conversation design. On Engage, implement this by using the Transfer to agent step.   To add a Transfer to agent step: From your block, add a Transfer to agent step.      2. From the Agent skill options, select a skill to ensure the conversation is assigned to an agent with the right competence. For this example, we have defined a language English skill.       3. Configure destinations when transfer to agent was: successful: when it connects to an open offloading. failed: when it connects to the offloading being closed. disabled: when it connects to the offloading being deactivated.   For optimal user experience, consider providing a message upon successful agent transfer, such as 'An agent will assist you shortly.' In the event of a failed transfer, communicate the reason for agent unavailability and suggest alternative contact times. Likewise, if agent transfers are deactivated, inform customers why agents may not be available and offer alternative means to get in touch.       4. Click Save.   See also: Conversation design intro [New bot builder] Manage intent Using skills     ... View more

Sometimes it proves useful create a slight delay between bot messages, to create a natural pause in the conversation. You can do this by using a Delay step that lasts between 1 to 20 seconds.   To add a Delay to your conversation: From your block, add a Delay Action step.     2. Set the Delay duration to your preference by moving your cursor.     3. Save your changes.   See also: [New bot builder] Blocks and steps [New bot builder] How to organize your flow [New bot builder] Clear session ... View more

Clear session allows you to delete variables and reset the conversation. This is helpful when you want to start a new conversation or remove any stored information from the current session.   For example, a train booking bot could end by saving a train ticket in the ticket ordering system. Afterwards, users should be able to book a new ticket to a different location. You can achieve this with the 'clear session' action, which removes the values of set session variables. This is useful when a user asks to correct a value or wants to start over and delete all variables.    The main difference between "close conversation" and "clear session" in a bot builder is as follows:   🚨 Please avoid using Close Conversation as it's not functional. 🚨     Clear session:  allows you to reset the conversation within the same session. It deletes variables and stored information from the current session, but keeps the chat session open for further interaction. This is useful when you want to start a new conversation or remove any stored information, while still allowing the user to continue chatting. To clear your session: Open your block. Add an Action step, choose Clear session. Under Clear variables, choose the variables that you would like to erase. Save your changes.     See also:  [New chatbot builder] Save a variable under button click [New chatbot builder] Add Collect input to your flow How long does a WhatsApp session last? ... View more

A Condition block allows to steer the flow based "if, then" logic statements. The Condition block checks variables and guides the user in different places of the conversation based on it.   🎥You prefer to watch a quick hands-on tutorial? Check out our new bot builder video tutorial!   To add a Condition block to your flow: Drag and drop a Condition block from the side panel onto the canvas. Click the + button. In the previous flow, we used a Collect input block to ask the customer how many pizzas they want to order, and we saved their response in the variable 'order_amount'. In this case, choose the variable 'order_amount' under the If condition. This selected variable will represent the amount of pizzas ordered. If a customer orders more than five pizzas, they will be automatically redirected to a special conversation where they can claim a complimentary pizza.    4. Next, select greater than and enter 5 in the next input field. This is because we will be offering free pizza for orders greater than 5.     5. If the user's response meets this statement, the conversation will go to a special block called 'Claim Free Pizza'. To make this happen, you can set up a Go To. You can either pick an existing bot dialog or make a new one. Since we don't have a block named 'Claim One Free Pizza' yet, we'll make a new one.      5. The Else statement is executed when none of the other paths are applicable. You have the flexibility to decide what action should be taken when none of the conditions are met. In this case, if the customer's orders are not more than five, the Else section will redirect them to the purchase flow.     6. Next, click on Create and add a message to the 'Claim free pizza' block and the 'purchase flow'.     7. Test the result to see how your changes perform.      See also:  [New bot builder] Save a variable based on button click [New bot builder] Collect input from your users [New bot builder] Make a channel router flow ... View more

In your Engage chatbot, you can ask for input of your customers and by adding Collect input blocks to your conversation.    🎥You prefer to watch a quick hands-on tutorial? Check out our new bot builder video tutorial!   To add a Collect input block: Let's say we created a match entity for a customer ID, called @customer_ID. To add Collect input, you can either: drag and drop a Collect input block from the left panel or start by dragging and dropping the connection arrow, then select Collect input The Collect input block opens. Click on Add step. Select Text.     5.  Select the created entity @customer_ID under Check if response matches dropdown.     6. Provide the name of the destination variable, which is where the value will be saved.     7. Create a new block that will be triggered when the user's input is validated. In this example, we will guide them to the next block where the bot will provide selectable options.     8. Next, create a text message in case the customer's response does not match. The bot will ask for the customer_id, which should consist of 6 numbers.   9. Save your changes and test your bot by clicking the Play button. Make sure that you have already trained your bot before testing it. Try it out: if a customer makes a mistake, the bot will redirect them to the correct answer until they provide the right customer ID.       See also: [New bot builder] Validate a user input [New bot builder] Detect entities in expressions [New bot builder] Variables and properties   ... View more

Collect input blocks are used to gather input from the user. They check if the user input matches a known variable. There are 2 possibilities: If the variable already has a value, the Collect input block is skipped and the bot proceeds to the next block. If not, the Collect input block is displayed and the input is asked.   In the Add step section of a Collect input block, you will find various options such as Message, Buttons, Carousel, WhatsApp list, and Media.     Collect input without validating the response   If you choose not to validate the answer, the information collected through the Collect input block will be considered correct no matter what the customer responds.   To skip validation, you need not to select any from the Check if response matches dropdown. In the Destination variable field, enter the variable name (e.g. name).   Collect input with validating the response    Collect input has two types of input recognition: general input types entity input types General input type The general input types are used to check if the user's input follows a desired format.   Date The Date input parser type will try to parse the response as a date. Sentences like 'next week Monday' are automatically converted to a DD-MM-YYYY date object. Supported formats (also in other supported NLP languages) are: 22-04-2018 22-04 22 apr 22 april 18 twenty two April 2018 yesterday today now last night tomorrow, tmr in two weeks in 3 days next Monday next week Friday last/past Monday last/past week within/in 5/five days Friday/Fri   Location The location parser will send the user's input to a Google Geocoding API service. When a correct address or location is recognized, the bot automatically create an object that contains all relevant geo-data.       To show the address as a full address (street, street number, zip code and city) you need to add some extra information to the variable: .fullAddress. So in the example above, the bot can display the entire location by using the following variable:{userLocationInformed.fullAddress}.   A bot message containing the following info: Thank you, shall I send your package to {userLocationInformed.fullAddress}?   Will display the following message to the user: Thank you, shall I send your package to Oudeleeuwenrui 39, 2000 Antwerpen, Belgium?     Hours This input type will parse and validate timestamps.   Image The image format type allows you to check if a user has uploaded an image or other file (such as pdf). Currently, this is only possible in the Facebook Messenger, WhatsApp, Instagram, RCS, MMS, Telegram and Instagram channels. If the bot user uploads a file, the URL to that file will be saved under the variable you save the response to.   If you want to save the attachment from a user no matter where in the flow the user is, you can use the defaultOnFileUpload variable. The URL that links to the attachment will be saved in this variable if the user uploads an attachment at a random point in the conversation.   Language This input type will parse and validate NLP supported languages.    English: (en-us): 'engels', 'English', 'en', 'anglais' Dutch (nl-nl): 'nederlands', 'Dutch', 'ned', 'nl', 'vlaams', 'hollands', 'be', 'ned', 'néerlandais', 'belgisch' French (fr-fr): 'French', 'français', 'frans', 'fr', 'francais' Chinese (zh-cn): 'Chinese', 'cn', 'zh', 'chinees' Spanish (es-es): 'Spanish', 'español', 'es', 'spaans' Italian (it-it): 'Italian', 'italiaans', 'italiano', 'it German (de-de): 'German', 'duits', 'de', 'deutsch Japanese (ja-jp): 'Japanese', 'japans', 'jp', '日本の Brazil Portugese (pt-br): 'Brazil Portugese', 'Portugese', 'portugees', 'braziliaans portugees', 'português'   Entity input types  You can also check if a given user input is identical to your entities. When you choose an entity type in the dropdown Check if response matches.     When response matches When user response matches, you can select or create a block to lead end-user to the next block.        When responses does not match   When a user gives a response that is not matched (for example, when the bot asks for an email address and the user replies with 'chicken'), the bot needs to let the user know their answer was invalid.   To do so, you can create a message to be displayed for when input validation fails once, and another message for when the input fails 3 times. Additionally, enable no response detection to handle cases when users provide no input. By doing so, the bot ensures a smoother interaction.   First two times and after two times When the user fails to give valid input 3 times in a row, you have the flexibility to redirect them to a block. For instance, you can guide customers with an error message like "Sorry, I can't seem to understand. Please contact our support team at [tel number]". Alternatively, the bot can also offer to transfer the user to a live agent for further assistance.       No response   With the introduction of no-response detection, your bot gains the capability to handle cases where users don't respond. It allows you to configure which block is triggered if the bot detects a silence, or in other words: when the user does not respond for a certain period or until a specific time.   You can set the amount of time after which the new block will be triggered in the duration field (in minutes or exact time). The duration of silence can range from 1 minute to a maximum of 1440 minutes (24 hours).       Configuration   Disable NLP Users are able to leave the Collect input block if an intent is recognized. For bots with a very small NLP model, this might trigger a false positive. The 'disable NLP' checkbox allows you to disable the NLP model while in the input validation, which makes sure that whatever the user says gets saved as input.       See also: [New bot builder] Add Collect input in your flow [New bot builder] Save a variable under button click [New bot builder] Condition   ... View more

The WhatsApp channel offers multiple interactive messages, including List messages. List messages are a menu of up to 10 options from which users can make a selection. On Sinch Engage, you can add ‘WhatsApp List template’ to your chatbot.   Here is how the list message looks like on WhatsApp channel:     To add a list message to your chatbot: From your block, add a WhatsApp List step. As a reference, the image below demonstrates how the fields should be filled out to appear correctly on WhatsApp:         2. It is essential to incorporate a body message and label the menu button. Including a header and footer is optional. In this example, our bot assists customers in choosing an ice cream flavor to order. We will add a message asking the end-user which ice cream they would like to choose. Additionally, we will name the button 'Open menu'.       3. Modify the button title and configure the Go to block that will be triggered when the option is selected. In this case, we will add 6 ice cream flavours: strawberry, lemon, orange, chocolate, vanilla, and cookies. We will name the button according to each ice cream flavour and configure a Go to for each flavour.       4. Click Save to save changes that you just make. You’re going to see all the connection arrows between the button and the configured block that will be triggered when the button is selected.     5. Add text message for each block connected to a button.     6. Now it's time to test your bot. You won't be able to test WhatsApp lists on the test emulator, but you can use the WhatsApp sandbox for testing:   a. Click the 'play' button to open the test emulator. b. In the selector, click on 'WhatsApp Sandbox'. c. Follow the instructions to set up your WhatsApp sandbox.       See also: Create a channel configuration Channel limitations [New bot builder] Make a channel router flow   ... View more

Videos, images, and gifs are a big part of how people chat online. Therefore, Media steps give great added value to your conversation design if you want to polish your bot's personality, mood, and naturalness.    With the Media step, your bot can respond media components of different types: Images: all typical image types, such as jpg, png and gif are supported on our platform. Video: the following video formats are supported: mp4, ogv, webm. Audio: currently we only support MP3 as an audio format. Files: currently only PDF is supported.   To add a Media to your bot: Add a Media step to your block.      3. Click on Choose file and locate the file on your workstation. Alternatively, you can also provide the URL to your file.      4. Select the media type. In this case, you select Image.     5. Click Save to save the change that you just make and name the block to Media.     6. Test the bot by clicking the play button to see the result.      See also: [New bot builder] Add buttons to your flow [New bot builder] Add carousel to your Facebook bot [New bot builder] Add List messages to your WhatsApp bot       ... View more

Buttons are handy for suggesting several options to your customers. The Buttons step consists in a message supplemented with clickable buttons.   🎥You prefer to watch a quick hands-on tutorial? Check out our new bot builder video tutorial!   There are 3 types of buttons: Go-to: button that leads to another block on the canvas. URL: button that is given a URL link and opens it when clicked. Call: button that will initiate a call. Buttons are restricted depending on the channel that you are using. Therefore, make sure you know about channel limitations on buttons before setting them up.   To add buttons to your chatbot: Inside your block, click on Add step.  Choose Buttons.     3. Write a text message.     4. In this example, we'll include two buttons: one Go-to button and one URL button. First, add the Go-to button.     5. Under the Title field, provide a name for the button. In this example, we will use "Pricing" as the name, as this button will lead to a block that explains the pricing at the gym.      6. Under the Go-to section, we will create a new block. This block will be triggered when users click the button, redirecting them to a specific block. You can create this next block directly within the text field, begin typing the name and then select the desired block type.     7. Follow the same steps to add a URL button. Name the button "Opening hours" and enter the link (https://www.sinch.com/) that the button should open.     8. Click Save.     9. Add a bot message when the customer clicks on the first message (called "Pricing" in this example): Open the ‘Pricing’ block that you already created by clicking on it. In the Text message field, enter the text for the bot’s response. In this example, prices are given to the customer. Do not forget to save your block to save the changes.     10. It's time to test your bot! Click the Play button and see the result.     See also: Channel limitations [New bot builder] Blocks and steps [New bot builder] Save variables under button click ... View more

You can save variables under button clicks if you're using a button step.   To save a variable under a button click: Add buttons to your flow. Click on Add a variable button to add a variable when a user clicks on this button within a conversation.  In this example, we will name the variable "pizza", under Variable, and assign it the Value "margherita".    4. Add another button and create a variable called "pizza" with the value of the pizza type.      5. Save your changes by clicking Save.     6. Open the following block in your builder. In this case, we will edit the 'margherita chosen' block. Add a step and select Text. Then, re-use your variable by typing '{' or clicking the ‘{}’ button on the bottom left of the Text Message input field and selecting the created variable 'pizza'.      6. After finishing editing your message, you can click Save.     7. Test your bot. Your bot now remembers the customer's choice under a variable and can use it again!         See also: [New bot builder] Collect input [New bot builder] Add buttons to your bot [New bot builder] Variables and properties   ... View more

For bots that live on Facebook, carousels are a way to visualize more than 3 options by using images. See below how the carousel looks like on a Facebook channel: Each carousel option can have up to 3 buttons with separate actions, but this is not required. These buttons are the same buttons as in the button template and use the same properties like payloads and URL, with the addition of an extra share button.    To add a carousel to your Facebook bot: Add a text message to your bot. In this example, we will use a carousel to display the pizza menu with an image. Click Add step.  Select Carousel. Add an image to your carousel by either: dragging and dropping your file. or by first clicking Choose file and then selecting your file.     5. Name your option under Title. In this example, the title would be 'Margherita'.     6. It is possible to include up to 3 buttons inside a card. In this example, we add a button called 'Add to order', which initiates the pizza ordering process when clicked. Under Go to, you can create a new block called 'Order pizza' by start typing the title and then click to create.      7. Follow the previous steps and add more cards for each pizza type. 8. Click Save to save your changes.    See also: Channel limitations Create a channel configuration [New bot builder] Add buttons to your flow ... View more

In Engage, you can show the customer a piece of text in the chatbot by adding a Text step.   🎥You prefer to watch a quick hands-on tutorial? Check out our new bot builder video tutorial!   To add a text message to your bot: Drag and drop any block to the canvas. Under Add steps, select Message.      3. In the text field, enter the text you want the customer to see.     4. Click Save to save the changes.   See also: [New bot builder] Canvas overview [New bot builder] Blocks and steps [New bot builder] Add buttons to your flow   ... View more