The Conversational Designer expands the old Article Editor capabilities so that, in addition to simple search-based articles, it supports building conversations with multiple questions and answers. Furthermore, the Providers usage and UI (called Entities from now on) are refactored, so that they function as a complementary part of the editor without the conversational business logic. The new Conversational Designer offers the following:
- Enhanced Editor to create composite conversations containing multiple bot questions.
- New Entity page (formerly known as Provider), which allows the customer to easily translate external data (for example, from CSV or JSON) to a conversation.
- Improved Conversational bot makes smoother conversations based on user queries. For example, the bot will not ask anything it already knows and can change the questions’ order.
The new Conversational Designer offers a simplified way of building rich conversations based on external data sources in two steps: link an external data source to the Bold360 ai Console and design a conversational business logic on top of it. The new process requires minimal to no technical expertise, depending on the complexity of the conversation.
To create a conversational article based on a CSV spreadsheet or a JSON file, you must do the following:
- Task 1: Create dynamic content for your conversational article in a spreadsheet or a JSON file
- Task 2: Create an entity in Bold360 ai
- Task 3: Create an article that references the entity
Task 1: Create dynamic content for your conversational article
If you have a CSV file
Decide what you want to provide as dynamic content and create a CSV sheet with the information.
- With a CSV editor, such as Google Sheets, create a new spreadsheet and name it, for example as CARS.
- Enter your data with one property per column. For example, TYPE, MODEL, COLOR, BRANCH.
Note: The column headers will serve as questions (which you can phrase later) and the data you enter in the spreadsheet will be displayed as potential answers to those questions.
- Publish the sheet to the web as CSV.
- In Google Sheets, go to File > Publish to the web.
- Make sure you publish as Comma-separated values (.csv).
- Copy the provided URL. You'll need it in Task Two.
If you have a JSON file
- Create a JSON file with the dynamic content that you want to display.
The following is a sample JSON file:
'name': 'Anna Banana',
'name': 'Mike Crueger',
'name': 'John McKornick',
- Upload your JSON file to an online storage, such as Google Drive.
- Publish the file to the web.
In Google Drive, right-click the file and select Get shareable link. This makes the file publicly available and copies the file share link to your clipboard.
- Paste the link to a text editor and copy the ID
- Create a new link by appending the ID to the end of the following string:
When you are done, you should get the following file link:
Task 2: Create an entity in Bold360 ai
- In the Bold360 ai Console, go to Knowledge > Entities.
- Enter a name and description for your entity.
- Paste the Data Source URL that you copied in Task 1.
- Select CSV for Response Type and click LOAD.
- In the When no entities found, bot will say field, enter a message that is presented to the customer when the person selects an option that does not retrieve any result.
- In the Properties Table, enter the lead-in questions to the Query fields. Customers will be asked these questions when selecting an answer. Again, you define the potential answers in the Excel sheet that you have created in Task 1.
Note: A property will be presented in the conversational answer even if you do not want to ask the customer a related question. In this case, do not add a question for the property. Typically, you would not ask a question related to a product's image, which is displayed in the conversational answer.
- Optionally, click the More options icon (...) next to a query to see additional settings:
- Split by - A syntax value such as comma or pipe that is used to separate multiple values on a certain field (such as when you want to use several synonyms or values that can appear in the visitor search that relate to the same key value)
- Auto Spelling - When selected, the visitor's question must contain an exact match to a value in the field
- Is Shared - You can share the selected property with another entity. For example, if you define COLOR in your Cars and in your Bikes entities, then you can share the color values across entities. In this case, when the customer selects a car color that does not exist in the Cars entity but exists in the Bikes entity (which is shared), then the message defined in the Missing Entity field is displayed.
- Autocomplete: Suggestions are displayed without words that appear more than once in a sentence.
- Use Quick Option - The customer can choose from options that a specific CSV column contain. For example, if the CSV spreadsheet contains Blue, Black, and Red in the COLOR column, then these options will be displayed for the customer to choose from.
- In the Default Fallbacks section, define how the answer should be displayed when there are multiple results to a customer's query.
- Save your changes.
Task 3: Create an article that references the entity
Create an article that serves as a generic template for the bot's response.
- In Bold360 ai, go to Knowledge > Articles > Add Article. A new article is created.
- Add a title to the article. For example "Where can I buy a car?"
- In the body of the article, follow these guidelines:
- Whenever you want to reference an entity, use this format: [[Entityname.ENTITYCOLUMN]]
- You can click the Entity button in the toolbar to select an entity from a list
- Use ALLCAPS, as shown
Your car type is [[Cars.TYPE]] and model [[Cars.MODEL]] with color [[Cars.COLOR]] are available at [[Cars.BRANCH]] the price is [[Cars.PRICE]]
Once you reference a property in the article body, the related entity is automatically "pulled" into the article and hence the conversation is created.
Note: All applicable entity properties (building blocks) are listed at the top of the Conversational Designer. The layout of the results when search retrieves multiple answers is also displayed for reference.
- Publish your article.
Once you created an article, customers will have the following experience in the conversational widget:
How does a conversational article "talk" to the customer?
The chat widget presents the questions of a conversational article in the most effective way so that the customer has to provide the least number of answers to find the result. This means that regardless of the order of the questions/headers in your CSV or JSON file, which also defines the order of questions when you set up the Entity, the bot rearranges the questions in the most logical way.
In addition, if a question in the conversational article has only one answer, then the bot simply skips that question and does not present that to the customer. For example, if a car dealership has only one black BMW, then the bot does not ask the customer either the price or the location of the car, but will display the result immediately.
How to use Quick options for a conversational article
For conversational articles, you can now define whether you want to use Quick Option buttons for a question directly from the Conversational Designer. When using quick buttons for a property (that is for a question), the widget displays results to the customer as a list of options to choose from. If you believe that your customers would see too many buttons as options, you can simply disable Quick Options while editing a conversational article, without editing the related entity.
Note: If you disable quick options for a property in the Entities editor, then you cannot enable it in the Conversational Designer.
How to view dependent Entities of a conversational article
On the Entities setup page, you can now see all dependent entities that are referenced from a selected entity. These dependent entities are parameters in the widget URL, such as www.company.com/getData?accountId=[Account.id]. In this case, Account is a dependent entity, from which the id property is referenced. By displaying dependent entities, you will have a better understanding of what properties compose the entire Entity.
Troubleshooting common problems
If your entity doesn't work as expected, check these common setup errors:
- When using a CSV file, open your entity sheet and make sure you published it as CSV (Task 1, step 4)
- Open your entity in Bold360 ai and make sure you set proper the Response Type (Task 2)
- Open your article and make sure you reference the entity name and fields (columns in your sheet) properly