Your chatbots are designed to respond to customer queries, offering accurate and relevant responses. This is possible because your bots are trained on their learning sources (such as FAQs and Q&As). While your bots can intelligently understand the context of the customer and the query for a more relevant response, you can also choose to improve the relevancy of your bot responses by mapping the sources (such as FAQs) to the customer queries (such as their preferences, choices, etc.). Your bot will filter the search results using these keywords while finding articles and sharing answers with the customer.
Note: While your bots fetch the most relevant information from across your knowledge base, filtering this search will allow your bots to search from within a specific subset of your sources.
TABLE OF CONTENTS
Who is this for?
Narrowing down the learning sources is particularly important for businesses handling multiple brands, verticals, regions, or any diverse customer groups serviced by bots trained on the same learning sources. For example:
Acme Flights has customers from multiple countries and thus has to accommodate different immigration procedures based on the customer’s country of origin (static user properties) and their destination (dynamic conversation property).
Acme Education has a vast library of online content, catering to multiple students across courses, universities, and countries (static user properties). These customers can also query their course prerequisites, which can change during the conversation (dynamic conversation properties).
Note: These properties (user properties, conversation properties, or bot variables) are already captured by the bots as part of the customer conversations. You might already be using these properties to trigger different flows, messages, and more. Filtered search feature will use these properties to narrow down the learning sources for your bots.
How to set this up?
Open the bot for which you need to narrow down the FAQs from which it learns. Now navigate to Natural Language > Configure > Natural language settings > Filtered search and toggle it on.
You will need to enter the properties that the bot needs to use to filter the search.
Any values mapped to these properties during a conversation will be used as a reference by your bot while searching for tags added to your articles/FAQs.
Choose between matching ALL of these properties or ANY of these properties to improve the coverage of the FAQs that the bot will narrow down while looking for answers.
- You can use the checkbox to include folders or articles that have no tags.
- This includes all your generic articles that are not specifically limited by properties. Essentially, it allows the bot to search for FAQs that have the matching tags AND for FAQs that do have no tags.
How does this work?
You can configure your bots to filter their search across knowledge sources using dynamic user properties, such as their currency.
During a customer interaction, your bots will gather the dynamic user property of the specific customer, whose value could be USD, EUR, INR, JPY, GBP, CNY, etc.
Your bot will then search for these values in your FAQs/Articles. For this to work, your FAQs/articles will need to be tagged with the values of each property.
In this example, the tags need to be currencies of different countries, which could be USD, EUR, INR, JPY, GBP, CNY, etc.
If the customer is using USD, your bot will search for FAQs/Articles that are tagged with USD, filtering out only the knowledge sources that have the USD tag.
Thus, use the right properties while setting up this feature and ensure you tag your learning sources appropriately:
Here’s how you can set up tags in your FAQs and here’s how you can set up tags in your QnAs
To continue with our initial examples, this is how bots narrowing down the FAQs will offer more relevant responses and improve the customer experience:
Acme Flights can now run a filtered search so that their bot gives accurate information about the immigration process for an Indian flying to London, which will be different for the same Indian flying to Maldives.
Acme Education can now run a filtered search and suggest course material that is different even if the same student is asking about different courses each day.
What are some other factors to consider?
The values that you enter that are to be used for filtering search are case-sensitive.
If the property used for filtering does not have a value associated with it during a conversation, the bot will ignore the property as it can not search for an empty tag.
In such cases, the bot will use other properties for filtering as long as they have values associated with them.
If the only properties available for filtering do not have any values associated, it will default to the fallback flow. Learn more about the fallback flow here.
If you select “Match ANY of the below properties” when one of the properties is a blank property will not cause any issues as the bot will be looking to match either of the properties and will skip the absent blank property.
If you select “Match ALL of the below properties” when one of the properties is blank, this will create a scenario where there are no matches at all. In this case, as mentioned above, the bot will default to the fallback flow. Learn more about the fallback flow here.
Filtered search will continue functioning even if your bot or customer converse in a different language. The respective language property value will be automatically picked up based on the locale.
Narrowing down articles will work for each bot, not for each version of the bot. If you delete a property from any version of a particular bot, the property will not be available for any version (including the previously published versions as well).
If you have used a property to define how your bot will narrow down the sources and then you try to delete the property, preventive validation will ensure that you do not delete that property. Learn more about preventive validation in your chatbots.