Using customer questions to train your bot
One of the benefits of FAQ Bot is the ability to see the exact questions your customers are asking, in their own words. You can then use that conversation data to train your bot.
It's fast and easy to use conversation data to train your bot
Some of our customers have even used conversation data to inform the development and structure of their website.
"Chatbot development provides multifaceted benefits for our organisation... He [chatbot] has identified the need for the Board to rethink the way the website is structured and how our information is presented."
We don't always chat to a bot with perfect spelling or grammar
When someone asks your bot a question, it will first try to find a match in your knowledge base. If it can't find a confident match, it will try to find answers that might be relevant, and present those as options your customer can pick from.
Previously, you didn't have control over which variant of a question your FAQ Bot would display in these "did you mean?" type answers. So if you had added question variants that were incomplete fragments or included spelling or grammar mistakes, there was a chance that they would show up in the bot.
While we do want to use those questions for matching, to improve the bot's understanding, we don't necessarily want to show them back to users.
This "did you mean" includes an answer with a lower case 'i'
Display questions for preferred question variants
With our new display questions feature you can now prevent inappropriate question variants being offered by your bot as "did you mean?" answers.
When you're reviewing your chat data, and you spot an inappropriate "did you mean?" presented back to the user, you can review that question in your knowledge base and add a preferred display question.
This is especially useful for chit-chat type questions.
👉 Ready to find out more about how FAQ Bot works?