People sometimes ask why you would need an FAQ chatbot when you can accomplish everything it does with a simple site search.
Here are some reasons.
Site search works well only with exact keyword matches
Usually, the way most site search boxes are set up, they only work with exact keyword matches. In contrast, a well designed FAQ chatbot is usually quite good at understanding variants of a user query.
Site search forces people to use simple keywords rather than full questions
Google search now allows people to do searches by asking full questions which are more natural sounding. But the search boxes on the majority of websites still expect terse, keyword focused questions. An FAQ chatbot can accommodate longer questions in more natural sounding language.
Site search cannot perform context specific queries
Dialogflow, for example, has a way to set contexts for your user's chat session. This could be something simple, like choosing a topic (portal), or something more involved, like asking for specific information about the user to personalize the results. While you don't have simple ways to do the same with site search boxes, an FAQ chatbot can be designed to be context specific and can return better results.
Site search can never be conversational
While chatbots have a long way to go before they become truly conversational, in the limited context of answering website-specific questions, they can use knowledge from the user's previous questions and provide better answers.
Designing an FAQ chatbot to do this can be a bit challenging, but this feature doesn't even have an equivalent in your typical site search box.
User: Are there any buses going from Redmond to Seattle?
Bot: Yes, bus number 545
User: When is the next one?
Here, the bot is able to understand that the word "one" refers to bus number 545. We can infer this from the user's first question.
Site search cannot collect user feedback in the midst of a task (unobtrusively)
The typical site search is designed to take a query and then show a set of 10 or so links. Within this UI it is not intuitive to also collect feedback from the user about any aspect of the service. Where will you ask the question? On the search results page? Or on the actual page which the user was looking for?
Think about the last time you visited the knowledge base section of a website. Somewhere on the page, there is usually a little widget asking "Was this information helpful?"
For example, this is a little widget at the bottom of the Microsoft Office documentation site.
Notice that it is intrusive to the reader at first (when you are actually reading the information)
But once you reach the bottom of the page, when you will actually be ready to provide the feedback, it also gets lost in a sea of gray. 🙂
While it is quite possible that the Microsoft team did plenty of UX research and decided that this is the design which works best, it is still not a great UX from the user's perspective.
Thankfully, they get an A+ for the UX on all of this (see animated GIF at the end of this post). But you didn't quite know what to expect did you?
With a chatbot, since the user remains in typing mode throughout, collecting this feedback fits more smoothly with the flow as you might imagine.
A chatbot helps you find out user's unanswered questions in their own words
Dialogflow has this concept of "Fallback Intent". When the agent cannot understand what the user says (i.e. none of the predefined phrases match) it will invoke this fallback intent and reply with a message such as "Sorry, but I didn't understand what you said". In your Dialogflow console, you can quickly scan and find these unmapped user phrases inside the Training tab.
Once you find these, you now have a way to see the questions your audience has but which aren't answered on your site (assuming your FAQ chatbot is quite thorough).
You can, in theory, get these queries by looking for empty results in your site search software also. However, there are a couple of problems:
- Sometimes it isn't easy to set this up, depending on what solution you use for the site search
- Even where setup is easy, you don't still get the full user query because most users are going to only type out the key phrase
When you find these unanswered questions, you might want to create a new post/article on your site using the exact question as the title (Google loves "How to.." articles). You can't do this quite as easily with a site search.
The problem with using a DialogFlow based FAQ Chatbot
At this point, you might be quite convinced that you need an FAQ chatbot for your website.
There is one major issue if you choose Dialogflow for creating this chatbot. The native web demo integration doesn't support clickable hyperlinks. And without hyperlinks, an FAQ chatbot is practically useless.
So how can you get over this? It depends on your needs. You need to write some custom code whichever choice you make. Read this article to learn more.
And here is the animated GIF which shows what happens when you click on one of the Yes/No choices on the Microsoft documentation site.
- Dialogflow vs RASA NLU
- Dialogflow vs Lex vs LUIS vs Watson vs Chatfuel
- Machine Learning vs non-Machine Learning algorithm
- BotFlo update
- Learn Dialogflow basics for free (till May 31)
- 10+ practical projects to learn spaCy in depth
- An Epidemiology Glossary for Programmers
- All my mini-courses are free this week
- Reader Question: What if a specific system entity isn’t available in all languages in a multi-lingual bot?
- How much can Machine Learning ACTUALLY help with answering free-form questions?