Learn all about creating and designing small talk for a chatbot that can actually have a real chat conversation, which can help build a deeper connection by strengthening it’s language with human qualities.
Have you ever had an opportunity to talk and chat with a chatbot, only to be disappointed in it’s ability to create small talk? In this post, I’m sharing with you some design principles, free available small talk data sets, and things to consider when implementing small talk with a chatbot.
When someone gives your chatbot a virtual knock on the front door, you’ll want to be able to greet them. To do this, give your chatbot the ability to answer thousands of small talk questions in a personality that fits your brand. When you add a knowledge base full of these small talk conversations, it will boost the users confidence in your bot.
What is small talk for chatbots?
Small talk are social phrases and dialogue that express a feeling of relationship and connection rather than dialogue to help convey information. Examples of categories of small talk for chatbots are greetings, short snippets of conversation, and random questions serving as a gentle introduction before engaging in more functional capabilities of the chatbot. General topics for chatbot small talk includes weather, politics, sports, television shows, music, songs, and other pop culture news.
Chatbot small talk is important because it allows users to test the limits of your chatbot to see what it is fully capable of. It is the user’s first foray into understanding how much conversation and dialogue that your chatbot can really do. When designing a chatbot, small talk needs to be part of the development process because it could be an easy win in ensuring that your chatbot continues to gain adoption even after the first release.
What is the difference between phatics and small talk for chatbots?
Phatics, small talk, and social talk for chatbots refer to the same definition, which is an informal type of conversation that does not cover any functional topics or transactions, serving users to help explore the capabilities of the chatbot.
Why implementing small talk, social talk, and phatics matter for a chatbot?
Implementing small talk for a chatbot matters because it is a way to show how mature the chatbot is. Being able to handle off-script requests to manage the expectations of the user will allow the end user to build confidence that the bot can actually handle what it is intended to do. This allows the user to potentially become a return user, thus increasing the rate of adoption for the chatbot.
The more the bot can perform, the more confidence the user has, the more the user will refer to the chatbot as a source of information to their counterparts.
Chatbots already have a preconception around being brittle bots that can’t talk about anything that they have not been trained on without personality or a long-term memory. This causes most chatbots that have been developed to fail, because they fail initially to confirm to their audiences that they can do more than the specific skill they are trained on.
How to Implement Small Talk with a Dataset
In most bot frameworks and platforms, there will be a way to create an intent for small talk. Being able to create intents and entities around small talk will help your NLU or NLP engine determine what types of questions get routed to the data set that can be answered.
By creating an intent and entity, as the chatbot intended users start asking questions, you can continually modify the model behind the intents and entities on the type of small talk that the specific audience of the end users prefer talking about, thereby continuing to learn the traits of that audience to provide a better small talk chatbot experience.
Considerations for Implementing Small Talk in Your Chatbot
As someone who has been in the trenches, when a feature comes out, I find that new developers try to drive towards getting it done by hardcoding rather than thinking through the entire lifecycle of a feature. I wanted to talk about the best practices on how to architect small talk so that it’s extensible and can be maintained.
First, ensure that the dataset that is being pulled from can be added to by a non-developer. If you have someone who is building a bot, you should also have a separate individual that is reviewing the dialogues when the chatbot is released. As the chatbot dialogue is being evaluated, there needs to be an easy way to add to the small talk intent so that the dialogue base continues to grow. Being able to tie the chatbot to a dataset that a non-developer can maintain will make it easier to scale your chatbot’s small talk data set.
Secondly, ensure that you create an intent and entity for small talk. Generally, I recommend one so that you can encompass all the things that the chatbot can talk about at an intrapersonal level and separate it from the specific skills that the chatbot actually has. Having an intent will allow you to train alternative utterances that have the same response with efficiency and ease.
FREE Conversational Datasets for Small Talk
Here is my favorite free sources for small talk and chit-chat datasets and knowledge bases. All of these are free and you’ll just need to extract them to use it as your own. I’ll continue to update this as I find more datasets:
They have a knowledge base provided along with the personality traits for small talk that you can use. The five personality traits for small talk and chit chat are: professional, friendly, witty, caring, and enthusiastic.
Types of Small Talk and Fallback Dialogue Categories to Include
There are generally a certain subset of topics and categories that are used to have conversational dialogue back and forth with the chatbot, which include:
- Fun Phrases
- Positive Sentiment
- Negative Sentiment
- Pop Culture Trivia About Latest Shows & Songs
- Questions to Test Your Bot
Greetings – Small Talk
- Hi / Hello / How are you?
- What’s up? / Yo / How you doin?
- Good Morning / Good Evening / Good Afternoon / Good Night
- What can you do? / What can I do with you? / What are you doing?
- Thanks / Thank you / Thanks dude / Thanks lady
- My name is <insert user name>
- How can you help me? / Help me please
- Talk to you later!
- Tell me something that you can do
- Say anything
Fun Phrases – Small Talk
- Will you marry me? / Are you Single? / Do you have a boyfriend/girlfriend?
- Do you know other chatbots?
- Do you want to rule the world?
- Are you busy?
- Say something funny
- Can you say anything else
- What do you think of me?
- Happy <insert holiday>
- Do you love me?
- Can you fly?
Positive Sentiment – Small Talk
- You are a genius!
- You are awesome / great / special
- Thanks so much!
- Thanks, that’s very helpful
- Yup, that’s true
- That’s accurate
- You are right
- Give me a fist bump
- I think you are great!
- Thumbs up
Negative Sentiment – Small Talk
- You suck
- I hate you
- Go away
- That’s not right
- That’s not true
- That was stupid
- You are not helpful
- That was not the right answer
Pop Culture – Small Talk
- Which baseball / football team do you like?
- You know nothing Jon Snow.
- Where’s Waldo?
- What’s the weather?
- Where’s the best restaurant in town?
- What should I eat today?
- Who is the best <insert role>?
- Who should I vote for?
- Who is the best digital assistant?
- Do you know Alexa?
Questions to Test the Scale of Your Bot
- What’s your age?
- Can you sleep?
- Who is your boss?
- Who created you?
- Do you know other chatbots?
- What is your favorite color?
- Why did the chicken cross the road?
- Happy Birthday
- Do you know me?
- Are you my assistant?
Lessons Learned from Implementing a Chatbot without Small Talk
As a product manager driving the roadmap for our internal chatbot that serviced over 30,000 employees, I decided to launch our chatbot without a full list of small talk and phatics. The reason was because I just wanted to get the chatbot out the door to see what people would ask it EVEN WHEN I told the audience that it could do one of three things.
As I analyzed the data that came back in the conversation log, the evidence was overwhelming. People WANTED the bot to have small talk. The data was telling us that our employees were using small talk to test the limits of what are chatbot could do, and our surveys afterwards told us that the small talk help boost their confidence on what the chatbot could do, helping them to understand whether they could trust that the time spent asking the question was worth revisiting the bot again.
It was only after three months that we decided to implement what we called a chit chat, which is basically another way to say small talk. This helped tremendously with our adoption and our ability to decreased our missed intent metric.
Related Topics to Small Talk Chit Chat for Chatbots
Which language is best for a chatbot?
The following programming languages are the most popular and commonly used for developing a chatbot: Node.js, Python, Java, Ruby, and PHP.
How can I make my small talk better in a chatbot?
Small talk with a chatbot can be made better by starting off with a dataset of question and answers that encompasses the categories for greetings, fun phrases, unhappy. users, pop trivia, and confidence testing questions. In addition, being able to go two levels deep with follow-up questions can help make the discussion better.
Summary of Chatbot Small Talk
The chatbot medium of engagement is still a new innovation that has yet to be fully adopted and explored by the masses. Being able to provide your chatbot with great conversational faculty will undoubtedly convey confidence to the user that the bot can surely help them with their functional question that the bot was actually intended to help them with.
Small talk for chatbots is a pretty simple feature to implement (it only took us a single two week sprint to implement) that makes a very large impact on the user interaction and experience to boost confidence and adoption.