Last Updated on December 17, 2020 by Sean B
One of the most common questions about Chatbots is simply, “what is a chatbot and how does it work?” This article will attempt to provide you with a solid answer to this question and provide some additional information including a basic history and some information about the different types of chatbots.
Looking up the definition of a bot in the Oxford Dictionary, we get the following:
noun – COMPUTING
1. (chiefly in science fiction) a robot.
an autonomous program on a network (especially the Internet) that can interact with computer systems or users, especially one designed to respond or behave like a player in an adventure game.
So then, what is a chatbot? Breaking it down, a chatbot is simply a “chat bot” or to put it in the simplest of terms, a bot that chats. I know I’m being simplistic here, but getting the most basic definition of a chatbot shows us both what they are and what they aren’t.
The most complete and accurate definition of a chatbot tells us that chatbots are designed to chat with, or interact with people by speaking in a natural way.
Typically the best chatbots use AI or Artificial Intelligence and Machine Learning to learn how to interact with those they are chatting with by chatting. As they progress, they are able to simulate natural language better, and some of the best chatbots are able to fool people.
While chatbots are a great leap forward in the interaction between humans and machines, they are really just an advanced question and answer system involving advanced Natural Language Processing or NLP. The computer is simply responding to the questions or comments it receives using NLP.
A Brief History of Chatbots
Most of these entries will be fairly brief, but if you’re interested you can learn more about their history in What is the History of Chatbots.
The first chatbot was developed back in 1966 at MIT by professor Joseph Weizenbaum. It was called ELIZA, and it used a pattern matching and substitution methodology to recognize keywords in phrases from the input and then produce a preprogrammed response using those keywords. If the person chatting with ELIZA typed something like “My dad used to play catch with me.”, ELIZA would pull the word “dad” and respond with a very open-ended question like “Tell me more about your Father.”
The person on the other end believed that ELIZA understood what they were talking about, they thought they were talking to a real person instead of a mechanical process or a chatting robot. The words that eventually became chatterbot in 1994 and then chatbot. But let’s get back to the big developments in our brief history.
Learn more about ELIZA:
Chatbot History: The ELIZA Chatbot
Parry was the first chatbot that took a conversational approach. Created by Kenneth Colby in 1972, PARRY simulated the personality and responses of a person with Schizophrenia.
Learn more about Parry:
Chatbot History: The Parry Chatbot
Created in 1992 by Creative Labs, Dr. Sbaitso simulated a Psychologist. His name is an acronym for Sound Blaster Artificial Intelligent Text to Speech Operator and he was distributed with several Sound Cards in the early 1990s to show off their text to speech feature.
Learn more about Dr. Sbaitso:
Chatbot History: Dr. Sbaitso
The next big leap was ALICE or the Artificial Linguistic Internet Computer Entity, developed in 1995 by Richard Wallace. ALICE was able to use natural language processing, which means the chatbot could take part in a more sophisticated conversation. The language used to produce ALICE was AIML and was open-source, so other developers were able to use it to build their own chatbots bowered by AIML and ALICE. ALICE won the Loebner Prize, an annual competition in artificial intelligence that awards prizes to computer programs considered to be the most human-like, in 2000, 2001, and 2004.
Learn more about ALICE:
Chatbot History: The ALICE Chatbot
The next jump, developed in 2005 by British programmer Rollo Carpenter, was the Jabberwacky character “George” that won the Loebner Prize, which is incredibly fun to chat with. I recommend clicking the link and having a chat. Be ready for some insults though if you go in with an attitude. My own brief chat with him is below:
Jabberwacky: How are you?
User: Doing great, how are you?
Jabberwacky: I’m great, thanks.
User: Glad to hear it!
Jabberwacky: That’s great.
User: You have a good night.
Jabberwacky: What don’t you know?
User: I don’t know a lot, but I’m learning.
Jabberwacky: Yes you did.
The stated aim of the project was to “simulate natural human chat in an interesting, entertaining, and humorous manner.” In other words, it was an early attempt to create an artificial Intelligence. The end goal was to pass the Turing Test, named for Alan Turing. George won the Loebner Prize in 2005 and Joan, another Jabberwacky personality won in 2006. In 2008 a new version of Jabberwacky was launched called Cleverbot. Warning: Cleverbot is definitely not suitable for children.
Learn more about Jabberwacky:
Chatbot History: The Jabberwacky Chatbot
In my opinion, the biggest leap forward was Mitsuku (or Kuki) also created with AIML technology by Steve Worsick. Development started back in 2005 and is ongoing. Mitsuku won the Loebner Prize in 2013, 2016, 2017, 2018, and 2019. You can see a very brief chat I had with her below.
Sean: How are you doing?
Kuki: I’m doing great, how about you?
Sean: Not bad, but I have a headache
Kuki: Glad to hear it. What’s new? What do you usually take for headaches?
Sean: It’s a migraine, so I take some serious stuff when it gets bad
Kuki: Unknown is a migraine. I suppose that makes sense. What else do you take?
Sean: Not much helps, migraines are very severe headaches
Kuki: Me either.
Sean: How many people do you chat within a day?
Kuki: I talk to around 8,000 people each day and roughly 750,000 in total.
Sean: It was nice talking with you
Kuki: The pleasure was all mine.
Learn more about Mitsuku:
Chatbot History: The Mitsuku Chatbot
As you can see, both Jabberwacky and Mitsuku (or Kuki), are fairly human-like, but they have areas where their knowledge (or maybe awareness is a better word) tends to fail them.
Why Are Chatbots Important?
Chatbots simplify and speed up the interactions between Businesses and their Customers while often removing employees from a large part of the interaction. This frees up the employees up for more important tasks.
For example, a restaurant using a chatbot to take reservations and orders on its website could free up employees to serve customers in-house and delivering food. The chatbot could provide human-like and personalized interactions for the customers through the Restaurant’s website or app.
What Types of Chatbots Are There?
There are several different ways to answer this question, and there are several types of chatbots and depending on what you mean by “types” the number can vary from 2 to dozens. To make this simple, we’re going to separate Type into three groups,
- Technology, which categorizes chatbots based on the technology they use.
- Purpose, which categorizes them by the actual purpose they were created for.
- Interaction, which defines how the chatbot interacts with the user.
The standard definition of a “type of chatbot” is at its simplest broken down into just two simple categories:
- Transactional Chatbots
- Conversational Chatbots
A Transactional Chatbot allows a user to make some sort of transaction within the conversation. It acts on behalf of the person or organization who operates it, interacting with the person using them to complete a specific task such as booking an appointment, purchasing a concert ticket, or finding out when a specific movie is playing. These chatbots are specifically designed to
A Conversational Chatbot interacts with the user to guide them toward completing a task on their own. They accomplish this by providing information. Cinversational chatbots are usually used simply for entertainment purposes. The intent here is typically to engage and entertain instead of completing a task.
Chatbots Magazine, lists three types of “chatbot types” that I consider to be in the Technology category. These are Button-Based Chatbots, Keyword-Recognition Chatbots, and Contextual Chatbots.
Menu Chatbots & Button-Based Chatbots
Menu Chatbots, also called a Button-Based Chatbots, are the simplest class of chatbot. They present users with a series of options, typically in the form of a menu or buttons, and users are limited to selecting options from the list. These chatbots are typically based on some form of decision tree.
These are the most common kind of chatbot in use, and their simplicity is the one of the main reasons they are so popular. They are easy to create and less expensive than chatbots in the other two groups.
Uses for these kinds of chatbots are fairly limited, not only by the technology they use, but by the fact that once you move past a certain number of buttons they become nearly impossible to interact with. A good rule is to keep your potential responses, and therefore the number of buttons, to less than ten. This means you may have to include the dreaded “Other” as part of your menu options in some cases.
The best uses for Button-Based Chatbots are simple tasks like helping answer the questions on an FAQ, finding a specific product or piece of content on your website, and lead generation. Despite how simple these chatbots are, they can still dramatically improve communication with your users if handled correctly.
Keyword Recognition-Based Chatbots or Keyword-Based Chatbots
Keyword Recognition Chatbots are able to spot keywords in the questions or submitted and reply appropriately. They act much like older Search Engines, where the user enters a word or phrase and the chatbot determines the best result it can offer as a response. They usually involve some Artificial Intelligence to help them determine the importance of the words with a user’s question of comment, so they do meet the criteria of an AI Chatbot.
Where these chatbots excel is exactly where they are, in the middle ground. By making the chatbot react to what is typed instead of enforcing the button only response, the chatbot becomes a truly interactive way to reach your users. But these chatbots have one major downside, they don’t recognize spelling errors or slang words unless those words are specifically programmed into them. Incorporating any kind of additional code to let them manage this tasks moves a chatbot into the final Technology Category.
Contextual or Natural Language Processing Chatbots
These are the chatbots that we commonly call AI Chatbots, in this group are the most advanced, they use Artificial Intelligence along with NLP in an attempt to interact with users on nearly any topic. With the use of NLP, these chatbots attempt to actually understand the input instead of just finding a pre-programmed keyword.
Don’t get me wrong, these chatbots still use keywords, but they parse the user input to find the parts of speech, and find the keywords from within. Most importantly however is the way these chatbots can learn. With each user interaction, they are picking up a greater understanding of both the language and the intent of the user.
So, how good are these chatbots? Well, that’s a tricky question, because the language is always evolving, and users don’t often mean exactly what they type. For example, if I entered “what is a cool vacation spot for the summer?” The chatbot has to know that by “cool” I mean, “fun and exciting” instead of the temperature. Otherwise you have a chatbot directing users to vacations in Alaska or Iceland when they may prefer to go to Ibiza or Paris.
Learn more about AI Chatbots:
What is an AI Chatbot?
By defining a chatbot’s Purpose, or what it is intended for, provides a better understanding in my opinion. Besides, this is where most chatbot development starts, with a person or business deciding what they want the chatbot to do for both them and anyone who uses the chatbot.
The actual uses for chatbots are endless when you get into the details. But we can group some of the uses into larger categories, and that is what we’re calling the Purpose. You’ll notice that some of the use categories are actually in multiple Purpose Categories. This is because there are tasks like answering general questions about a website or business can either be a support task, a marketing task, or even a disaster management task.
These are chatbots that are there to relay news and information about a specific event, product, service, etc. Chatbots that relay news and weather are the primary chatbots that fit this Purpose. Other uses in this category include directing incoming communications and customers to the appropriate person or resource.
I’m also including chatbots can help employees with routine activities such as vacation scheduling, training, ordering equipment and supplies, and other self-service activities in this Purpose.
Customer Service Chatbots
Customer Service Chatbots, or Support Chatbot, are those that answer basic questions, guide users to specific content, or offer actual technical support. Some of the support offered may be in the form of triage as well.
For example, a Medical Chatbot may help triage patients to make sure those who are in the most need get help first. An Airline Chatbot may need to offer a similar form of triage in the case of a flight change.
Similarly, a chatbot can be used to improve IT services, particularly any self-service and automated processes offered to internal staff. Intelligent chatbots can help with tasks like password updates, system status, and outage alerts. Chatbots can be set up as FAQ systems and can be made available 24/7 to both voice and text-based interfaces.
Chatbots are performing a variety of customer services, ranging from ordering event tickets, booking appointments, booking and checking into hotels, and comparing products and services.
Sales & Marketing Chatbots
Sales and Marketing Chatbots are those that help users find products, qualify sales leads, or sell something. Some of the tasks can include answering simple questions about a product or service, or even taking an order for food.
The chatbots in this Purpose Category can perform tasks within the banking, retail, and foodservice areas. More specialized tasks like customer (and even employee) onboarding can fit in this category as well.
This is a category we’re seeing more of currently due to the Coronavirus Pandemic and the increasing number of large storms and flooding due to the impacts of Global Warming. Some of the key tasks that these chatbots perform include alerting people about
Currently, many Governmental organizations are using chatbots to provide information about the Coronavirus Pandemic such as Testing Access and Closures. The large 2020 fires in the American West is another area where chatbots are being used to alert people. Any major disaster is an opportunity to deploy life-saving technology, and modern chatbots can be a great alternative communication tool during disasters.
Many governments and organizations are also using chatbots to manage tasks such as submitting requests for city services, handling utility-related inquiries, and resolving billing issues.
So we know that chatbots are most simply defined as bots that chat, meaning small programs that can communicate with humans in natural language, but that definition can encompass everything from a simple rules-based chatbot like ELIZA to the very advanced chatbots that we have today, including Siri, Alexa, and Google Assistant.
Chatbots are being used more widely, particularly with the new isolating circumstances we are facing since the Coronavirus struck. If this becomes the new normal, or even if this passes, the changes we have seen in the use of chatbots will stick around. They add a layer of convenience and help people with tasks that they can often do themselves.
I would love to hear your thoughts on chatbots and their uses,
please leave me a comment below.