Can you recall the last time you had to dial a toll-free number for customer service? Do you remember the long wait time before you could even start talking about your problem? Then, all of the sudden, it crossed your mind you had pushed the wrong button leading you to the wrong division… No doubt this sounds familiar to most of us.
Fortunately, there is a powerful new technology that has come to relieve us from this frustration. It goes under different aliases such as (conversational) AI assistant, virtual assistant, or simply chatbot. Researchers, though, insist on calling it a dialogue system. Let’s stick to this name then. You use dialogue systems day by day, from the assistant on a smartphone to email systems that sift through your incoming mail.
The global COVID-19 pandemic has been especially prominent at stimulating the trend towards self-service automation through dialogue systems. Now more than ever, businesses are looking for ways to boost self-service facility. A general direction is to automate routine customer inquiries and relieve support team for higher-value tasks.
What Is a Dialogue System?
It’s a computer program that supports spoken, text-based, or multimodal conversational interactions with humans. Some people believe that dialog systems are one of the most important natural language processing (NLP) applications of this century. For the first time in history we can speak to a machine in our own language, and we can’t always tell that it isn’t human.
Eliza: The First Chatbot
Although chatbots may sound like a recent buzzword to you, they’ve been around since researchers invented a way to communicate with computers. As a matter of fact, the very first chatbot Eliza was built in 1966 – even prior to launching the early personal computer. Its creator was Joseph Weizenbaum at the MIT Artificial Intelligence Lab. Posing as a psychiatrist, Eliza inspected the keywords in the user input and generated the rules for transforming the output. This specific methodology of producing responses is still extensively being employed when developing dialogue systems.
If you feel like reading more about where chatbot technology originated and how it evolved, you can find a nice brief history in [Raj, 2019, p. 13-16]. It will surely impress you how far the technology advanced since the beginning.
Limitations of Early Dialogue Systems
While there is much to be gained from the accomplishments of the early dialogue systems, in many cases they had one or more of the following limitations:
- they were usually extremely fragile and would fall over or crash due to the slightest departure from the anticipated input;
- the systems performed well for the goals for which they were developed but did not scaleup or transfer smoothly to other domains;
- dialogue decisions were handmade and thus could not be assured to be the best;
- the systems were frequently built using proprietary toolkits and languages that were not necessarily available for public use; and
- they concentrated only on voiced or written language and didn’t take into account other modalities that are crucial in natural conversation.
Dialogue Systems Today
Driven by ever improving pattern-matching technology, chatbots evolved a great deal since the times of ELIZA. And brand new approaches have been devised to reinforce pattern matching. This greater sophistication is probably the answer why chatbots are often referred to as dialog systems in recent literature.
Early chatbots resided either on dedicated servers in universities and industrial research laboratories or as voice-enabled user interfaces via phone networks. These days you can come upon dialogue systems on a large variety of openly available platforms and devices. A dialog system now can be shaped as a messaging app on your smartphone, or as a personal digital assistant on your tablet.
Given the recent progress in voice technology, IT front-runners such as Apple and Amazon launched artificial intelligent agents for voice. Now you can ask Siri or Alexa to assist you with hiring a car, turning off the lights, playing your beloved music from Spotify, etc.
In What Ways Modern Dialogue Systems Are Superior?
Contemporary dialogue systems have addressed many of the limitations of early systems:
- they can be developed and deployed on popular messaging apps that people are already used to;
- the user does not need to download and install separate apps for each application;
- usually the systems can retrieve contextual information about users, such as their
location and physical wellness, that may have been obtained via sensors;
- the systems are capable of learning from experience, so they can dynamically improve their performance;
- an increasing number of systems are able to communicate in multimodal fashion by interpreting the user’s eye gaze, gestures, and head movements;
- developers can now use many easy-to-master toolkits that take advantage of the latest advances in AI, machine learning and NLP.
Dialogue Systems from the Business Perspective
Dialogue systems are a very promising technology that is gaining acceptance now because of the remarkable progress in machine learning and artificial intelligence. The use cases described below should provide you with ideas where dialogue systems could make a difference to the user experience today.
For regular things such as checking a balance, finding a branch nearby, or requesting a money transfer to another account, a chat interface could be especially handy. In addition, the dialog system could easily accommodate standard customer support issues such as blocking a stolen card or requesting a new one. The dialog system has a direct interface with the back-end system and is given the right permissions to execute operations on the user’s behalf.
Insurance activities usually require a rather lengthy dialog between the customer and the insurance company. Fortunately, most of the time, the data exchanged between the two parties is well structured and can be easily automated. Example use cases for a dialog system could be registering an insurance claim, finding out the status of a claim, and getting information about other insurance products. In addition, dialog systems lend themselves for cross-selling various other products based on the buying pattern of the user. Another use case where a dialog system could assist the user is deciding on the right plan based on some initial questions. Users often have no idea of the offerings they might be eligible for, and dialog systems can help drive sales higher by capturing and utilizing the sales data.
Travel is a huge market with a great deal of customer interaction taking place before a sale is
made. Price is one of the key drivers for sales in the travel industry; users always search for the best
price to pay when booking a flight or a room. Some companies now even provide real-time prices of flights and hotels. One use case could be building a dialog system that communicates with a couple of back ends to get the real-time pricing. Imagine having a message prompted as soon as the pricing of a seat changes!
Food and Restaurant
There are many simple-to-use and simple-to-build use cases that can be automated on top of a dialog system in the food industry. Booking a table is among the most popular use cases with a surprisingly big share of bookings still being taken over a phone. Instead, it would be much more convenient to access a dialog system and book a table for any number of people while on the go.
In this sector, there are mainly two functions that a dialog system could carry out: product search and customer support. Automating customer support for e-commerce is an enormous market, and with the breakthroughs in the NLP, automated systems will take care of all customer support queries in the near future.
Utilities and Bills
Everybody uses utility services, and paying a bill is a use case calling for automation as well. In fact, dialog systems that aid users in administering their utilities are among the fastest-growing applications today. Telecommunications and electricity companies might benefit by providing their customers with a dialog system running on a web site or on various social media platforms. End users will gain a lot by having convenient bill fetching and payment services offered by such dialog systems.
This was the first article in the technology.org series on Dialog Systems, where we focused on the business side of building dialog systems. In the end, you want your clients to be satisfied and able to accomplish more by using your product or service. Dialogue systems allow users to always remain in touch with the brand, and they offer your business a unique opportunity to conveniently engage the user.
In the second part of the technology.org series we will attempt to give a gentle introduction to the technical aspects of developing dialog systems for real-world applications.
Darius Miniotas is a data scientist and technical writer with Neurotechnology in Vilnius, Lithuania. He is also Associate Professor at VILNIUSTECH where he has taught analog and digital signal processing. Darius holds a Ph.D. in Electrical Engineering, but his early research interests focused on multimodal human-machine interactions combining eye gaze, speech, and touch. Currently he is passionate about prosocial and conversational AI. At Neurotechnology, Darius is pursuing research and education projects that attempt to address the remaining challenges of dealing with multimodality in visual dialogues and multiparty interactions with social robots.
Andrew R. Freed. Conversational AI. Manning Publications, 2021.
Rashid Khan and Anik Das. Build Better Chatbots. Apress, 2018.
Hobson Lane, Cole Howard, and Hannes Max Hapke. Natural Language Processing in Action. Manning Publications, 2019.
Michael McTear. Conversational AI. Morgan & Claypool, 2021.
Sumit Raj. Building Chatbots with Python. Apress, 2019.
Sowmya Vajjala, Bodhisattwa Majumder, Anuj Gupta, and Harshit Surana. Practical Natural Language Processing. O’Reilly Media, 2020.