Talk to me: what voice bots can do today

18 years ago, in the anime series Ghost in the Shell: The Lone Syndrome, they showed absolutely fantastic robot operators. In addition to very advanced physical capabilities (you would have to type a code with such fingers!), They were perfectly able to communicate with people in voice. Today, bots will not surprise anyone. We meet with them everywhere, in various areas of our lives. They are used in medical institutions, banks, in sales, in customer service, in social networks, etc. And if we are already starting to get used to chat bots, then voice bots are still a novelty for many. Yes, and the technology itself is still far from perfect. What is the situation with voice bots now?

The story of voice bots began with chat bots back in 1966. The first such development was created by Joseph Weizenbaum and was called "Eliza." The bot conducted a dialogue as a psychotherapist, questioned the phrases and demanded to continue them. At the first communication, many did not realize that they were communicating with the machine. Phrases were built according to the type:

- My head hurts.
“Why does your head hurt?”

In fact, “Eliza” was a parody of a psychotherapist and did not perform any functions. But, nevertheless, it was the first chat bot.

The next bot appeared in 1972 and bore the name "Parry". He played the role of a patient with schizophrenia and managed to deceive many doctors who spoke with him.

A significant milestone in the development of bots was ALICE (Artificial Linguistic Internet Computer Entity). This was the first bot that could conduct a practically full-fledged dialogue with a person, using heuristic analysis to compare received phrases with those available in the database. The answers were fairly general, but even such communication for 1995 was a breakthrough. Since then, the interaction algorithms of bots with users are improved every year by leaps and bounds. With the advent of neural networks, bots have become much smarter. When communicating with different people, bots using machine learning replenish their database and learn to speak as a person.

With the development of speech recognition technology and voice synthesis, voice bots have appeared. This is the next step to a full-fledged virtual assistant who can solve all sorts of simple tasks without human intervention. There are many areas of use for such assistants: replacing call center operators at the initial stage of processing applications; technical support in simple and most common questions; automation of sales and ordering goods; customer alerts and surveys; electronic assistant secretaries who do not get tired and do not forget anything; smart answering machines, etc. Gradually, voice bots replace entire call centers and process all incoming calls from clients. And if the bot cannot solve the issue that the client has contacted, it switches the call to the human operator.

Also, various “smart” devices begin to equip voice bots. Thanks to the development of IoT, virtual assistants are increasingly appearing in our lives. For example, in 2018, every sixth adult American had smart speakers . This is a device with an integrated voice bot that connects to the Internet via any available interface and can play music from streaming services, order delivery or a taxi, read mail, find information on some request, tell a child a fairy tale, voice the weather and much more. The sales growth of these gadgets in 2017 was 128%. In Russia, foreign devices that do not speak Russian have not received distribution, but recently domestic developments have appeared on the market. They have the same set of capabilities as their foreign counterparts, and in some areas surpass them.

Soon we will also release our smart column “Capsule” ( pre-order is now open) She was born a little later than similar devices. The only button on the column mutes all 6 microphones, the rest is controlled by voice or using a tactile sensor on the top of the head. The first connection is quite simple - you need to download the Maroussia application on your smartphone and connect the speaker through it to Wi-Fi. Authorization in VK allows you to listen to music from your playlist and call friends: “Capsule” is the first smart speaker in the Russian market that can call over the network. Or you can listen to your own selection of music on your smartphone via Bluetooth. You can also use the column as a calculator: ask her to add or multiply something, and she will immediately say the result.

Its capabilities are constantly expanding. For example, recently added “office sounds”: you are sitting at home on a remote place, and so that the changes are not so painful, you include the usual office noise. We have also taught Capsule to play games: quizzes, Cities, Erundopel, Believe - Do Not Believe, Fanta and others. And in the near future we plan to teach her to entertain children: to make the voices of various animals and interesting sounds, to play educational games using RFID cards, etc. Now Maroussia is able to conduct dialogs and answer questions, but soon there will be an opportunity to order a taxi, food, movie tickets, etc.

In the near future, voice assistants will be able to perform the tasks of the referent. For example, there is a version of Google Duplex Voice Assistant. So far, only owners of Pixel smartphones in some US cities can use it, but it is planned to introduce it on iOS and Android. The bot's presentation took place in 2018 at the Google I / O conference . The bot called the restaurant and reserved a table, communicating in a voice indistinguishable from the human. In the near future, a significant expansion of the capabilities of Duplex is planned. He will be able to make calls on behalf of the user, call repair services, arrange meetings and much more. In this case, the person at the other end of the wire will be notified that he is communicating with the bot on behalf of the owner.

Call centers based on voice bots are becoming more and more popular. So, based on the same Duplex, Google introduced the intelligent voice system Google Contact Center AI. Artificial intelligence will be able to receive all incoming calls, process them and decide on the need to connect a specialist to solve the problem. In addition to Google, voice assistants are being developed by many companies, including KVINT, Robovoice, Glagol, etc.

Bots as call center operators can be used both for outgoing calls (informing customers, polls, marketing research coordination, clarification of orders and much more), and for processing incoming calls from customers, solving simple tasks described in the database. If necessary, the bot switches the call to the operator of the corresponding department: sales, service, technical support, etc.

Voice bots can help when you need to make or receive thousands of sample calls per day. One robot replaces dozens of operators 24/7. Naturally, to completely replace a person so far will not work: there are many situations that the robot will not cope with. But at least it can significantly reduce the load on call center operators.

For example, some Russian banks already use voice bots ( 1 , 2 , 3 , 4) The robot can tell the balance, address of the nearest ATM, currency exchange rate, etc. For example, VTB Voice Assistant successfully processes about 25% of customer requests. The robot answers the question within a few seconds and it does not depend on the time of day or the number of free operators. When the client sets out the essence of the issue, the bot will automatically redirect it to the appropriate service. The call is transferred to the operator only if the answer to the client’s question is not in the database. The system is based on a neural network and is self-learning. Since the introduction of the voice bot, the level of speech recognition of bank customers has grown to 85%. The robot automatically collects and structures information at the request of users. Thanks to these data, VTB improved some services and launched new ones (SMS informing about the date the card was ready, about branch addresses, etc.).The bot significantly reduces the burden on the employees of the bank's call center, and also saves the time of specialists.

Voice bots have flaws. Although technology allows you to recognize most of people's phrases, but when the question is different from the typical one, many robots begin to lead the user in simple circles in circles, trying to find a solution to the problem without involving a specialist. This is very annoying to customers and causes dissatisfaction. Also, the bot can erroneously recognize the question and give the wrong answer, which also takes away customers' time and nerves.

Active corrections are underway to correct these and other shortcomings. Databases and machine learning technologies are being improved, and methods for synthesizing and recognizing speech are developing. The voice bots of the next generations are already being tested. For example, a bot for outgoing calls is being tested at VTB, which can offer banking products, talk about the conditions for their use and respond to customer objections.

Today, many startups are developing solutions in the field of voice bots. Some of the interesting ones include:

ELLIQ from Intuition Robotics. It is a voice assistant designed for the elderly. It makes it easier for them to dive into the digital world and communicate with the Internet of things. ELLIQ will talk about how to use chats, instant messengers, social networks, notify you if a new message has arrived from friends or family or a call has been received from someone, or will offer to call your relatives yourself if they appeared on the network. The assistant can select a playlist, order a medicine or products, voice the weather for the coming days, remind you of the need to take medicine, etc. ELLIQ will help to understand even the very poorly versed people in the world of digital technologies.


Chris by German Autolabs . This is a voice assistant for drivers. It allows you to connect the applications necessary for the driver, make calls and answer them using voice commands and simple gestures. Also with its help you can turn on music, switch between tracks and radio stations. Chris can tell you the way, find the best route and show you how to save time by going around traffic jams.


Ada Health . A specialized medical assistant developed by a team of doctors and scientists to help take care of our health. Currently, the assistant is a chatbot in the form of a separate mobile application, but in the near future it is planned to introduce voice mode. The bot helps to diagnose and select a treatment based on the entered symptoms and description of health problems. Startup works with leading healthcare companies and builds on their expertise. Since the launch of the project in 2016, the application has spread to 140 countries and has 15 million ratings.


Dasha.ai . A universal system offering outsourcing services for calling and answering calls. “Dasha” is able to call customers, sell goods, send letters based on the results of a conversation, chat, and answer questions. Of course, she needs a detailed script to work. “Dasha” understands synonyms, remembers the words and context used by the interlocutor; recognizes when she is interrupted or asked to call back. A person hears normal “human” speech on the phone, and many do not even understand that they are communicating with the robot:

https://dasha.ai/ru/wp-content/uploads/2019/09/dasha_demo_1_mp3.mp3

https: // dasha. ai / ru / wp-content / uploads / 2019/09 / dasha_demo_2_mp3.mp3

* * *

In the future, voice bots will completely replace numerous call centers and will solve most customer requests, allowing to minimize the work of specialists. Electronic assistants are also developing, going beyond the usual gadgets. They will settle in all household appliances, cars and even in clothes. Instead of numerous assistants on different devices, there will be one assistant for each user.