VBD 
“Virtual assistant” is a term whose prominence has increased significantly in recent times, as it has been integrated into many smart devices like smartphones, smart TVs, cars, etc. A virtual assistant is a product of the combination between artificial intelligence (AI) and machine learning (ML) models and algorithms. Virtual assistants are developed to assist users in many operations and therefore, optimize customer experience. This article concentrates on their exact definition as well as their breakthrough functions.
What is a virtual assistant?
Virtual assistant, which can also go by the name “AI assistant” or “digital assistant”, is an application program that can understand natural-language voice commands or texts, and use such input to assist users in many operations. Such operations can include reading messages, e-mails, searching for phone numbers, setting up meetings in calendar, making phone calls and reminding users of previously arranged appointments.
Developed by artificial intelligence, virtual assistants act as a personal secretary for users. They are normally integrated into daily used smart devices so that users can easily connect with and give commands to them at any location or moment they wish. Researchers are always attempting to make virtual assistants smarter, more flexible, more understanding of users and finally, more capable of handling instructions simultaneously. Prominent virtual assistants in the world include: Siri (Apple), Alexa (Amazon), Google Assistant, etc.
Interaction methods of virtual assistants
Alongside with the continual development of AI and ML models, virtual assistants obtain an increasingly diverse number of ways to interact with users. Such interaction methods are becoming more and more flexible and adaptable to daily natural-language conversations.
Texts
Texts are the first interaction method that is used for virtual assistants. Almost all current virtual assistants support this kind of interaction. Users only need to give instructions by sending a message and virtual assistants will respond to the instruction immediately. A typical illustration of this is a chatbot system. As users initiate a conversation with, for example, an online vendor on Facebook, they can receive a response within seconds and this might be due to the vendor using a chatbot.
Chatbots are growing in popularity and can handle user instructions in a smarter and more friendly way. Nowadays, users do not need to input commands based on a set template as has happened in the development state of virtual assistants. They can now input natural-language voice commands which ensures a better and more comfortable customer experience. Chatbots are now based on state-of-the-art AI and natural language processing (NLP) models, with the ability to handle thousands of commands at the same time, high processing speed, high accuracy, satisfying the demand of businesses for an optimal customer service experience.
Voices
In addition to texts, with the purpose of being able to communicate with users more easily, current virtual assistants also support interaction via voice commands. They are sometimes named “voice assistants” accordingly. Voice assistants operate based on the following principle: Users only need to activate them using “wake-up words” (voice commands that require the assistant to turn itself on), then communicate with them as in a regular conversation. Not only can voice assistants handle commands on smart devices, but they can also be a companion which users can confide in or rely on for any piece of information.
However, the most prominent voice assistants, developed by technology giants like Apple Siri, Amazon Alexa, Google, etc., only support commands in English. This means many Vietnamese users find it difficult to communicate with their virtual assistants.
Recently, ViVi, a Vietnamese voice assistant, has been developed by Vingroup Corporation’s Institute of Big Data (VinBigdata) and introduced to the public for the first time. The next plan is to integrate the voice assistant into existing VinFast car models.
Experience ViVi: here.
Images
Some virtual assistants that are able to process user-input images have been in existence. Despite this, image data processing and the corresponding interaction method are still very much within development, and image-based virtual assistants are currently not a focus. However, in the future this can turn out to be a breakthrough.
Top virtual assistants in their field like Apple Siri, Google Assistant or Samsung Bixby are making much effort in assisting users in a diverse range of interaction methods. However, the smartest and most flexible method is still either via texts or via voice commands.
Platforms that support virtual assistants
Current virtual assistants are supported in many different types of devices and platforms to satisfy the diverse and frequent needs of users. The following are some of the most popular platforms:
Operating systems of smart devices
Integrated right into the operating system (OS) of smartphones, digital watches, computers, smart TVs, etc., virtual assistants have ceased to be a strange term for users. They are portable, can handle many instructions that concern both professional and private life and their appearance has truly made our lives easier and more convenient.
Functions of household devices
Recently, virtual assistants have also been integrated into speakers or other household appliances. For example, users can easily command the virtual assistant to turn on the music, adjust lighting in a room, change the speed of fans or the temperature of air conditioners, etc. This makes our house a “smart house” and brings about a comfortable, convenient customer experience.
Integration into messaging services of websites and social media
Text-based virtual assistants are often applied and integrated in this way. Organizations and businesses often use virtual assistants as a customer service employee which can respond to customer requests in real time. They can provide information about the business and collect customer information, which optimizes processes that are aimed at improving customer experience.
Capabilities of virtual assistants
Handling requests
Based on huge databases and state-of-the-art ML and AI algorithms, virtual assistants are capable of receiving many requests simultaneously and process them in real time. Virtual assistants analyze natural-language voice commands or texts from users, and then convert this data to another form and send it to the local station for processing. All of this can be done in a matter of milliseconds. They utilize pre-programmed and pre-trained knowledge to handle requests or reply to questions suitably. In recent times, based on NLP technologies, virtual assistants can now analyze natural-language, daily conversational language which might include dialect-exclusive words and phrases. This means that virtual assistants can extract information accurately to satisfy requests in a more flexible way.
Automatic Speech Recognition (ASR)
For voice-based virtual assistants, besides the ability to handle requests, speech recognition is also truly a breakthrough from the part of AI researchers. Virtual assistants listen to the voice of users, record it, and transform it into another form that can be understood and processed. Utilizing feature-based algorithms that take into account characteristics of voices like frequency, energy, or amplitude, voice assistants have obtained a high accuracy. Voice assistants also have text-to-speech (TTS) technology integrated so that they can converse with users more flexibly and accurately.
Locating and extracting information
Users often demand virtual assistants to perform such operations as turning on a song, updating new information or adjusting volume of a speaker. Those operations require virtual assistants to process a gigantic amount of data. They have to sign on to digital music platforms and then conduct searching to turn on the correct song, or go into news platforms to extract the most up-to-date information, or quickly access operation systems to change system parameters. All of the aforesaid operations can be processed in a fraction of a second.
On-the-fly learning from user behaviors
This is a highly rated ability of recently developed virtual assistants. They extract information based on customer behaviors to provide recommendations that are most suitable to their demands. The more users utilize the services of virtual assistants, the more flexible they become and the more capable they are of offering correct recommendations. For example, instead of having to instruct in a specific way such as “Turn on “Peaches” by Justin Bieber!”, users can now say “Turn on my favorite song!” or even more simply, “Turn on the music, please.”.
Automatic integration of applications
This is a newfound strength of virtual assistants that is being developed by researchers. A specific example: Virtual assistants can now help users note down the minutes of a meeting, extract a schedule that has just been agreed on in the meeting, and then access alarm clock software to set up a reminder. Users now do not need to manually write down or memorize the dates of the appointments, as virtual assistants have performed such tasks for them.
Applications of virtual assistants
Virtual assistants for personal users
Communicating with users, providing information, responding to queries and requests: Based on state-of-the-art AI and NLP technologies, current virtual assistants can easily research and process data based on user inputs. Thanks to the help of an ever-increasing database, virtual assistants now learn from an expanding dictionary and therefore are more flexible in dealing with natural-language commands.
Virtual assistants can answer user questions in a wide range of topics, whether it be about the weather, politics or medical information. Some virtual assistants can even capture and understand the portrait or emotion of users and converse with them with situation-specific lines.
Hand-free operations on smart devices: This is a highly popular function of voice-based virtual assistants. According to a May 2017 research conducted by Pew Research Center, nearly half of the adult population in the United States use smart devices or other appliances that have voice-based virtual assistants incorporated. Those assistants can assist users in making phone calls, sending text messages, opening software applications, giving directions to a location on a map, searching for information online, setting up an appointment on a pre-installed calendar app, etc. Users can have their requests fulfilled at very high speed without having to lay their fingers on any device.
Recently, voice-based virtual assistants have been integrated in many more devices and thus have become much more flexible. Besides successfully operating on smartphones, those assistants can control many household appliances: smart cars, smart speakers, TVs, air conditioners, etc. with new exciting features: unlocking houses and cars, turning on lighting, adjusting room temperature and speed of fans, changing television channels, decreasing volume and much more.
Virtual assistants for businesses
Boosting customer experience: Using manpower to take care of the requests of hundreds of customers logging onto a fanpage daily can easily lead to errors, and also cost customers a lot of time waiting for their queries to be dealt with. This leads to a significantly decreased customer experience, which in turn affects profits and more importantly, the brand of the business. Instead, businesses can opt for digital assistants, and more specifically, chatbots, to respond to their customers in real time with accurate, pre-programmed answers.
Changing production methods: Virtual assistants have also been a part of automation in factories. For example, in a laboratory setting where employees need to wear gloves and tight protective clothing to ensure safety and prevent spreading pollutants, they might have great difficulty in pressing buttons or typing on a keyboard. Voice assistants can change everything by automating all operations.
Changing working and management methods: For companies that integrate state-of-the-art technologies, many employees use smart speakers for different tasks in meetings. Voice assistants can search for information, write down notes, set up appointments and meetings in a calendar software, or even draw out to-do lists. This helps in saving time and forces everyone to be more concentrated on the meeting. In addition to this, some assistants are granted access to company’s internal data, which allows them to process and translate statistical data, automate operations and partially assist in text reports.
Some difficulties of virtual assistants
Privacy concerns
Privacy is a matter of concern for many users, especially for smart devices that are used on a daily basis. After being activated by wake-up words, voice assistants will start recording sounds from the surrounding environment. Such data will be sent to a local station for processing and a response will be returned. Therefore, users are often worried about the possibility of their private information being recorded or tracked. However, researchers have recently introduced new features that allow users to access saved recordings of voice assistants and make a decision as to keep or delete them.
Accuracy
Voice assistants are not perfect in understanding everything that users input. There will be some tasks that users need but the assistant has not been trained on yet. A factor that affects the accuracy of voice assistants is the distinctive feature of languages and dialects. Users come from different places, share different backgrounds, and therefore they might have a widely different set of vocabulary and a distinctive way to pronounce words. Voice assistants might fail to recognize words that are highly location-specific but have not been normalized and standardized.
Security
Even though virtual assistants send information to local stations with encrypted connections, there still exists risks of hacking and information leaking. In the beginning of 2018, some users of Amazon’s Echo reported that the assistant blurted out devilish laughter without any reason and concluded that somebody had infiltrated their speakers. Amazon conducted an investigation on the matter and announced that this was because Echo had heard words similar to “Smile, Alexa!” and therefore it started laughing.
Because voice assistants can recognize and respond to any voice from the surrounding environment, a stranger can check or even change information found in the devices of the original owner. Or perhaps even more dangerously, voices can be faked to control and manipulate devices.
Virtual assistants, along with their many breakthrough features, are predicted to be a smart companion with human beings in the future. A new era of breakthroughs in this field is opening, and no person or business should stand outside of this trend. Let’s hope for the development of smarter virtual assistants all over the world and even in Vietnam!
Virtual assistant is a technological solution that has been a focus of VinBigData’s research activities, due to its breakthrough features and high applicability. Recently, Vivi – voice assistants exclusive to the Vietnamese people – has been experimented on Vinfast car models with the accuracy of Vietnamese language recognition being 98% for common words. The voice recognition error rate is smaller than or equal 6% with Internet connectivity and 15% without Internet. ViVi supports more than 100 operations concerning information, services, entertainment and can be readily integrated into smart cars. Research about VinBigData’s smart voice assistant here.
