What Is It, How Does It Work and Why You Should Care?
One of the most exciting things about first dates is the chance to get to know someone. With most people attached to their smartphones, it’s getting harder to meet people in a traditional setting. For years now, dating apps have been using algorithms to help people find love, but what if finding love meant divulging deeply personal details … to a chatbot ?
That’s exactly what a group of young adults are doing in Japan to help break the ice in social settings and find a desirable match. The technology stack that makes this possible is conversational AI, fortified by humans’ willingness to give up their data to find love — but it doesn’t stop there.
Results from a recent In 2018, report by Voicebot show that voice technology adoption is on the rise across multiple devices. 53% of U.S. adults actively used a voice assistant each month. The number of U.S. households with smart speaker devices grew 78% in the past year. As the CTO of a company that provides search and conversational AI platforms, I know that to convert these users to customers, businesses and brands must establish a voice strategy that acclimates to the new ways users are interacting with the world around them.
Understanding Conversational AI and the Advantages of Its Technology
Conversational AI is a technology that allows users to use their voice to have conversations with applications, devices and computer interfaces. Put another way, it is what allows us to use natural language to interact with intelligent assistants, chatbots and smart speakers.
Through conversational wake words that instantly deliver information and services on command, voice-driven interactions provide deeper insights into users’ intent and moods by recognizing behavioral patterns and preferences.
The origins of conversational AI began with chatbots built for specific tasks within a limited domain or scope to provide basic responses. Since then, it has expanded to provide an abundance of “skills” tailored to produce a richer user experience framed by interactions. Future iterations of conversational AI will provide personalized assistants that both serve and predict users’ needs. Its greatest strength will reside in its ability to engage in human like discussions across various scenarios such as finding the right gift, offering fashion advice, recommending new restaurants, or perhaps even simulating a human relationship, as depicted in the film .
A Simplified View of How Conversational AI Works
Conversational AI combines powerful technologies such as machine learning, speech to text, user authentication, natural language understanding, intent and domain prediction, and text to speech. Here’s an overview of these key components:
ASR/NLU: Automatic speech recognition (ASR), or speech to text, transforms voice to text. The natural language understanding (NLU) module, using natural language processing (NLP), extracts entities and additional semantic information. The result feeds the system with inputs represented in a structured format with semantic labels.
Dialogue control module: This handles the management of the dialogue, tracks its context and decides the pragmatic adaptations needed to make the conversation realistic. This component needs to excel at dealing with human utterances, external noises, multiple users and multiple conversations (to name a few).
Task prediction module: This determines the intention of the user (to buy something) and the associated domain (groceries), filling in the values of the attributes related to the task (specific supermarket, delivery date, etc.).
Dialogue experts: Each expert is associated with a specific task/domain and generates a structured representation of the next answer in the conversation. Using the conversation status, each expert outputs the actions related to the task at hand when appropriate. For example, this could be managing a home sensor or sending a buying request to an e-commerce platform.
NLG/TTS: The natural language generation (NLG) phase generates the answer text that will be translated to voice by the text-to-speech (TTS) software.
Most of these modules are based on different flavors of machine learning and manage/interact with two databases:
- User model and context state: This database maintains the knowledge of the system for each user, including interests, conversation history and more, as well as the current state of the conversation and its context — which is key for disambiguating domains or named entities when needed.
- Knowledge base: This is essentially an ontology (i.e., a set of concepts and their relations) that helps to determine the right task and domain as well as answer questions.
Why Business Leaders Should Pay Attention: Planning Your Conversational AI Strategy
A recent Conversational Commerce study by Capgemini predicts that three years from now, 40% of consumers will use a voice assistant instead of a mobile app or website. To meet this demand, the global conversational AI software platform market space is set to expand at an impressive CAGR of +31% from 2019 through 2025. These platforms facilitate the development of cognitively enabled applications, including intelligent assistants, leveraging a wide range of both structured and unstructured data sets.
The evolutionary rise of conversational interfaces will impact industries on multiple levels, so consider the following when planning a strategy:
1. Start with a clear and focused use case. Identify aspects of your business that benefit from conversational AI and deliver the highest value to your users.
2. Decide whether to partner with a major player or use a platform to build your own conversational interface. As you consider a near- and long-term strategies, flexibility in how and what you can build, along with who owns the data, will help you decide.
3. Focus on opportunities to personalize and add value. Users consent to use their personal information in direct correlation to the value they receive by doing so. Create experiences that are worthy of this exchange and keep users engaged.
As conversational AI evolves, there’s no doubt that both consumers and brands will benefit from more intelligent touch points enabled by this technology. Though your product may not be able to promise the value of a potential partner or soulmate, it can surely offer your customers a reason to fall in love with your brand.
Originally published at https://www.forbes.com on April 16, 2019.