Deepbrain Li Xuanfong: When the semantic skills store is applied to millions of levels, the voice-interactive age is truly mature | WARE 2017__ Voice interaction

July 15, in the "ware 2017 speech Intelligent Platform and application summit, which was produced by the hardware innovation community in Shenzhen Bay, Deepbrain co-founder and CMO Li Xuanfong published" The Development opportunity of the pioneering company in the field of voice interaction. " "As a keynote speech, we show the intelligent ecology with semantic skills as its core, which should be paid attention to and how to build the application framework."

Deepbrain's core team is one of the first teams in the country to focus on natural language processing technology (NLP), and after Apple released the Siri Smart Voice assistant in 2011, it launched the first Chinese voice helper based on smartphones, Smart 360, and accumulated tens users. The team in 2014 also made the first domestic voice intelligent speaker-small Tony-level speakers, than the introduction of Echo was six months earlier. Internet of things four landing scenes, voice interaction into new traffic entry

In his speech, Li Xuanfong said, deepbrain that smartphones are not a strong voice interaction scene. At present, the mobile phone as the main application of Siri, Cortana and other voice assistants, and did not highlight the product's intelligence, and then enter the Alexa in the smart speaker this pure voice interaction scene lit up.

The Deepbrain team was aware of the need for a semantic open platform early on. In 2013, when smart hardware was less hot, the Deepbrain team released a semantic open platform and made a voice version of the WiFi-infrared remote control. June 2014 issued a smart speaker, and in the public platform audio products to obtain excellent results. In addition to the speakers, Deepbrain later tried to begin to embed voice-interactive technology in hardware such as intelligent robots to try to make the scene of smart products.

Li Xuanfong that the internet of Things voice interaction has four landing scenes, are smart home, children, black and white power and car (including the front and rear mounted). Based on cost considerations, in a longer period of time, the terminal price of more than 100 yuan may be the first to voice.

With the development of the Internet of things, voice interaction will be the next new traffic entry. The existing forms of interaction, including keyboards, mice, and touch screens, are based on vision while also taking up both hands, which does not apply to vehicles that need to hold the steering wheel at all times, as well as home environments that are busy doing housework.

Voice interaction can easily liberate the hands, allowing users to focus on the work at hand while the voice of the task can be issued. It is based on such market analysis and consideration that Deepbrain launched the intelligent Cloud brain with the core of semantic skill ecology in May this year. for hardware manufacturers to provide five major capabilities, semantic skills ecology is the core

Li Xuanfong Introduction, Deepbrain can provide hardware manufacturers with five capabilities, namely, semantic skills store, AI man-machine conversation engine, robot functional components, fine data and home internet services. Among them, the semantic skills store is one of the key to building a semantic-skills ecological plan.

Before the product enters the market, first must investigate the market demand, locates the accurate product localization. For example: whether the product is a male or female user. Home or office scene. How to design the appearance. How to build content services. User operations and how to promote them.

How to retain users by grafting the intelligent service on the hardware, with the upgrading of software and hardware to form a virtuous business chain cycle, is the common demand of the traditional manufacturer AI transformation and upgrade.

Semantic skill ecology is the cornerstone of AI services, and developers in various areas of the industry have developed semantic skills applications that are more suitable for user expectations in this scenario. Li Xuanfong that current voice assistants don't really show openness and intelligence, such as how many features you ask Siri. People don't know because it's not a good way to render a skill.

Consumers do not know how many functions The Voice assistant has, and do not know what to ask questions will have accurate results, which issues are not specific feedback. This is why voice assistants are often in trouble: User expectations are high and results are not used. When one day more than 90% of the user's problem has a precise result, the Voice interactive intelligent era is really coming.

Li Xuanfong that the future semantic skills based on voice interaction must be able to reach tens of thousands of, hundreds of thousands of, or even millions of them, to make the voice-interactive operating system truly mature, and this process may take another 5-10 years.

Currently, Deepbrain's semantic skills platform has been deployed to thousands of developers, developing more than 1000 semantic skills based on smart home, audio, entertainment and child education, which is expected to reach 3000 by the end of the year. And how to apply NLP technology, build an open ecological platform that can help developers shorten development time, service voice interactive industry chain vendors, assist industry products and scene Landing, guide users to adjust the expectations to a reasonable range, is what Deepbrain is doing.

Editorial: About ZHE @ Shenzhen Bay

Original: Deepbrain Li Xuanfong: When the semantic skills store is applied to millions of levels, the age of voice interaction is truly mature | WARE 2017