The only child gave birth to the children's voice robot market, but trust has already been overdrawn

Speech technology has brought a lot of good imagination to human life. It is expected that people can use voice technology to communicate with robots in any scene to provide convenience to humans, as described in the movie "She". However, in reality, the most likely to let voice technology fall, but a group that does not require high voice recognition.

For children aged 3-7, use speech recognition, semantic understanding and other technologies to communicate with children and provide services for companionship and early education. Such an application scenario sounds very promising. Especially in China, the company of the only child has always been a concern of the society.

Wu Yijian, CEO and founder of Shanghai Yuanqu Information Technology Co., Ltd., said in a recent interview with the news that early childhood education companion robots have different requirements for voice interaction technology than adults, and do not pursue very high recognition rates, but they need machines to understand. The way children interact is the just-in-time need for children to interact. Only in the process of market promotion, but because of the abuse of early market participants, overdraft consumer trust in such robots.

The uniqueness of children's voice interaction

Face-to-face communication is the best way for human communication, but it is not easy to make robots communicate with humans.

"First, you need to let the robot hear and recognize what you are saying; second, it has to understand what you are saying; finally, the robot has to output what you said, that is, communication." Wu Yijian told the news.

Wu Yijian graduated from the Junior College of China University of Science and Technology. He has been responsible for the research and development of voice technology and products in domestic institutions such as Keda Xunfei, Microsoft, and Shanda Innovation Institute. He founded the information technology company is committed to doing the best domestic early education companion robot. Following the launch of the first generation of desktop robots Xiaobai two years ago, the company has now launched the second generation of portable companion robots.

In a number of companies have engaged in the research of voice technology, but ultimately choose the children's market cut, Wu Yijian has his own reasons. First, children's need for voice interaction is just needed. Unlike home or adult interactive robots, children are not yet literate, so they can only communicate with the outside world through voice, which makes the children's companion robots just need. Second, relative to adult or general-purpose speech recognition, dialect problems do not need to be considered in children's voice interaction. This saves the startup company the energy and input to study dialects. Third, young children have relatively low requirements for the correct rate of speech recognition. Compared with GM or adult recognition rate of more than 90%, the recognition rate of young children only needs to reach 80% to meet many interactive requirements.

In addition to technical reasons, Wu Yijian has a little bit of selfishness in entering the early childhood robot market. "I used to watch my daughter often play iPad, the head is low, and facing the screen, I think it is not conducive to her development. So, I want to design a robot that can interact with her, but it will not let her indulge in it." Wu Yijian said.

The uniqueness of the child's voice interaction determines the particularity of the market, but the relatively low technical requirements do not mean that the market will make good products very easily.

"The pronunciation of a child is different from that of an adult. He may have a slurred speech and occasionally no grammar. But the child's voice interaction also has its own special place. For example, if the child says that I want to see Sun Wukong, the robot needs to understand the animation of the Journey to the West. In addition, you listen to Siri's voice is mechanical, but in fact, children's robots need to have emotions, robots need to bring emotions in the process of speaking." Wu Yijian said.

In Wu Yijian's view, understanding the interaction of young children has become the key to occupying the market. To do this step more critically is to collect enough data from children to interact with the robot, so that the robot knows the child better. Currently, the company wants its second-generation robots to get as much interactive data as possible.

The market has been overdrawn by early participants

The special nature of early childhood education and accompanying robots has made many companies' technology entry barriers lower, but on the other hand, it has also had a bad influence on the market.

澎湃News From the World Robotics Conference in previous years, it was discovered that many robot manufacturers usually use low-cost voice chips, add them to an iPad, and add a camera, which is called a voice interactive robot.

"Now the market for robots is very chaotic. The reading machine that is as small as one or two hundred is called a robot, and the toy that can make a toy is also called a robot. Before the consumer has really had a good experience, some participants in the industry will consume it. The trust of the people is overdrawn, over-marketing. This is actually negative for the whole industry. This is a very big challenge for companies that want to make products well and build user reputation, because consumers are easily affected by the previous experience. The stick killed the entire industry." Wu Yijian said.

In Wu Yijian's view, there is no threshold for the speech recognition engine in the current market, but it is very difficult to do it well. Especially in the voice interaction, there is no product that satisfies the consumer. Future competition may not be an algorithm, but more important to big data and interactive experiences. In terms of big data and application experience, China may have an advantage over foreign companies.

“I went to college very early and did a lot of research on speech, but I found that those studies did not produce much value in the end. Now, I hope to turn my previous research into one. Volkswagen's technology, this is what I hope the company can do in the future." Wu Yijian said.

Wu Jun, a natural language processing and search expert and Silicon Valley venture capitalist, once put forward his own point of view: a technology, if the application scene is particularly large, it is worth doing a good job, because it is enough for a lifetime. This sentence may also apply to the market for young children's companion robots.