36Kr learned that "Zhujian Intelligent" in the field of natural language processing has received a US$30 million Series B financing. This round was led by China Development Financial Holdings and followed by Cathay Financial Holdings, and old shareholders Ecovacs and Shangshu Investment. Light Source Capital serves as the exclusive financial advisor. The funds will be used for technology research and development and commercial market expansion.
Zhujian Intelligence (Emotibot) is an AI startup company that 36Kr continues to pay attention to and report. It was founded in 2015 by Jian Renxian, former vice president of the Microsoft (Asia) Internet Engineering Academy. It focuses on multimodal emotion recognition and computing . It has previously launched a "Xiaoying" robot for the C-end.
In recent years, voice interaction and emotional recognition are no longer new. Many large companies and startups have entered smart speakers and chat robots, but most of them still use sound and text to identify them. Few people touch on multimodal emotional recognition. Since its inception, Zhujian Intelligent has integrated text, voice, image and other information for identification and calculation, thereby improving machine perception capabilities and establishing and optimizing multimodal human-computer interaction technology.
For example, Zhujian Intelligent once cooperated with the Sharp flagship store. When customers approach the product, the camera installed near the product can collect user's expressions, eyes, and body data. Then, the emotional recognition robot in Zhujian will "understand" user behavior and have voice conversations with the user.
So how do this be done? This mainly comes from Zhujian Intelligent multimodal dataset and high-precision model algorithm .
Currently, few people are exposed to multimodal emotion recognition computing worldwide, one of the reasons is the lack of relevant data sets. In addition, based on the complexity of Chinese itself, training data sets and model algorithms for English cannot be completely migrated. Therefore, Zhujian Intelligent uses self-manufacturing and cross-domain expert team tagging methods to build a multimodal data set that can be combined with self-developed model algorithms.
Look at the model algorithm again. Zhujian Intelligence combines symbolism with neural networks, and realizes algorithm training under the premise of small data volume by using computer symbols to simulate human cognitive processes. According to the company's founder Jian Renxian, in the absence of training data, the algorithm accuracy of Zhujian Intelligent can reach more than 80%; in the case of training data, the model accuracy can reach more than 93% in a short period of time through the reinforcement learning mechanism, and it can be fully commercially available.
It is reported that currently can realize cross-field industry intention and emotion analysis. In the understanding of intentions across more than 3,000 fields, the average accuracy rate can reach more than 95%, among which the analysis of facial vision, emotion, expression, and behavior can reach more than 90% accuracy rate .
After three years of technical accumulation, Zhujian Intelligent currently has two major product directions: one is type brain dialogue robot , which mainly includes virtual customer assistant (VCA), virtual enterprise assistant (VEA), intelligent customer contact center (AICC, which has supported multiple rounds of conversations and can reach Google Assistant level), virtual personal assistant (VPA, such as Xiaoying robot), etc.; the other is multi-modal emotion recognition system , which is used in emotion recognition, face expression recognition, impression analysis, advertising effect analysis, call center quality inspection, classroom emotion analysis, etc.
It is worth mentioning that Zhujian Intelligent launched the AI intelligent platform Bot Factory™️ this year, so companies can rely on Zhujian’s dialogue engine to build their own business chat robots. The platform provides enterprises with full-stack AI solutions, assisting the implementation of technology in private cloud + SaaS to ensure data security. Currently, the charges are mainly for annual service fees, and may be charged according to authorization in the future.
In addition to Zhujian Intelligence, domestic startups in the NLP field also include Triangle, Moran Cognition, Ain Interaction, etc. The main implementation directions of such startups are mainly around intelligent customer service and intelligent hardware (speakers, robots).
Among them, smart hardware is mainly used to cooperate with speaker or mobile phone manufacturers, such as Triangle, which cooperates with Vivo, Oppo and others to connect to its own SDK. Zhujian Intelligence in hardware implementation includes: providing bamboo service robot interactive brain for UBL Cruzer, providing bamboo dialogue engine for a leading domestic manufacturer, and providing bamboo related technologies for Changhong TV.
Intelligent customer service is Zhujian Intelligent's main direction . Jian Renxian told 36Kr: Taking the out-of-call scenario as an example, it includes more than 200 subdivided scenarios and more than 50% will be replaced by machines in the future. Zhujian Intelligent currently has a considerable share in this market. It is reported that around intelligent customer service, Zhujian Intelligent has reached cooperation with leading companies such as Bank of Communications, China Merchants Bank , Minsheng Bank , Ping An Group, Pacific Insurance, China Life Insurance, Evergrande Real Estate Group, Vipshop, Ecovacs, Industrial Securities , SAIC, China Travel Transport, China General Nuclear Corporation, China Mobile, Foxconn and other leading companies.
When selecting customers, Zhujian Intelligent tends to cooperate with the leading in the industry. Jian Renxian said: company cooperates with cloud platforms, traditional integrators and solution providers to go deep into each industry and introduce experts from various industries to train the algorithm model of Zhujian Intelligent, so as to provide the industry with accurate customized services and copy this to each vertical field.
At the commercial level, smart customer service is currently the main source of revenue for Zhujian Intelligent. Jian Renxian told 36Kr: 2018 is the first year of productization and commercialization of Zhujian Intelligent. It is estimated that revenue will grow several times next year and achieve break-even. As for the
team, Zhujian Intelligent has gathered many experts in the field of NLP and computer linguistics. Jian Renxian, founder and CEO of , was the vice president of Microsoft Engineering Academy, responsible for the Asian Bing search and the development of Microsoft Xiaobing and Cortana. Most of the members of the team come from Microsoft, Baidu , Alibaba , Tencent, Amazon , IBM, Google and other companies, and have established cooperation with MIT Artificial Intelligence Laboratory, MIT Media Lab and scientists in this field. The person who proposed the concept of emotional computing, Professor Rosalind Picard of MIT Media Lab and his team also had in-depth cooperation with Zhujian Intelligent.
Regarding the investment logic of this round, the leading investor, China Development Financial Holdings, said:
"Artificial intelligence is an important potential field we are optimistic about. Semantic understanding is the entrance to connect human beings with the future smart life, and it is also the key for enterprises to enter the industrial Internet revolution. Zhujian Intelligent adheres to technological advantages, accumulation of massive data and polishing and expansion of business scenarios, provides full-stack AI solutions to assist enterprises in intelligent transformation, and greatly improves commercial implementation and cross-industry capabilities through the AI platform Bot Factory™️, meets users from all aspects of life, and creates commercial value for corporate customers."
The author believes that for most NLP companies, intelligent customer service is the main source of income, and leading customers in the fields of finance, education, operators, etc. are also the main targets for everyone to seize. For NLP startups, it is feasible to grasp category differentiation and customize algorithm models in segmented fields, but over time, it is easy to become a case-making company. How to turn customization into standardization and become a replicable module in each vertical industry may be the key to increasing the volume. In addition, startups have already built open AI platforms like Bot Factory™️ to lower the threshold for enterprise-level users to construct chat robots. In addition to competing in technology and performance, perhaps more effort may be needed in pricing strategies.
————
I am Zheng Yin, a founding student of 36Kr Superman Academy. I pay attention to AI, AR/VR, Internet of Things, edge computing, and social networking. Please add WeChat zhengbaobao097825 for project communication and indicate the company, position, and name.
Superman Academy believes that early ecology promotes social change, and its mission is to cultivate leaders and talents for early ecology. For details, please click 36Kr wants to build a "Superman Academy" for young people entering the venture capital circle. Do you come to
