2021年8月3日

The industrialization practice of Jingdong science and technology intelligent human-computer interaction technology

Yan Xi, an intelligent human-computer interaction platform independently developed by JD Technology Group, is one of the intelligent human-computer interaction systems widely used in the industry. It is an end-to-end intelligent customer service solution with high EQ, personalized and full link, which helps merchants reduce service costs by 65% and improve reception efficiency by 45%. “Yan Xi” does not serve JD itself. At present, it has been launched in retail, government affairs, finance, education, medical care and other scenes to help the digital transformation of the real economy.

Recently, jingdong group vice President, head of jingdong technology group intelligent customer service products division, jingdong artificial intelligence research institute, the executive vice President of the IEEE/CAAI Fellow xiao-dong he in China Association for Science and Technology of academic journal of science and technology leader, the journal published titled “the practice of the intelligent interactive digital age services and thinking” the signed article, from the perspective of technology development and industrial application, reveal jingdong intelligent human-computer interaction platform for everybody “rhinoceros” behind the technical capabilities.

The following is the full text of the signed article by He Xiaodong:

In recent years, China’s digital industry has developed rapidly, with active innovation and rapid growth in the Internet and related service industries. With the development of digital economy, data itself has become an important factor of production of digital economy.

Digital economy includes digital industrialization and industrial digitalization. Digital industrialization mainly refers to the information industry itself. Industrial digitalization is the transformation and upgrading of traditional industries by exponential word technology. According to China’s information and communications research institute issued a white paper “the economic development of China’s digital (2020), in 2019, the industrial added value of digital China is about 28.8 trillion yuan, which accounts for about 29% of gross domestic product (GDP), among them, China’s agriculture, industry, service industry added value proportion of digital economy, 8.2%, 19.5% and 37.8%, respectively, digital services accounted for accelerated growth. It can be seen that the digitalization of service industry has a large number of industrial opportunities and growth space.

The service sector covers a wide range of industries, from retail to finance to education. The service object of the service industry is people, which involves more basic technologies such as human-computer interaction. How to make computers understand people and human-computer cooperation to improve service efficiency? The service industry needs to improve the efficiency of the industry, reduce the cost and further improve the user experience through intelligent technology. Intelligent man-machine dialogue is a very important technology.

Intelligent human-computer interaction technology industry value is very large, issued by the State Council in 2017 of the new generation of artificial intelligence development planning, points out a new generation of key technology of artificial intelligence, the United States in 2018, the Internet and information technology research and development plan (NITRD) think the human-computer interaction is a core technology, invested more than $1.8 billion over the past two years to carry out the research. According to market research firm Gartner, human-computer interaction technology will support a trillion-dollar industry in 2022.

Human-machine interaction technology describes how to make people and computers communicate and interact with each other, mainly to improve the efficiency of man-machine collaboration. Looking back at the development history of human-computer interaction over the past 70 years, from punched paper tape to command-line terminal, to graphic interaction based on mouse and keyboard, and then to natural interaction like touch screen, this interaction process is the process of improving human-computer collaboration efficiency.

The era of intelligent interaction is characterized by the use of artificial intelligence and other technologies to enable machines to have situational knowledge and communicate with people through multi-modal and multi-round dialogues, so as to better understand people’s intentions and complete high-value tasks.

Each stage of development in the past 70 years has also brought about a new upgrade at the level of trillions of dollars, from the earliest representative enterprise IBM to Apple, Microsoft, and then to Apple, Google, Huawei, Samsung and other smart phone industries. In the era of intelligent human-computer interaction, new trillion-level industries can be predicted to take place.

There are plenty of opportunities and challenges in the intelligent interaction stage. Technically, we hope to build intelligent service robots to better serve human beings. On the one hand, we hope that robots can be smart enough to understand people’s intentions and have the necessary knowledge to truly complete valuable tasks for people. On the other hand, we hope that robots can understand people’s emotions and communicate and serve with empathy. From the perspective of industry, in the era of intelligent interaction, there will be a large number of intelligent machines to provide services for people, and services provided by intelligent machines will be everywhere. At the same time, it also poses a challenge to the technology, and strengthens the basic technology layer, including speech semantic analysis, affective intelligence analysis, large-scale language modeling, knowledge graph construction and reasoning decision-making intelligence, cross-language and cross-vision multimodal representation learning, etc. On the basis of these basic studies, a series of core competencies should be established, including language understanding, language content generation, multi-round human-computer interaction and dialogue, multi-modal information processing, etc. Only at the core competence level can a series of applications that can be perceived by users be established, such as intelligent customer service, intelligent marketing guide, intelligent interactive media, personalized intelligent assistant and a variety of intelligent hardware and other interactive intelligent platforms.

Breakthrough of key technology laid the basis of intelligent interactive aspects to build new applications, such as through several rounds of dialogue and decision-making under complex scene technology in-depth semantic understanding and emotion recognition so as to accurately grasp the intention of the people, through the study of the intelligent analysis of data for structured knowledge so that they can use the knowledge reasoning and auxiliary decision-making, through creative intelligence can let the machine to generate more attractive conversation thus able to multimodal optimization experience with the user interaction, and so on. Based on these technologies, JD has created a series of new human-computer interaction and collaboration tool systems to support all-channel, all-process and all-scene customer service intelligence. Take JD’s intelligent customer service system as an example. The whole system is a modular integrated service system based on a variety of advanced technologies. From the beginning of multi-channel consulting access, to the background human-machine collaboration, human-machine cooperation service scenario, complete the closed loop of service, including machine reception, machine assisted labor, seamless handoff between human and machine, and automatic docking with business process. In addition, the system also provides real-time service quality monitoring and service data analysis, as well as intelligent staff and business management, thus improving the overall efficiency and experience of the service.

There are many scenarios for intelligent customer service, such as intelligent interactive marketing guides. JD has developed an intelligent shopping guide system, JingXiaoZhi, which provides more than 130,000 merchants on the JD ecosystem platform with the ability of intelligent marketing shopping guide. The task of a marketing guide presents many new challenges compared to a simple response to an inquiry, such as explaining why the product is worth buying or what advantages it has while answering the inquiry. Where once only experienced salespeople could do this job well, now machines can do it, and even have the advantage of efficiently mastering large amounts of product knowledge and using big data to accurately understand user profiles. At present, good goods have been found on JD platform, and digital and intelligent marketing services have been provided for in-community purchase and store consultation to improve the service experience.

When it comes to human-computer interaction, people tend to think only of the interface, but to truly provide a closed-loop experience requires a full process automation in the background. For example, it is not only the recognition of users’ voice, but also needs to do a lot of things like automatic query, automatic operation, self-starting work order execution. Only by combining artificial intelligence and process automation, can service efficiency be improved. This scenario presents a new challenge for the technology, where digital robots can learn to evolve themselves in the process of automated service. In response to the intelligent interactive shopping guide scenario mentioned above, JD launched an Alpha⁃ In Sales project, we learn and evolve through self-learning and unsupervised learning and user feedback. For example, when guiding users, we should learn when to grasp users’ pain points and interest points and recommend products that meet users’ needs to users, so as to further improve the efficiency and experience of shopping guide.

Smart government services are also an important scenario. Combined with JD’s accumulation in the field of intelligent interactive RPA and data intelligence, JD has built a “brain for government affairs consultation” and built an intelligent center for urban consultation through the systematic deployment of “one platform + two systems +N applications”. One platform refers to the consulting brain platform, covering JD Technology’s profound accumulation of data (including government and public data and market data), computing power (including cloud computing capability and physical machine), algorithms (voice, semantic, visual and other cutting-edge artificial intelligence algorithms). The two systems cover data and intelligence. One is data management and analysis, including privacy management, data sensitivity management, and data value mining. Another is intelligence consulting and decision-making, how a knowledge based on data mining, can provide users with better consultation and decision-making, through artificial intelligence aided decision-making technology at the same time, can be hot issues, high-frequency events and evaluation analysis for managers to provide more decision aid and decision support; The following set up a number of subdivision modules. N Applications include people’s livelihood, government services, fine management, public security and other fields. Because the technology has been thoroughly tempered under the Jingdong scene, it has the characteristics of low threshold and easy deployment, and can be quickly landed in multiple scenes and widely used.

Enumerate a few concrete examples of the actual industrial landing of intelligent human-computer interaction technology. JD Intelligent Customer Service is one of the largest intelligent human-computer interaction systems actually deployed in the industry, providing a large number of intelligent customer service capabilities for the 400 million users of JD. Data showed that the monitoring efficiency of risk public opinion improved by 65%, and the accuracy rate reached 95%. Provide a series of business service robots to reduce business service costs by 65% and improve reception efficiency by 45%. During the “618 Shopping Festival” in 2020, 78 million person-times of consultation will be served, and the customer service based on crowd tactics will be transformed into intelligent reception mode, and the efficiency of the customer service center will be improved by 90%. Further the public service sector, in the fight against COVID – 19 outbreak, jingdong outbreak of rapid development of intelligent artificial intelligence department outside the system, help the grassroots community, enterprises and medical institutions to quickly complete screening for outbound outbreak cases, notifications, and information collection, review, etc., hundreds of times higher than that of manual telephone efficiency, help the government and enterprises to control the outbreak, overcome the outbreak.

At the same time, breakthroughs in intelligent human-computer interaction technology often open up new digital service scenarios. For example, the latest digital human technology, through the synthesis of voice, emotion, visual image and three-dimensional graphics modeling, personalized voice customization, image style customization, through voice and emotion drive facial actions and expressions, the voice interaction into a better virtual image interaction, so as to create a virtual host.

Another example is live streaming with goods, which is a new retail format rising rapidly since 2020. As a labor-intensive industry, live broadcasting and bringing goods requires manual live broadcasting, which leads to live broadcasting and promotion of only a small amount of goods. However, there are hundreds of millions of minimum inventory units (SKUs) in the market, which cannot be fully covered in the traditional human labor-oriented live broadcasting industry. However, intelligent interactive technology can help the scale of live broadcasting with goods industry. At present, Jingdong has automatically excavated selling points through product knowledge graph, automatically generated promotion copy, background music and virtual image for release, and equipped with consulting and shopping guide customer service, making the whole process automated. In this way, the form of live broadcasting can be quickly promoted to hundreds of millions of broadcast and commodity sales, so that this new marketing form becomes truly scalable.

In the field of service industry, there are a lot of space and growth opportunities for digital intelligence, and intelligent human-computer interaction technology will also bring more and better opportunities. In the next decade, we hope to be a technology-based enabler of the service industry together with ecological partners, so as to push the digitalization of service to a new level.

Leave a Reply

Your email address will not be published. Required fields are marked *

Scroll Up