Introduction:In 2014, Alibaba quietly launched the smart voice project. Six years later, it has grown into the first in China. IDC, an international authoritative research organization, released the semi annual research report on China’s AI cloud service market. It was found that the AI on Alibaba cloud performed well. Voice AI took the first place in the market share of intelligent voice and conversational AI, accounting for 44% and 57% respectively.
“As soon as possible, please take out your package in the smart cabinet.” For the busy city people, the express information on the mobile phone provides a lot of convenience.
What many people don’t know is that the calls made before the express delivery and the information of the delivery cabinet cannot be separated from Alibaba cloud voice AI.
In 2014, Alibaba quietly launched the smart voice project. Six years later, it has grown into the first in China. IDC, an international authoritative research organization, released the semi annual research report on China’s AI cloud service market. It was found that the AI on Alibaba cloud performed well. Voice AI took the first place in the market share of intelligent voice and conversational AI, accounting for 44% and 57% respectively.
From serving Ali’s economy to serving all industries, Ali voice has become a dark horse
If you have a sudden power failure in your home, you call the power repair phone. Before last year, the line was often busy and it was difficult to get through. However, the emergency repair telephone has suddenly become more people-friendly this year, and it can be connected as soon as it is dialed. This can not be separated from the credit of patch, the first virtual artificial intelligence distribution network dispatcher in China.
Last year, patch was launched in Hangzhou. Its brain stores hundreds of thousands of words of text materials, such as dispatching regulations, safety regulations, analysis reports, etc., as well as hundreds of TB (terabytes) of basic data such as equipment, personnel, power grid topology, and 5000 hours of voice data. It uses knowledge mapping technology to process and store these knowledge, forming its own judgment and understanding, and ultimately replacing manual implementation Distribution network dispatching. Patch can make up to 200 calls at the same time, can work 24 hours a day, and accurately complete the monitoring of massive data.
When patch finds the power grid fault, it will send out the fault warning at the first time, contact the relevant emergency repair experts by telephone, and accurately calculate the time and navigation path required by the emergency repair expert to arrive at the repair site.
Before patch took up his post, the distribution network dispatchers needed to connect more than 100 calls every day for more than 200 minutes to monitor 500 messages in real time. The number of incoming calls at the same time in the morning and evening peak hours was as high as 40 times, which was too late to connect. The on-site personnel needed to wait for a long time. The intelligent dispatcher greatly improves the dispatching efficiency, and the traditional power dispatching work can be easily completed by patch. Today, the waiting time for field personnel has been reduced to one minute.
Patch’s voice technology comes from the voice lab of Dharma Academy. Patch can understand the phone calls from the staff, organize the language to communicate with each other, and conduct multiple rounds of man-machine dialogue. He can also speak Hangzhou dialect and Xiaoshan dialect.
In the future, the electric brain like patch can replace more than 50% of human work in the production and command field of distribution network. Taking Hangzhou as an example, it can reduce more than 200 people such as distribution network dispatching, emergency repair command and customer service, saving more than 30 million yuan per year.
Ali voice AI serves nearly 1000 customer service center systems across the country, helping nearly 100 ecological partners with intelligent customer service solutions, and enabling nearly 100 million users across the country to experience the powerful capabilities of alivoice.
Ali’s voice technology comes from the Dharma Institute. In 2014, Ali established the predecessor of the Dharma academy, and its initial service object is the internal demands of Ali economy. “The place where voice technology started to show its skills is the customer service call center. Taobao customer service, Ali group, ant group, nailing, Gaode and other traditional customer service call centers have carried out voice intelligent transformation.” Yan Zhijie, a voice AI expert of Ali cloud Damo academy, told reporters that voice assisted the internal business of Ali economy and achieved great success. For example, the nail of the fire during the outbreak of the epidemic this year was also provided by the Dharma Academy. Users can see the text information in a moment when they are pinning their voice. One of the heroes behind this smooth voice to text experience is Dharma hall Advanced speech recognition technology.
Give the customer’s life back to the customer, Ali voice relies on two magic tricks to find the market key
How do you get the voices of Chen qiao’en, Xu Weizhou, Zhu Zhengting and Doraemon in your family? It turns out that Ali phonetic AI “changed” after learning the recording corpus of stars within half an hour.
As long as a star has a small amount of voice corpus, Ali voice AI can learn perfectly. If ordinary people want to customize a voice AI with their own voice, as long as they record 20 sentences according to the requirements, they will automatically generate their own accompanying voice.
Since 2017, Alibaba’s intelligent AI technology has been exported. “We have a slogan that Dharma institute technology goes to the cloud with zero time difference. All good voice technologies used by Alibaba will be provided to all customers on Alibaba cloud in the shortest possible time through the cloud. Zero time difference represents an attitude. This technology does not mean that only Ali can use it, but also everyone can use it. In addition, we will focus on how this technology can be commercialized and generate customer value. ” Yan Zhijie said that unlike the traditional voice technology manufacturer’s “self production and self marketing” mode, Alibaba cloud has adopted a new service mode of integrated and self-learning, which has rapidly opened up the market.
The original voice AI, which mainly serves Ali’s economy, was a little uncomfortable at the beginning of marketing. “Originally, when working in Ali, all departments could make up for each other’s positions. However, there is no possibility of filling positions for customers in other industries. ” Yan Zhijie said that in the process of exploration, they quickly adjusted their playing methods.
Taking the court, one of the subdivided application scenarios of voice AI, as an example, the basic task of traditional voice AI manufacturers is to transform the whole court trial process into words and become structured, which is conducive to the application of subsequent judgments. Ali’s intelligent AI wanted to do the same at first, but they soon found out the problem.
“First of all, there are many provinces in China, and there are some problems with accents and dialects. At the same time, there are many non AI elements in this scene. How can we make a judge handling system, a court file management system, and an application system that displays the contents on the screen when a court session is held. We created the integrated pattern. In short, we only do the voice AI content we are good at, and other non AI content will be handed over to some head integrators in the judicial field, such as allowing companies like Huayu and Yunjia to integrate us, and then make a comprehensive court application system. ” ‘This asset light model quickly opened up the market, ‘Mr. Yan said.
Ali’s voice capability covers more than 40 scenes of government affairs. Among them, intelligent court speech recognition covers more than 8000 offline courts in 20 provinces of China, with a coverage rate of nearly 50%. Internet court trials cover more than 15000 online courts, covering more than 90%. Ali and more than 20 government ecological partners have reached cooperation. At present, among the three scenarios of call center, telecom operator and court trial, Ali voice AI has the highest customer recognition.
Since 2017, Ali voice began to upgrade its self-learning. “We found that in addition to the basic common model out of the box, many customers also need to make personalized customization based on industry data and knowledge. All of a sudden, we find that we just rely on ourselves to do it, which is not replicable and promotional. The reason is that we can not lay down so much manpower and material resources in every field. At the same time, we are not the most professional. To change a way of thinking, we can give voice AI customization ability to practitioners in the industry by launching self-learning products. Without much professional knowledge in the field of voice, we can easily use our self-learning products and infuse data and knowledge in the industry in a safe environment, so as to achieve the world’s top voice interaction in his industry The effect. With this self-learning ability, it is equivalent to fully releasing productivity. ” Yan Zhijie said that Alibaba cloud launched the voice self-learning platform in 2017, opening up the AI customization ability, teaching people to fish, and helping users customize voice AI independently. At present, Alibaba cloud’s customers and partners have developed more than 30000 models on the platform.
Alibaba cloud’s voice self-learning platform provides a set of customized training process for acoustic model and linguistic model, which allows users to infuse industry data in a secure environment, and quickly and conveniently customize their own voice model without knowing the voice language algorithm. In the China Mobile project, based on the Ali voice self-learning platform, the partners used only two weeks and dozens of hours of data to optimize the recognition rate of Hubei and Fujian provinces to more than 92%; in the Hangzhou virtual artificial intelligence distribution network dispatcher project, the partners used the self-learning platform to successfully improve the recognition rate from 76% to 93%; during the epidemic period, the intelligent epidemic machines in Hubei Province When people use Ali voice AI self-learning platform, the recognition rate of Hubei accent voice has increased from 62.5% to 94.4%; a partner has built a new Russian and Arabic speech recognition model from scratch within a month using the self-learning platform, with recognition rates of more than 85%.
“In the traditional voice technology service mode, if customers have voice optimization needs, they need to polish with the technical side, and even need to hand over the industry data to the voice technology provider. With the self-learning platform, users can build their own models and really control their lives in their own hands. ” Yan Zhijie said that Alibaba cloud will not touch customer data, and customer data privacy is absolutely safe.
At present, Ali voice AI has more than 50000 customers, including internal customers of Ali economy, such as Taobao customer service, tmall genie, Gaode map, rookie logistics assistant, etc. Outside the Ali economy, Alibaba cloud voice AI technology has been applied to China Merchants Bank, Guangfa bank, Zhejiang high court, China Mobile, CCTV, Huayu, byte skipping, Haier, Konka, Didi, Sina Weibo, funny headlines, Haidilao, HP, vipkid, Shanghai Metro, xiaoi robot, lilac doctor, Himalaya, palm reading and other customers, covering education Finance, Internet, home appliances, travel, media, transportation, catering, communication, medical and other industries.
**This year, China’s cloud based AI market is nearly $2 billion, with Alibaba cloud accounting for 44% of the total
According to the semi annual research report on China’s AI cloud service market released by international authoritative research organization IDC, AI on Alibaba cloud has performed well, ranking first in six dimensions in the three fields of intelligent voice, conversational AI and machine learning, ranking first in the market share of the three fields with a market share of 44%, 57% and 29%, surpassing cloud service providers such as Baidu cloud, Tencent cloud, Huawei cloud, AWS and Microsoft azure.
The IDC report investigates the cloud AI services of major cloud manufacturers in China, covering six categories: human face, image and video, ASR & TTS, conversational AI, NLP and machine learning. Ali AI won the first place in the number of products, market share and API calls in the field of intelligent voice; the first in market share and API calls in dialogue AI; and the first in machine learning.
On February 28, Alibaba voice AI technology was successfully selected as “the top ten global breakthrough technologies” in mit2012. MIT believes that Alibaba already has better AI voice technology than Google, which can complete complex human dialogue functions and even understand human potential intentions. Alibaba is also the only Chinese technology company on the list.
“Generally, a five-year compound growth rate of an industry reaches 50% or 60%, which is already a very high potential market. The blue cloud market is absolutely a blue ocean. ” According to IDC analyst Lu Yanxia, the growth rate of China’s AI cloud service market is 93.6% from 2018 to 2024, which is a very high compound growth rate.
Lu Yanxia said that in terms of voice service providers on cloud, Alibaba occupies a dominant position. In 2019, Alibaba cloud’s voice service will account for about 44% of the market share, nearly half of the market. “In fact, I didn’t even realize that Alibaba cloud can do so well in voice. We have indeed seen that Alibaba cloud has gone very fast in terms of ecological partners in recent years, and has gathered many partners. ” Lu Yanxia said that in the smart cloud voice service market, cloud service providers represented by Alibaba cloud occupy an increasingly important position, and may even surpass some voice technology manufacturers who originally deployed locally in the future. In the short term, the mainstream trend in AI is still private deployment. However, the wave of cloud services and hybrid clouds will also drive the rapid growth of AI cloud service market. With the upgrading of technology, in the next three to five years, the AI products that users have deployed today may also be replaced by a new generation of more intelligent products.
For the next smart cloud voice market, Lu Yanxia believes that from a technical point of view, the whole AI is facing many technological breakthroughs, such as face recognition, human body recognition, voice interaction in noisy environment, etc., and there will be many technological breakthroughs in the application field in the future. In terms of application, only about 20% of the scenarios use AI at present, and there are many application scenarios to be implemented in the future. In addition to consumer products such as courtroom, living room and call center, AI intelligent voice will be gradually implemented in the fields of conference services, medical records to text, pan industry and other fields in the future.
“You may as well go to see which scenes are not available or not well done by today’s technology, and they are of high value. These scenes will be gradually unlocked with the development of technology.” Yan Zhijie said that Ali voice team has a slogan called “ubiquitous voice interactive intelligence”. Their dream is that in the future, whenever and wherever, whether at home, in the office, in public space, or in the car, one day there will be contact points or entrances for voice interaction, which can interact with people.
Transferred from: https://www.thehour.cn/news/385576.html
Zhang Yunshan, an hour news reporter of Qianjiang Evening News