Never thought that DingTalk was actually an AI platform

Temple Qiongqiong from the bottom of the recessed non-
qubit reports | Public number QbitAI

60 seconds of speech can be converted into text in 1 second. Even Guangpu and Trump can be accurately "translated" into Mandarin.

Open the document, click the voice shorthand, and you can record the content of the meeting in text without cracking the keyboard during the meeting.

This is the intelligent ability displayed by office software.

Well, it smells so good.

And these technologies seem to be familiar to me, who often writes cutting-edge AI papers.

So when I was curious and wanted to pick up the technology behind the application, I really found the relevant literature.

Right, there is INTERSPEECH on the left, SOTA on the right, and Ali’s technical signature Dharma Academy , the R&D staff is a big cow, and even the head of the Speech Laboratory of Dharma Academy Yan Zhijie personally participated...

Huh? That nail that makes people "shivering" has always been remembered as an application-oriented product. How can it be related to the technical-oriented Dharma Academy?

Asked again, sure enough.

Dingding himself also generously admitted that after integration into the Alibaba Cloud system, including Dharma Academy, Alibaba Cloud's latest and cutting-edge technologies will begin to accelerate.

Now, please call Dingdingwei-Alibaba AI and other cutting-edge technology user interface.

The productization position of cutting-edge technology?

Just talk about the small function of converting long speech to text .

In Dingding, you can send up to 5 minutes of voice, can also convert text in seconds, and support multiple dialects such as Sichuan dialect and Cantonese.

The mixing of Chinese and English does not affect the accuracy of text conversion.

Although the function is common, the real chapter can be seen in the details. And what is used behind it is the latest end-to-end speech recognition technology of Dharma Academy.

At the Yunqi Conference in September 2020, Dharma Academy announced technological breakthroughs in voice AI:

Launched E2E-ASR end-to-end speech recognition technology, based on the SAN-M network structure proposed by Dharma Academy, and a streaming end-to-end speech recognition framework based on SCAMA. While improving computing efficiency, it also improves The speech recognition error rate has been reduced by nearly 30%, refreshing the published SOTA of the online ASR (speech recognition) method.

More importantly, this technology can achieve a voice recognition effect close to the cloud on a mobile terminal.

Similarly, there is an AI translation function that supports 10 languages ​​such as Chinese, English, Japanese, and Vietnamese.

Not only for text conversations and documents, but with OCR technology, DingTalk can also recognize and translate text with one click for image files in chat.

During the epidemic, behind the DingTalk video conferencing function, there are also strong outputs from Alibaba Cloud and Dharma Academy.

You know, the number of DingTalk users in 2020 has doubled compared with 2019, breaking through 400 million .

To smoothly support the office collaboration and remote class needs of users of this magnitude, for the technical staff behind the video conference, it means an unprecedented high concurrency challenge. I still remember that during the epidemic, various online education and remote office platforms continued to collapse.

It not only requires sufficient servers and cloud computing resources as basic support, but also poses more difficult technical challenges to video codec algorithms and video conference architecture.

For example, the traditional video conferencing architecture uses a centralized architecture, which has natural disadvantages in large-scale deployment and elastic scaling.

But because Dingding relies on Alibaba Cloud's cloud computing and edge computing capabilities, and adopts a distributed microservice architecture, it can perform ultra-large-scale system computing resource scheduling and network resource scheduling. And can dynamically expand and contract according to the system load, and finally can maximize the shared use of system resources.

In addition, due to factors such as the popularization of 5G and the enhancement of user terminal network capabilities, the continuous increase in video traffic has also put forward higher requirements on the delay of the distribution network. Audio and video traffic needs to be distributed more intelligently. deal with.

It can be said that the test of audio and video for technicians is all-round.

Dingding audio and video has now connected to the basic technical strength of Alibaba Cloud, which can fully integrate the video codec algorithm, voice 3A processing, network QoS, AI transcribing, machine translation, AI noise reduction and other technologies provided by Dharma Academy. Comprehensively enhance the user's video conference experience.

Similar to the intelligent noise reduction of Dharma Academy’s Voice Laboratory, under 0db SNR, MOS (Mean Opinion Score) can still reach 3.5 points, and echo cancellation ERLE (Echo Return Loss Enhancement) can reach 52.2db, leading the industry.

When DingTalk was integrated into the Alibaba Cloud system, Zhang Jianfeng said that DingTalk would only be stronger.

Nowadays, I am not deceiving.

How can Dingding be stronger?

The recent major version upgrade of Dingding is an intuitive display.

This time, even the positioning is different: directly from the collaborative office platform to an enterprise collaborative office and application development platform.

The main changes are as follows:

  • Launched nails should take other low-code development tool that allows non-programmers users can quickly develop new applications.

  • Launched application connectors, which can connect DingTalk, DingTalk ecological applications, user-built applications, original IT systems, etc., to break information islands;

  • At the same time, through the 1300+ API interface, the underlying product capabilities are opened to customers as an application development platform to reduce costs and increase efficiency for enterprises' digital transformation.

……

May wish to combine specific cases to see what kind of ability this is.

Mengniu Group faced the suspension of offline bidding during the epidemic in early 2020. How to quickly move the bidding to online and resume normal work as soon as possible?

Mengniu has already moved its organizational structure to the cloud through DingTalk, and decided to build a bidding platform with Yidiao, combining DingTalk's group, video conference and other functions to complete supplier bidding and auditing online.

According to Zheng Jiong, IT Director of Mengniu Group:

It originally cost 1 million yuan to purchase an audit system, but now it is developed with almost zero cost.

At present, the company has built more than 100 suitable applications to replace a large number of IT systems purchased, effectively reducing the operating costs of the enterprise.

So it can be understood as the current Dingding:

The middle and back-end technologies are stronger, and the use of cutting-edge technologies such as Dharma Academy provides more support and guarantee for front-end applications.

The middle and back-end technology has further lowered the threshold for front-end technology development. Developers can develop suitable programs according to local conditions and become simpler, so Dingding can do more.

In other words, the newly launched low-code development products should be built, and the applications built are still cloud-native .

How to say? Since Yida is naturally built on Alibaba Cloud, the applications built with it have native Alibaba Cloud capabilities such as distributed computing, elastic capacity expansion, remote disaster recovery, CDN acceleration, and enterprise-level cloud security.

Moreover, Yida has componentized various cutting-edge technologies and basic technologies of Alibaba and Alibaba Cloud, so that each user can directly call Alibaba's OCR, data engine, DataV and other technologies and products.

The DingTalk used to be a single soldier providing weapons, but now it is mobilizing developers and the people.

Why is Dingding "renewed"?

The secret can be applied to the mass line: rely on the masses, mobilize the masses, come from the masses, and go to the masses.

However, before implementing this route, Dingding has preparatory preparations.

And this is Ali's "cloud nailing one" strategy.

Zhang Jianfeng, president of Alibaba Cloud Intelligence, said that in the cloud intelligence system, the industry applications of the uplink and the infrastructure of the downlink are nailed.

In the industry applications of the uplink, the low-code tools and connectors mentioned in the previous article are now embodied, making the creation, development, connection, and data exchange of industry applications easier.

How does the down-link infrastructure reflect? The analysis is to call Alibaba Cloud's underlying computing, network, storage services, and industrial solution capabilities, and to productize Alibaba Cloud's cutting-edge technologies and algorithms in the cloud, AI, and big data fields on DingTalk. Turning all kinds of cutting-edge technologies that sound out of reach for ordinary users is turning into reality and becoming products and tools that users can touch.

Therefore, DingTalk's new look is due to the initial success of the integrated cloud and nailing strategy. DingTalk is really becoming an integrated platform for Ali's technology.

Secondly, after all, Ji Ali's technical success is limited, so Ali has lowered the development threshold in practicing the "mass line", so that more scenario applications can bear fruit.

This is why the low-code issue is so noticeable in Dingding's upgrade.

If you look at the dots in a line, you can understand why DingTalk must be adjusted into the Alibaba Cloud system.

On the one hand, coordinated operations and unified leadership can focus on the development of elite and cutting-edge technological artillery.

On the other hand, Alibaba Cloud also has a suitable business application window for end customers and users, and various cutting-edge technologies have a more direct use position.

Precisely because of this, Dingding can now be said to have strong technical strength in the background to support.

Of course, looking at the previous judgment of Zhang Jianfeng, the helm of the cloud nail integration strategy, I will also be impressed with Ali's strategic layout and vision.

From a historical perspective, Zhang Jianfeng believes that the history of global software development is divided into three stages:

In the first stage, the IT infrastructure is a mainframe or a minicomputer. Enterprises purchase large-scale software systems to solve all problems, but the implementation costs are high, operation and maintenance costs are high, and it is difficult to develop again.

The second wave is the rise of SaaS software, such as Salesforce's CRM system. At this stage, the IT infrastructure is unified, but the software is provided by different vendors, and data islands are still formed between the software and the software.

The third stage is the goal of the integrated evolution of cloud nails. The main feature is the cloud-based capabilities, allowing enterprises to develop applications from integrated or SaaS-based software development in the past to low-code development, allowing enterprises and organizations to keep up with the trend of digital transformation at a lower cost.

Zhang Jianfeng also said that there will be many uncertainties in the next ten years, but there are also clear-the greatest certainty-the popularization of digital technology and the overall digital trend of the entire social economy and life.

So don't worry about Dingding "dominating" your study and work.

Because it will transform everything related to your digitization and intelligence.

- Ends -

This article is the original content of the NetEase News•NetEase Featured Content Incentive Program account [qubit]. Unauthorized reprinting is prohibited.

Join the AI ​​community and expand your network in the AI ​​industry

Qubit "AI Community" is recruiting! AI practitioners and friends who are concerned about the AI ​​industry are welcome to scan the code to join, and pay attention to the development of the artificial intelligence industry & technological progress with 50,000+ friends :

Qubit  QbitAI · headlines on the signing of

վ'ᴗ' ի Track new trends in AI technology and products

One-click three consecutive "Share", "Like" and "Looking"

The frontiers of science and technology are seeing each other~

Guess you like

Origin blog.csdn.net/QbitAI/article/details/112791047