The “National Standardization Development Outline” proposes to ” promote the transformation of standardization work to digitization, networking and intelligence “. Tian Shihong, deputy director of the State Administration for Market Regulation and director of the National Standards Committee, proposed in the report of the National Standardization Work Conference to “actively promote the work of machine-readable standards and explore new mechanisms for national standards management under digital conditions.” development trend, and strengthen the basic theoretical reserve of standardization.” Standard digitization has become one of the important contents of standardization work in the future.
Standard digitization conforms to the development of the era of digitization, networking and intelligence
The 2022 Government Work Report proposes to accelerate the integration of digital technology and the real economy. This integration is manifested in many aspects, including how standardization integrates with digital technology. Standard digitization refers to the process of using digital technology to empower the standard itself and the entire life cycle process, so that the rules and characteristics carried by the standard can be read, transmitted and used through digital devices. The digitization of standards is proposed to better realize the formulation of standards and the promotion, publicity and implementation of standards with the help of a new generation of digital technology. The digitization of standards includes two aspects, one is the digitization of standard expressions, and the other is the digitization of standardization methods, which promotes the development of standardization work through digital technology.
Standard digitization is not only in China, but also internationally. The International Organization for Standardization ISO gives a definition called SMART for standard digitization, that is , machine-readable, usable, understandable, and interpretable standards can be realized without human participation. This should be a milestone in the development of standardization work. The transformation from the simple textual expression in the past to the current machine-readable method is not only possible due to the development of digital technology, but also a need for social development. Digitalization has penetrated into all aspects of the real economy and industry, and is used in many application scenarios. Machines have been replaced by people, and it will be the general trend of social development for machines to understand standards.
Let’s take the video shot by the smart city camera as an example. Hundreds of thousands of traffic surveillance cameras on the road are connected to the city’s traffic control center, but the video wall in the monitoring room of the traffic control center cannot fit so many video screens. Now it is basically displayed by road time-sharing, for example, there is one road per minute, and there are many cameras on a road, and relying on people to stare at it is very inefficient. Now we hope to synthesize all the surveillance videos on the road into a city-wide traffic dynamic video through artificial intelligence technology, which can significantly improve the efficiency of human monitoring. There is also a method that does not require synthesis, and directly reads these scattered videos through a computer. According to the labels of the spatio-temporal positions of monitoring points, the machine replaces the human to obtain a global understanding. This kind of traffic surveillance video in the future is not for people to watch. When people are used as the viewing object, the frame rate and resolution of the video should be considered to correspond to the resolution capability of the human eye. The video coding algorithm standard of machine vision is followed by people. The standards of reading will be different, so we have to consider that from human reading to machine reading, not only the way has changed, but even the content of the standard and even some of the connotations, regulations, and indicators in it will be different.
We are now entering a new era of information technology, and the entire social economy is transforming and developing towards digitization, networking, and intelligence, and standardization reflects the digitization of standards.
What standard digitization can bring us
First, the first is the change in the standard presentation form. We know that standards were basically texts in the past, so in the future, to make them readable by machines, standard texts will need to change. Machine-readable is first of all machine-retrievable, that is, the standard can be found and found from the machine. Then there is machine readable, that is, the standard content can be read out. Further development to the comprehension of the standard, the machine can understand the connotation of the standard, can interact, and finally can execute it. Machine-retrievable is relatively easy, machine-readable is difficult.
The technological development of artificial intelligence can change Chinese into English and English into Chinese through semantic understanding. What is the difficulty of machine readable? First of all, the standard is professional. The models and algorithms of artificial intelligence can only be optimized after a large amount of data training according to the processing tasks. This kind of training cannot cover all majors, not to mention that there are professional thresholds for people to read standards. Secondly, the standard contains unstructured data such as formulas, charts, and curves. It is generally difficult for computers to understand unstructured content. The visualization and readability of standards in these aspects is not simple, and a lot of work needs to be done.
How to make it machine readable can start with a standard description language. Now some people use the language of XML to describe standards. XML is an extensible markup language, a computer language that can describe data objects, and it contains an explanation of the data objects more than the general data language. It reduces the difficulty of computer understanding of data.
To be machine-readable, programs can also be introduced to describe standards. For example, some algorithms can directly become standards. This introduces another aspect of the digital representation of standards—open source software and the application of open source software. The standard digitization mentioned in the “National Standardization Development Outline” involves two aspects: machine-readable and open-source software. Open source software is now used more in the open source community. It is a kind of software, or it describes a technical specification as an algorithm, and the computer is easy to execute. Obviously, open source is also a form of standard digitization. However, compliance with standards expressed as open source software is limited.
The digitization of standards is not only in the form of standards, but also in the digitization of standardized methods, allowing digitization to run through the entire life cycle of standardization, facilitating the formulation, implementation, and promotion of standards. Start by speeding up the process of standards development and revision through open source crowdsourcing. In the past, we used to use the method of meeting and letter review to formulate standards. Now we can use the method of open source crowdsourcing to establish an “online collaborative writing” platform to allow more people to participate in the process of standard formulation, and to carry out collaborative development of standards in a digital environment. Ability to form deliverables in weeks, significantly reducing standards development cycles. According to the conventional method, the standard is revised only once a few years, and open source can iterate quickly, dynamically update, and speed up the formulation and revision of the standard. There will be some management system issues involved here. Whether or not national standards can allow this in the future still needs to be studied. However, group standards can be explored, and everyone can quickly participate in the open source platform. As long as they agree, they can be updated. This applies to emerging or A standard in a rapidly evolving technology field.
Second, it is easy to find the background of standard-setting and its connection with intellectual property rights. Through artificial intelligence and big data technology, we can quickly compare international standards, national standards, and group standards, find out the reference relationship between standards, and find out which contents are consistent and which are inconsistent. How much is the standard mining rate, and what is the difference in the scope of application? It can also be found in a series of technical standards such as terminology, technical system, specifications, test methods, product requirements, etc. Whether there are any contradictions or needs to be improved in order to express more rigorously. The digitization of standards also facilitates the linking of standards to intellectual property, including easy retrieval of claims made by interested parties on intellectual property rights during the standard development process, and whether patent holders are committed to FRAND (fair, reasonable, and non-discriminatory) principles for patent implementers Licensing, or whether the relevant intellectual property agency has authorized the patent, can not only know how many standard essential patents (SEPs) a standard involves, but also identify whether a patent is a SEP. In short, standard digitization combined with a new generation of information technology can well clarify some details of the standard.
Third, assist in the supervision of the implementation of standards. Because the standard is digital, regulators can find out whether it complies with the standard from the behavior data of the implementation of the standard, and it is easy to check the compliance of the implementation of the standard. In the past, it was difficult for government departments to read so many standards, but if there are traces of data on the things we do, we can compare them with the standards to find out whether the work meets the standards, and we can do better. The governance of society, which is also a benefit that standard digitization hopes to achieve.
Fourth, fast retrieval of standard knowledge. A standard text is often very long. In an implementation project, only a certain point or aspect of the standard may be used, not necessarily all the standards, but the implementer needs to read the entire standard from beginning to end. Some The work involves multiple standards, and it may be difficult to read all relevant standards. Why is it that some work does not meet the standards well? In fact, I feel that there are too many standards, and I don’t know which standard this matter is related to, so the implementation may not meet the regulations. Now it can be intelligently pushed through machine readable without downloading and reading a large amount of standard text, realizing fast knowledge retrieval, and even compiling directly into work instructions or rules through the machine, which can accurately guide the implementation. The digitization of standards is not only beneficial to the standard setting department and drafting unit such as the Standardization Technical Committee, but also to the standard supervision department such as the National Standardization Administration, and it is also helpful to the departments we implement.
Technical and institutional challenges
The digitization of standards is a milestone in the history of standardization development since the Industrial Revolution. It is the main feature of standardization in the digital economy era. It presents a rare opportunity for innovation in standardization work and promotes the reform of standardization methods and management systems, but it also brings unprecedented challenges.
One is the challenge of standardized management. The digitization of standards itself is not entirely technical work. There are many technical things to do here, but corresponding reforms in management are also required. It is necessary to re-examine the standard formulation and revision process and approval management authority. Our original standard regulations are all for people to read and people to implement. Although each person may have different interpretations when reading the standard, each standard will indicate the ownership of the right of interpretation, and the implementer of the standard will apply for a problem when it encounters a problem. explain. Now we turn to the machine to read. The algorithms executed by the machine are compiled by humans, and different machines may have different understandings. The ownership of the right of interpretation has not changed. Will it recognize that the standard clauses may have ambiguity of understanding and take the initiative to ask for interpretation? In the era of standard machine readability, it cannot be handed over to machines and still requires human intervention. When and how human intervention should be involved, management needs to be clarified. In addition, according to the current management regulations, the group standards organization needs to have the legal personality registered with the Ministry of Civil Affairs, while the open source community responsible for formulating open source software standards does not have such qualifications. The reform of the standardized management model needs to adapt to the digitalization of standards to explore.
The second is the challenge of artificial intelligence technology. Use artificial intelligence technology to analyze big data, clarify the ins and outs from intricate things, and give decision-making reference. The application in many fields has shown the wisdom beyond the staff, but it is also necessary to pay attention to the artificial intelligence through the internal neural network operation. The interpretability of the result is poor, and it is often unknown why it is known, and it cannot explain why the result is. In some important applications, it is inevitable to have doubts about the credibility of the results given by artificial intelligence. Standard machine-readable technology will use artificial intelligence technology to understand semantics. It also faces these dilemmas in artificial intelligence technology. If some very important things are operated according to machine-readable standards, it may be right 99% of the time. , but it does not rule out that it will be misjudged 1% of the time . How to avoid misoperation caused by artificial intelligence? When there is a misunderstanding of the human-readable standard, it is of course the responsibility of the human being. When there is a problem with the machine-readable standard, it is difficult to define whether the designer of the machine algorithm or the user or owner of the machine should bear the responsibility. We cannot give up efforts to standardize digitization because standard machine readable will fall into the misunderstanding of current artificial intelligence technology. It is impossible to improve any technology, but it is necessary to promote the application of machine readable. In order to avoid risks, machine readable standards can be Start with standards where machine-readable misjudgments do not have a big impact.
The third is the challenge of network security. Machine reading will inevitably use computer systems. Information technology is a double-edged sword. The network system itself is vulnerable to hacker Trojan horse attacks. Some risks on the network will also have such risks after the standard is digitized.
Can you imagine the consequences of using “poisoned” machines to understand standards? The digitization of standards requires great attention to the security of the information systems they are associated with, and important standards cannot be fully interpreted by machines. Humans must check the results of machine interpretation.
A long way to go
The task of standard digitization is arduous, and long-term planning is required to advance in an orderly manner. Currently, we need to start from the following aspects.
One is the urgent need for a standard for machine-readable standards. Identify the elements of machine readability and what technologies are used to support machine readability. Regarding the standard description language, the XML extensible markup language was mentioned above, and some people proposed to use the JSON language. In the future, there may be other formats of languages, which need to be preferred. If multiple languages are side by side, the compatibility problem must be solved. There must be a unified machine representation for formulas, algorithms, programs, flowcharts, circuit diagrams, and graphs in the standard. In short, the most basic metadata in machine-readable standards needs to be stipulated first, and the National Standards Commission has deployed the China National Institute of Standardization to do research in this area. China National Institute of Standardization and Zhejiang Institute of Standardization have also held some seminars to discuss how to do the basic work of standard digitization.
At present, ISO, ITU and some foreign standardization organizations are all studying the digitization of standards, but so far, there has not been a recognized machine-readable standard in the world, and it is not easy to start in this regard. However, some fields have begun to explore, and passports and visas are required to travel abroad. In the past, it was seen by border guards, but now it can be read by machines. A national standard for machine-readable passports emerged in 2017, including three parts: machine-readable passports, machine-readable visas, and machine-readable travel documents. Modern passports have chips in them, just like our domestic second-generation ID cards. Machine-readable ID cards and passports are actually reading the chip, and the information on the chip is read out and compared with the pre-stored information. The model of machine-readable passports is difficult to use machine-readable standards, and the method of embedding chips into each standard text is difficult to generalize.
Second, the standard is readable and understandable, and a corresponding professional platform needs to be established to support it. The standards are basically professional, and the understanding of these majors is not sufficient for ordinary artificial intelligence, and requires training. Use big data technology to build knowledge graphs from massive standard texts, build standard knowledge “ontologies”, and discover machine-readable rules. The China Institute of Electronic Technology Standardization is studying the specification of the standard knowledge graph, which is also a basic research, and has not yet formed a standard that can be published. However, there are many fields covered by the standard, and it is not enough to have a general knowledge map. It is necessary to develop corresponding standard knowledge maps by professional fields, and develop and build a number of standard machine-readable cloud platforms for professional fields, such as machinery and electronics. , biological, chemical, metallurgy, construction and other industries standard digital technology platform. These work require the overall coordination of the national standardization management department, instructing the relevant departments to develop and establish such a platform to help promote the application of digital standards. It is understood that my country’s National Standards Commission has deployed to explore the research and possible technical support of machine-readable standards in the field of aviation.
The third is talent training. We originally wrote standards should say that many of them are veterans, know how to write standards, and use standard language to be more concise and less ambiguous. However, the original standardization veterans may not be competent to write machine-readable standards described in languages such as XML, and ordinary IT experts are not familiar with standardization rules and industry technologies. It is necessary to be familiar with the profession, the rules written by the original standard, and the use of machine-readable technology. Such talents are absolutely scarce. The issue of talent development for standard digitalization needs to be put on the agenda as soon as possible.
The fourth is to establish standard digital management norms. The digitization of standards is a new issue. It involves the management of the whole life cycle of standardization. We need to revise the corresponding regulations and processes of standardization in a targeted manner. Otherwise, the legality of standard machine-readable text form and its production method will be a problem. Standardization has a strong international character, and the research on standard digitization needs to strengthen international cooperation. In addition, the digitization of standards cannot be rushed. It is necessary to plan and conduct pilot projects before launching. The most effective application scenario of machine-readable standards is that the equipment on the production site automatically executes the process according to the standard. Standard machine-readable standards in these fields can be implemented first. advance. It takes a lot of work to become machine-readable with so many existing standards, and it needs to be considered in the long run. In a word, standard digitization is a rare opportunity for the development of standardization, and it is also a new challenge for the departments engaged in standardization research, standard formulation and implementation, and standardization management.
The source of this article is “China Standardization 2022-04-26”
Note: Unless it is marked as original, it is all submitted by netizens or institutions for sharing. If you need to publish, please contact [email protected].
This article is reprinted from: https://www.dx2025.com/archives/26786.html
This site is for inclusion only, and the copyright belongs to the original author.