Article 25 of the Patent Law provides for six types of subject matters for which no patent right should be granted. No matter how practical, novel or inventive they are, specific results covered by such subject matters are excluded from patent protection. Nevertheless, the determination of ineligible subject matters does not require evidence to aid comparison as in the case of commenting on novelty or inventiveness, which can be concluded directly and merely by deductive reasoning. Therefore, it is particularly important to clearly define the concept of the various statutory exclusions in the article.

The "rules and methods of intellectual activities" (the "rules of intellectual activities") are one of the six ineligible subject matters. With the development and popularization of Al technology, a large number of computer-aided software inventions have emerged. The so-called "software" is a computer program in the broad sense, which can include the external expression of the program code, as well as the internal operation logic of the program algorithm. Therefore, software inventions often involve "the computer program per se", "the language and computation rules of the computer", "the mathematical theory" and other specific types of rules of intellectual activities specified in the Guidelines for Patent Examination (the "Guidelines"), thus resulting in the blurring of the conceptual line between rules of intellectual activities and patentable computer technical solutions.

Provisions for Program Codes Not Eligible

The Guidelines defines the ineligible subject matter "the computer program per se" as a sequence of coded instructions that can be implemented by a device with information processing capabilities, such as a computer, in order to obtain a certain result, or a sequence of symbolic instructions or symbolic statements that can be automatically converted into a sequence of coded instructions ^[1]", specifically including "source programs and target programs". By combining this definition with basic concepts of the computer sector ^[2], it is clear that a "target program"means a program written directly in machine language (i.e., binary digits), while a "source program" is one written in assembly language or high-level computer language (such as C) that is run by the computer by automatically compiling it into binary digits.

The above definition of the "computer program per se" appeared in the 2001 edition of the Guidelines and has been there to date. Both the "source program" and the "target program" are written in program codes of non-natural language, which, in theory, can be easily distinguished from technical solutions written in ordinary text. However, in the practice of China, this concept used to be interpreted in an expanded manner for a long time. For example, the 2009 Review Practice Protocol explicitly states that all claims with words like "program" and "instruction" in the subject name are to be regarded as the "computer program per se", and can be excluded without examining the other contents of the claims ^[3]. It was only when the Guidelines were revised in 2017 that the original sense of the "computer program per se" was strictly followed and limited to program code.

The exclusion of the "computer program per se" from patent protection is currently the common practice and consensus in various countries and regions around the world. For example, the European Patent Convention (the "EPC") explicitly states that programs for computers are not patentable inventions ^[4]. The Patent Act (the "Act") in the UK also explicitly states that a program for a computer is not an invention under the Act^[5]. The PCT International Search and Preliminary Examination Guidelines promulgated by the World Intellectual Property Organization (WIPO) makes it even clearer that, as a general rule, claims that have only recorded program code are excluded subject matter ^[6]. Although the U.S.Patent Act does not explicitly exclude computer programs from patent eligibility, the Manual of Patent Examination Procedure (MPEP) promulgated by the United States Patent and Trademark Office (USPTO)lists "computer program per se" ^[7] as a "judicial exception" and excludes it from the four categories of statutory subject matter ^[8] of Section 101, based on U.S. jurisprudence.

Two views are currently prevailing on the rationale inherent in this pervasive legal exclusion, as follows.

The U.S. Federal Circuit holds that product claims for intangible collections of information, even if they are created by human effort, do not fall into any statutory category. In the prejudication of the 2007 Microsoft case(Microsoft Corp. v. AT&T Corp.), it was stated that software expressed in code or a series of instructions and detached from any medium was an idea without physical substance. Therefore, product claims for software programs that do not have at least one structural qualification (e.g., a "device plus its functions"qualification) do not have a material or tangible form and thus do not fall within any legal category ^[9].

This interpretation has emphasized the tangible nature of the product of invention. However, with the development of computer technology, the essence of the product of invention has been increasingly leaning toward intangible results like software. The U.S. Supreme Court has also noted this trend, and in the prejudication of the 2010 Bilski case (Bilski v. Kappos), it made clear that the "machine or transformation test" (M-or-T test), which is a tangible test for the product of an invention, is not the only criterion for the determination of eligibility. Subsequently, the U.S. Federal Supreme Court further gave the new standards for the determination of eligibility in the prejudication of the 2014 Alice case (Alice Corp. Pty. Ltd. v. CLS Bank Intl). In 2017, the USPTO developed the Alice/Mayo test based on the prejudication of the Alice case ^[10], which uses the more superior attribute of "inventive concept" as a criterion for eligibility, and specifies that "tangibility is not required for the Alice/Mayo test" ^[11]. Therefore, it can be seen that the doctrine of outright denial of patent eligibility for program codes solely on the basis of intangibility has, in fact, been rejected by the U.S.patent justice and administration.

The UK Intellectual Property Office has explained the reasons why software is not suitable for patenting in terms of patent examinability by arguing that the inability or difficulty for the patent offices to examine software-based patents would directly lead to the production of poor quality patents, which would, in turn, hinder the development of the industry ^[12]. WIPO has also specified in the PCT specifications from an examination perspective that computer programs are excluded subject matters to the extent that the international search unit is not equipped to search for prior art related to the program ^[13]. Zheng Chengsi, a scholar in China, has also made a more specific interpretation from such a perspective: "If software is given full patent protection, the novelty search would be the first issue pending for solution. More software products appear in the world than new inventions being developed each year, and it will be difficult to identify whether a piece of patent-pending software is novel." ^[14]

Nevertheless, when such assertions were made, the assertors could hardly have foreseen the continuous surge in global patent filings. According to WIPO statistics, from 2005-2019, patent applications for inventions worldwide kept a general year-on-year growth, with the number of applications in the single year of 2019 alone exceeding 3 million ^[15]. In addition, the length of individual claims was becoming increasingly longer, with claims of more than a thousand words abounding. However, in the practice of examination, a patent application must not be denied on the grounds that the amount of relevant literature is excessive and the length of claims is too long that it is making search and examination difficult. Therefore, under the context of the actual situation, the difficulty of examination is also not a sufficient reason to exclude program codes from patent subject matters.

Therefore, it could be seen that, as technology evolves and circumstances change, the old doctrine is no longer sufficient to adequately support the artificial provisions to exclude computer programs from patent subject matters. According to the Resolution T11797 of the Appeals Board of the European Patent Office, the European Patent Office (EPO) even made a variation on the EPO Examination Guidelines as early as the 2001 edition, which contradicts the European Patent Convention("EPC"): although "programs for computers" have been included among the items listed in Article 52(2) of the EPC, they are not covered by the provisions of Article 52(2) and (3) of the EPC if the subject matter requested is of a technical nature ^[16]. This provision has been in use until now, with only minor adjustments in its wording ^[17]. Therefore, it can be seen that the actual criteria for the EPO subject matter review restrict the meaning of ineligible "computer programs".

Although the EPO has made a major breakthrough in terms of the rules, and it seems that even computer program codes can be eligible subject matters, in practice, there is almost no one that will claim program codes as a patent application as program codes lack the generalization ability of natural languages and its scope of protection is too narrow. Therefore, the significance of the above provisions for practice is mainly for the further determination as to whether the inner conception of a computer program - the program algorithm, is patent-eligible.

Challenges in the Determination of the Eligibility of Program Algorithms

An algorithm is a description of a program operation that can be expressed in natural language ^[18]. Patent laws of different places around the world do not expressly exclude algorithms for computer programs. In fact, many software inventions in the computer sector have also been protected by patents precisely through algorithms described in natural language. However, patent laws of different places would explicitly exclude many closely-related subjects, such as the USPTO's "abstract ideas" ^[19], the EPO's "mathematical methods","schemes, rules and methods of intellectual activities, games or business"^[20] , etc. The rules of intellectual activities prescribed in China include, in addition to the "computer program per se", "management methods in organization, production, business implementation and economy", "mathematical theories" "rules and methods of all kinds of games and recreations","methods of statistics, accounting and bookkeeping", etc.^[21] , and all of these subjects can be expressed in computer program algorithms. Therefore, algorithms described in natural language could still fall into the above category and be found to be patent-ineligible subject matters.

Theoretically speaking, to decide whether a scheme involving an algorithm is patent-eligible, it is only necessary to determine whether the scheme has "technical properties":if it does, it is a patentable subject matter, otherwise it is not. However, the practical difficulty of such a criterion is that it is a self-evident conclusion to identify an industrial property right, such as a patent, as a technology; and it is difficult to further define the basic concept that plays a key explanatory role here - what "technology" is ^[22]. If we could not effectively understand the more fundamental concepts, then using only "technical properties" to explain the "patent subject matters"would produce circular arguments.

China's Guidelines explains "technical solution" as the aggregation of technical means applying the laws of nature to solve a technical problem ^[23]. For "inventions involving computer programs", the "three elements", i.e. technical problems, technical means and technical effects, are also used to assist in the determination of whether they are technical solutions. These "three elements" can further explain what a "solution" is from the three perspectives of "problem","means" and "effect", but in terms of the key challenge - what "technology" is - they are still tautological synonyms. Therefore, it is not easier to adopt this judgment method than to directly adopt the "technical solution" approach to explain the "patent subject matter".

The EPO has made extensive revisions to the subject matters of invention excluded in the 2018 edition of the Examination Guidelines, focusing mainly on software-related inventions, bringing to the forefront the requirement that "further technical effects must be produced" [i.e., "technical effects beyond the 'normal' physical interchanges between the program (software) and the computer(hardware)running the program"] as the primary point for consideration;while the requirement for "whether further technical effects exist should be assessed without reference to the prior art"is adjusted to follow the above requirement, becoming an effective complement to the former. The EPO also explains the specific situation "without reference to the prior art"according to the prejudications of its Appeals Board: for example, where a computer program takes less time to compute than a "prior art program" or performs better than a manual program for the same task, it is not sufficient to confirm that the computer program has further technical properties.

This judgment rule can be construed as more demanding than the "three elements" rule in the Guidelines of China. It requires not only that the eligible software invention has a technical effect, but also that such technical effect is not merely a general technical effect that the computer necessarily has. However, these rules can only be applied to "software-only" claims, and once a claim also expressly defines the computer for the implementation of the program or the computer readable storage medium for recording the program, the entire solution will be deemed to have technical properties and become a "computer-implemented invention", thus satisfying the eligibility requirement of Article 52(1) of the EPC. This is echoed in the "Business Methods" of the 2018 edition of EPO's Examination Guidelines, which states that "a business method is not excluded if the subject matter for which protection is sought qualifies that at least some of its steps are realized through technical means, such as computers, computer networks or other programmable devices" ^[24]. Obviously, this is an approach of judgment that focuses too much on the writing form.

In contrast, the current USPTO method for the determination of eligibility, i.e. the Alice/Mayo test, focuses more on the substance of the program. This test avoids the vague concept of "technology" and does not deliberately look for technical properties such as "computer" recorded in the solution. Instead, it first identifies the "judicially excluded subject matter" recorded in the claims, and collectively refers to "the other properties (including technical properties) other than the "judicially excluded subject matter" as "additional elements", and if the latter is "significantly more" than the former, then the patent is eligible, and if not, it is ineligible. For example, a claim that merely mentions the use of a computer to achieve an abstract idea does not suffice the level of "significantly more" ^[25], and even if it contains technical terms such as "computer", the overall solution is not patentable subject matter.

The Alice/Mayo test, while better able to capture the nature of the solution, will lead to new challenges in terms of the determination of "significantly more", and thus still produce many uncertainties and even misjudgments. For example, the prejudication in the 2011 Content Extraction & Transmission LC case (Content Extraction &Transmission LLC v. Wells Fargo Bank, N.A.), the patent in question claimed protection for an application interface that comprised a scanner that extracted data from physically copied documents, a processor that identified specific information from the extracted data, and a memory that stored the identified information. The Federal Circuit ruled that the claim was of abstract ideas about "data collection, identification and storage" and did not meet the patent-eligibility requirement. The solution consisted of a number of defined hardware (scanner, processor, memory) and did not involve any obvious abstract idea, which was very close to a scanner in the traditional sense, and yet it was still ruled to be actually an idea "of itself ") ^[26]. Arguably, this is an analogical argument quite extreme for prejudication of the Alice case, and the conclusion has also gone beyond the intuition of average people.

Analysis of the Ineligibility of Program Codes through "Mapping"

The U.S."Alice/Mayo test" has opened up a new perspective and inspired practitioners in the patent industry to understand the meaning of the ineligibility of patent subject matters from a broader perspective.

As a kind of information property right that is disclosed in exchange for protection, the peer mapping of information is crucial. Extant studies have pointed out that the accurate mapping relationship between what is documented in a patent and the technical solution has been the most critical link of the patent document. ^[27] In other words, for a claim for which the inventor is claiming patent protection, information about the technical solution should be mapped out accurately in the mind of the readers through information carriers such as characters, while the effectiveness of accurately mapping information depends largely on the reliability of the information carrier. In an actual patent application, for a technical invention to obtain patent protection in a particular country, the claim and specification must be written in the official language commonly used in the country, which is not only a political issue relevant to national sovereignty but also a technical issue concerning the accuracy of information mapping.

In this sense, the exclusion of the "computer program per se" from the subject matter of a patent is a reasonable provision. For human beings, even program codes written in a high-level computer language are not as easy to read and understand as the text of an algorithmic process written in the native tongue, let alone binary digits specifically designed for computer recognition. Describing a claim for program codes is similar to expressing them in a foreign language. Although it can be loaded with information about the invention, it is obviously not easy for the public in the country of the patent protection to understand the intrinsic conception of the invention, thus making it impossible for them to create a smooth and accurate information mapping.

The fundamental requirement of a patent for disclosure dictates that the boundaries of such intangible property right should be made explicit in the most effective manner. Therefore, patent information should be disclosed in the country of protection in the official common text where information could be most fluently disseminated, rather than in a foreign language, the C language, or other semi-cryptic forms, and the exclusion of "the computer program per se"from patentable subject matters in national or European patent laws should not be changed with the increased importance of software technology. Even if an algorithm inherent in a computer program "has a technical property", it is not suitable for the claim to be written in the form of program codes for patent protection.

Judging the Eligibility of Program Algorithms by"Superordination"

Superordination: two-way information mapping The claim of a patent must not only forms a one-way information mapping in the mind of the readers but also a two-way information mapping test with the relevant product entities when judging the existence of an infringement. The determination of whether a claim is identical or equivalent to an entity is obviously not a comparison of the similarity in the appearance of characters and components, but a comparison of the match between the two in terms of the intrinsic information. Hofstadter, an American researcher on Al theory, proposed the concept of "superordination" based on the theory of mathematical logic, which is specifically defined as "two complex structures that can be mapped to each other, and each part of either structure has a corresponding part in the other structure." ^[28] We could draw on this concept to express the information equivalent relationship between two different subject matters, i.e. the claim and the infringing entity.

From the point of view of information "superordination", if a solution described in text cannot find any "isomorphic"entity in the real world, there will be no need for such text information to be a subject matter of patent protection -Since there is no "isomorphic" entity, there will be no industrial exclusivity for production, use, or sale. From this perspective, it is also easy to understand why patent laws in various countries or regions generally exclude purely subjective ideas from patent subject matters, and why the U.S.courts used to place great emphasis on the tangible properties of the products of invention when determining the patent subject matters.

In fact, the so-called intangible products such as computer software, unlike purely subjective and intangible ideas formed only in the mind, implies its capacity to be definitely preserved with the help of computers and other general-purpose hardware, that is, software inventions have the prerequisite to finding the relevant physical products in the real world. However, a software product with actual utilities is inevitably a designed and specific algorithm based on objective needs, so the practical application of the software product may only form "superordination" with a specific and informative algorithm claim, and it is difficult for it to achieve "superordination" with a general and abstract algorithm claim. Therefore, the eligibility of algorithmic claims should be judged on a case-by-case basis.

Analysis of Guidelines Cases with "Superordination"

The 2019 revision of the Guidelines has added two very informative examination examples to explain the extent of subject matter examination of program algorithms in both positive and negative aspects. In both cases, it involves the modeling of computer neural networks. The solution of Case 1 is summarized as "the training samples of tasks of other classifications related to those of the first classification are also used as training samples for the mathematical models of the tasks of the first classification, thus increasing the training samples and ...... and improving the accuracy of the modeling". ^[29] Case 2 is a training method for a CNN model of a convolutional neural network with the solution outlined as "after the convolutional operation and maximum pooling operation on the training image at each convolutional level, the feature images obtained after the maximum pooling operation is further subjected to a horizontal pooling operation so that the trained CNN model can recognize images to be recognized of any size when trying to recognize the image category". ^[30]

The Guidelines has found, from the perspective of the "three elements", that Case 1 "does not involve the combination of the subject matter, process and result with the specific application field, but is an optimization of the abstract mathematical method, and the whole solution does not include any technical features". Therefore, the patent application is not a subject matter for patent protection;while for "Case 2", the Guidelines has "made it clear that the data processed in each step of the model training method is image data and how the image data is processed in each step, re ecting that the training algorithm of the neural network is closely related to the processing of the image information. This solution addresses the technical problem of how to overcome the capacity of the CNN model to only recognize images of fixed size by using different processing and training means of images on different convolutional layers. It has used a technical approach that follows the laws of nature to obtain technical effects where the trained CNN model can recognize images to be recognized of any size". Therefore, the patent application for the invention is a subject matter for patent protection. ^[31]

Case 1 has constructed a mathematical model of higher accuracy by "adding training samples", and obviously, this modeling method is not a purely subjective idea. Instead, in Case 2, the modeling method only has mathematical terms such as "convolution," "pooling," and "iteration" and does not involve hardware as a typical technical attribute. Arguably, the difference between the two solutions lies more in the details of different levels in describing the improvement of the individuality of the inventions.

Specifically, the improvement of modeling accuracy by increasing the number of training samples, as described in Case 1, is a general principle of computer neural network algorithms, and it is the effective increase of the number of training samples that can reflect the improvement of the individuality of the invention. However, for the specific addition of "training samples for the tasks of the second classification", the claims only refer to "related to the said tasks of the first classification", and it is impossible to know the material difference and association between these two types of customized training samples. This makes the content that re ects the individuality of the invention a purely subjective and vague definition, which leads to the fact that the whole solution has remained at the level of "mathematical theory" and thus falls into the category of rules of intellectual activities.

In contrast, Case 2 has described in detail the use of specific computational means for the individuality of different stages along the modeling process, which has ultimately achieved the dynamic adjustment of the resolution for image recognition. Such content that reflects the personality of the invention is quite objective and relevant, no longer the generic and grounded theory of the CNN model, so it is not considered to be in the category of rules for intellectual activities.

For inventions that improve conventional computer functions, the US MPEP also requires that the disclosed content must provide sufficient details to enable a person of ordinary skills in the art to recognize that the claimed invention has provided an improvement for it to constitute an eligible solution, and provides that indications of the claimed invention producing an improvement may be included in the description of the specification, which identifies the technical problems and explains the details of the non-conventional technical solution, and is also expressed in the claim. ^[32]

From the perspective of "superordination", the reasonableness of the judgment that Case 1 is not a patent subject matter can also be understood in the following way:since the information of the "training samples of the tasks of the second classification", which reflects the individuality of the invention, is subjective and vague, it is impossible to locate a precise correspondence to it in the real world, which, as a result, leads to the impossibility of the claim to form a "superordination" with any software product in the real world.

It could thus be seen that claims that do not limit the individual details of inventions cannot be "isomorphic" to each other even if there are related entities; and an epistasis algorithm that does not reflect the details of the improvement of the invention individuality is thus not eligible as a subject matter for patent protection.

It is worth mentioning that the requirement to reflect the details of the improvement of an invention's individuality is not the same as prohibiting the claims from making an epistatic generalization. A patent-eligible epistatic claim is a collection of superordinate concepts with similar attributes, and each of these superordinate concepts would be clear to a person skilled in the art. For example, the claim records a wire made of metal, and a skilled person in the field can clearly know that it contains a series of specific metal wires such as gold wire, silver wire, copper wire, and iron wire. The superordination of the algorithm claim described in Case 1 is essentially a blurring of information due to the use of subjective customized "training samples for tasks of the second classification" for key characteristics, and is thus difficult for those skilled in the art to accurately extract implementable subordinate concepts thereof.

Social Implications for Refined Eligibility Criteria of Software Invention

For conventional inventions, the corresponding product of an invention can eventually be implemented in a specific structure, so it is clear whether the superordinate claim can be subordinated or not. However, for software inventions, the virtual nature of the invention products makes it quite difficult to judge the implementation of the superordinate claims according to traditional standards, and thus also difficult to assert whether the whole solution is "unclear""unsupported" or "inadequately disclosed" and other legal issues.

It is not only difficult to deny such sup superordinate software inventions with clauses such as "unclear", but it is also challenging to locate prior art or common knowledge to deny their inventiveness due to their virtual means and advanced ideas. Once a patent for such software invention is granted, the potential pitfall is that it would not be difficult to propose such superordinate solutions of key means which are vague but generally feasible ahead of time, but it would be obviously an unfair consequence if others overcome practical difficulties and create practical and feasible results, which will fall under their protection and thus incur infringement proceedings.

The "patent trolls" that have been proliferating in the U.S. since the 21st century have made this potential trouble a reality. The main litigation tool of the "patent trolls" is software patents, which are overwhelming the U.S. industry with the pains they bring. According to the US National Economic Council and other administrative agencies, "patent assertion entities" (mainly "patent trolls")not engaged in manufacturing initiated 2,500 infringement lawsuits between 2010 and 2012, mainly in the sector of information technology with nearly 82% of the infringement lawsuits arising from software patents. ^[33]

The frequent occurrence of patent blackmails has, to a certain extent, fulfilled the above-mentioned concern of the UK Intellectual Property Office: the difficulty of examination has led to the creation of poor-quality patents, which has, in tum, hindered industrial development. Therefore, the use of patent ineligibility, the artificially-drawn legal red line, to exclude specific subject matters is the most convenient and effective means of deterrence. And this is also why the U.S. Patent and Trademark Office and courts at all levels have abandoned the prejudication of the 1981 Diamond case (Diamond v. Chakrabarty), which held that "everything made under the sun is patentable," and have bucked the trend of lowering the threshold of patentable subject matter in the world today by setting increasingly stringent standards for software The case of Diamond v. Chakrabarty, in which the principle of "everything made under the sun is patentable,"bucked the trend of lowering the threshold of patentable subject matter in the world today, and set increasingly stringent eligibility standards for software inventions.

Software inventions are different from both physical inventions in traditional technology and purely subjective ideas in the mind, which is a new problem faced by the judgment of patent subject matters. In recent years, all major countries or regions in the world have all made frequent revisions to their criteria for patent eligibility, mainly to formulate patches to the rules to address this new issue in order to make the old patent system theory compatible with the booming new thing. According to the pragmatic situation of each country or region, it is necessary to narrow down the implication of unpatentable subject matters in software inventions, which has also brought the huge difference in the definition of the connotation and eligibility criteria of the patent subject matters of software inventions in different countries or regions, which is evident from the history of the relevant laws and examination guidelines of China, the United States, Europe and other countries or regions that this paper has discussed.

In the information age of open-source sharing, only by carefully and reasonably singling out eligible subject matters for patent protection from software inventions could we reconcile the contradiction between the sharing of social resources and individual enthusiasm for innovation, so that the development of the patent system can keep pace with the times and avoid the proliferation of "patent trolls"while promoting the development of social economy and technology.

References

¹ Guidelines for Patent Examination 2010 (revised in 2019), Part II, Section 1.

² See Tan Haogiang. C Programming, Tsinghua University Press, second edition, 199, p.7.

³ See Examination Operating Procedures-Substantive Examination Volume, Chapter 9, Article 1.1 and 1.2.

⁴ See European Patent Convention (2010) Part II, Chapter I, Article 52-(2)-(c).

⁵ See Intellectual Property Office: Patent Act 1977 (amendment 2011) Chapter 37, Part I, Article 1-(2)-(c).

⁶ WIPO: PCT International Search and Preliminary Examination Guidelines (2018) Part 鈪�, Chapter 9.

⁷ USPTO:MPEP(2019)$2106.03,2020,page 2100-21.

⁸ The four types of patent-eligible subject matters specified in Article §101 of the U.SS. Patent Act are methods, machines, articles of manufacture, and compositions.

⁹ Same as Note 7.

¹⁰ See USPTO: MPEP(2017)S2106,2018,page 2100-16.

¹¹ USPTO:MPEP(2019)$2106.05,2020,page 2100-61.

¹² Qiu Lei'an et al. A Study on Patentability of Overseas Software-Based Technologies, Intellectual Property, No.7,2016, p.127.

¹³ WIPO: Patent Cooperation Treaty Regulations (2015), 39.1-(vi).

¹⁴ Zheng Chengshi, Intellectual Property Law, Law Press China, Second Edition, 2003, p.32.

¹⁵ See WIPO: World Intellectual Property Indicators 2020, page 24.

¹⁶ EPO: Guidelines for Examination in the European Patent Office (2001), Part C, Chapter IV,2.

¹⁷ See EPO: Guidelines for Examination in the Eurcpean Patent Office (2018),Part G,Chapter ll,3.6.

¹⁸ See Tan Haoqiang. C Programming, Tsinghua University Press, second edition, 1999, p.13&19.

¹⁹ See USPTO: MPEP(2019)$2106,2020,page 2100-15.

²⁰ See EPO: Guidelines for Examination in the European Patent Office (2018), Part G, Chapter ll, 3.3, 3.5.

²¹ See Guidelines for Patent Examination 2010 (revised in 2019), Part II, Chapter I, Section 4.2.

²² See Yin Xintian. Introduction to the Patent Law of China, Intellectual Property Press, first edition, 2011, p. 18.

²³ See Guidelines for Patent Examination 2010(revised in 2019), Part 鈪�, Chapter I, Section 2.

²⁴ EPO: Guidelines for Examination in the European Patent Office (2018), Part G, Chapter 鈪�, 3.5.3.

²⁵ USPTO:MPEP(2019)$2106.05(f),2020,page 2100-78.

²⁶ USPTO:MPEP(2017)$2106.04(a)(2),2018,page 2100-33

²⁷ Qiu Lei'an et al. A Study on Patentability of Overseas Software-Based Technologies, Intellectual Property, No.7,2016, p.130.

²⁸ [U.S.] Douglas Richard Hofstadter, Godel, Escher, Bach: An Eternal Golden Braid, The Commercial Press, first edition, 196, pp.667.

²⁹ See Guidelines for Patent Examination 2010(revised in 2019), Part 鈪�, Chapter IX, Section 6.2.

³⁰ Same as note 29.

³¹ Same as note 29.

³² USPTO:MPEP(2019)$2106.05(a),2020,page 2100-51.

³³ See Patent Documentation Department of China National Intellectual Property Administration. U.S. Report Shows "Patent Assertion Entities" Hinder U.S. Innovation and Economic Growth, Foreign Intellectual Property Information, No.16,2013.