Xiaoli SU, Partner, Patent Attorney

 

On December 21, 2023, the China National Intellectual Property Administration (CNIPA) released the amended Patent Examination Guidelines, which came into effect on January 20, 2024. For the sake of narration, the Guidelines amended in 2023 are referred to below as the new Guidelines, and the previous Guidelines as the old Guidelines. This article will be presenting and interpreting the main amendments made to Chapter 9 of Part II of the New Guidelines concerning the examination of invention patent applications relating to computer programs.

 

I. More Types of Subject Matter of Invention Patent Applications Relating to Computer Programs

 

The old Guidelines provided that the claims of an invention patent application relating to a computer program could be written as process claims or product claims, such as claims for a device for executing the process. In the new Guidelines, it is further pointed out that in addition to the device for executing the process, the subject matter of the product claims may also be a computer-readable storage medium or a computer program product.

 

For the first time in the new Guidelines, it is explicitly allowed to write about the topic of computer program products. However, it should be noted that a computer program product is not a computer program per se, but should be understood as a software product that executes its solution primarily through a computer program. This is due to the fact that with the development of Internet technology, more and more computer software no longer relies on traditional physical storage media, such as optical discs or disks, but are transmitted, distributed and downloaded in the form of signals through the Internet. The new Guidelines allow computer program products to be used as the subject matter of protection, so that the protection of computer programs is no longer limited to tangible storage media, thus facilitating more comprehensive protection of invention patent applications relating to computer programs.

 

So far, the claims of an invention patent application relating to a computer program can be drafted in the following four common ways:

 

1. A method for removing image noises, characterized in that it comprises the following steps....

2. A computer apparatus/device/system, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to realize the method of claim 1.

3. A computer-readable storage medium in which a computer program is stored, wherein the computer program is executed by a processor and a step that executes the method of claim 1.

4. A computer program product, comprising a computer program/instruction, characterized in that the computer program/instruction is executed by a processor and is a step for executing the method of claim 1.

 

II. More Detailed Standards for Examination of Subject Matter of Invention Patent Applications Relating to AI and Big Data

 

The invention patent applications relating to AI and big data usually comprise features of algorithms or features of business rules and methods, and such applications need to be examined as to whether they are the technical solutions as described in the paragraph 2 of Article 2 of Chinese Patent Law (hereinafter referred to as Article 2.2).

 

Under the old Guidelines, when an invention patent application relating to AI or big data is examined under Article 2.2, it is necessary to determine whether the claimed solution falls into a specific technical field, and the algorithm-processed data is data with exact technical meaning in the technical field. To a certain extent, this limits the scope of protection accorded for invention patent applications relating to AI and big data, and is not conducive to promoting the research and improvement of the core or basic algorithms of AI and big data.

 

The new Guidelines further add (1) standards for examination of subject matter of algorithmic improvements that do not define specific technical fields in the claims, and (2) standards for examination of subject matter relating to big data processing.

 

(1) Standards for the examination of the subject matter of algorithmic improvements that are not defined in the specific technical field in the claims

 

Section 6.1.2 of the new Guidelines clearly stipulates that if the claimed solution is related to improvement of AI and big data algorithms of deep learning, classification, clustering, etc., the algorithm has specific technical relationship with the internal structure of the computer system, and can solve the technical problems of how to improve the computing efficiency or execution effect of the hardware, including reducing the amount of data storage, reducing the amount of data transmission, and increasing the speed of hardware processing, so as to obtain the technical effect of improving the internal performance of the computer system in accordance with the laws of nature; then the solution defined by the claim is a technical solution as described in Patent Law 2.2.

 

For example:

A training method for deep neural network models, comprising:

computing training time of training data respectively in a preset candidate training program when size of the training data changes;

selecting a training program with the least training time from the preset candidate training program as the optimal training program, wherein the candidate training program comprises a single-processor training program and a multi-processor training program based on data parallelism; and

model-training the changed training data in the optimal training program.

 

Analysis: The claim does not limit the processed training data to data with definite technical meaning in a specific technical field. However, in order to solve the problem of slow training speed, the model training method selects a single-processor training program or a multi-processor training program with different processing efficiency for the training data of different sizes, and the model training method has a specific technical relationship with the internal structure of the computer system, which improves the execution effect of the hardware in the training process, so as to obtain the technical effect of improving internal performance of the computer system in accordance with the laws of nature. Therefore, the solution of the invention patent application is a technical solution as specified in Article 2.2, so a patentable subject matter.

 

(2) Standards for Examination of Subject Matter Relating to Big Data Processing

 

Section 6.1.2 of the new Guidelines also clearly stipulates that if the solution of the claims deals with big data in a specific application field, and uses classification, clustering, regression analysis, neural network, and the like, to excavate the internal correlative relationship in the data that conforms to the laws of nature, so as to solve the technical problem of how to improve the reliability or accuracy of the big data analysis in the specific application field, and obtain the corresponding technical effect, the solution defined by the claim is a technical solution as described in Article 2.2.

 

Example:

An analysis method for tendencies to use e-coupons, characterized in that it comprises:

classifying the e-coupons according to information of the e-coupons, to determine e-coupon types;

obtaining user sample data according to application scenario of the e-coupons;

excavating user behavior characteristics from the user sample data according to user behavior, wherein the user behavior comprises browsing a web page, searching for keywords, adding follow, adding to a shopping cart, purchasing, and using e-coupons;

training a tendency recognition model for different types of the e-coupons, by using the user sample data as training samples and the user behavior characteristics as attribute labels; and

predicting probability of the e-coupons being used by the e-coupon tendency recognition model, to obtain the tendencies to use e-coupons of different types.

 

Analysis: In the field of big data processing and analysis, the individual behavior of a single user has a certain degree of subjectivity and randomness, but the behavior of a group of users is often regular, and the correlation between different behaviors can reflect and conform to specific natural laws. Therefore, the means used to excavate the correlation between different behaviors of group users can also constitute technical means. The solution relates to an analysis method of the tendency to use e-coupons, the method deals with the big data related to e-coupons, through classifying e-coupons, obtaining sample data, determining behavioral characteristics, and carrying out model training, excavating the intrinsic correlation between user behavior characteristics and e-coupon usage tendencies, and the behavioral characteristics, such as long browsing time, many search times, and frequent use of e-coupons, indicate that the use tendency of the corresponding types of e-coupons is high, and this internal correlative relationship conforms to the laws of nature. Accordingly, the technical problem of how to improve the accuracy of analyzing the user's tendency to use e-coupons was solved, and the corresponding technical effect was obtained. Therefore, the solution of the invention patent application is a technical solution as specified in Article 2.2, so a patentable subject matter.

 

Counterexample:

A method for predicting price data of a financial product, wherein the method comprises:

training a neural network model by using historical price data of N + 1 daily indicators of the financial product to obtain a price prediction model, wherein the historical price data of the indicators of first N days is used as a sample input data, and the historical price data of last 1 day indicator is used as a sample result data; and

predicting the price data of the financial product of next day, using the price prediction model and the historical price data of the indicator of last N days.

 

Analysis: The solution relates to a price prediction method of financial products, which deals with big data related to financial products, and uses neural network models to excavate the internal correlation between the past price data of financial products and the future price data; however, the price trend of financial products follows the laws of economics, because the level of historical prices does not determine the trend of future prices, therefore, there is no internal correlation between the historical price data of financial products and the future price data that conforms to the laws of nature. The problem of how to predict the price of financial products is not a technical problem, and the corresponding effect obtained is not a technical effect. Therefore, this patent application is not relating to a technical solution as described in Article 2.2, nor is it a patentable subject matter.

 

It can be seen that, compared with the old Guidelines, the new Guidelines no longer limit the solution of the claims to specific technical fields and data with exact technical meanings, and further improve the subject matter examination standards for invention patent applications relating to AI and big data. Therefore, in order to satisfy the requirements of Article 2.2, for invention patent applications relating to AI and big data, it is feasible to draft or prosecute applications from any one of the following perspectives:

* The claimed algorithm is relating to a specific technical field, and the data processed is the data with exact technical meaning in the technical field.

* The claimed algorithm is relating to improvement of the internal performance of a computer system.

* The claimed algorithm excavates the intrinsic correlation in the data that conforms to the laws of nature.

 

III. Improved Inventiveness Examination Standards for Invention Patent Applications Relating to AI and Big Data

 

The inventiveness of an invention patent application relating to AI and big data that comprises both the technical features and the algorithm features or business rules/methods features, is examined following the principle of comprehensive consideration: The algorithm features or business rules/methods features are functionally mutually supportive and interactive with the technical features, and should be considered as a whole with the technical features. The phrase "functionally mutually supportive and interactive" means that the algorithm features or business rules/methods features are closely integrated with the technical features, and, together, constitute a technical means to solve a technical problem, and obtain a corresponding technical effect.

 

The old Guidelines set forth the following two provisions on how to determine "functionally mutually supportive and interactive":

 

1) If the algorithm in the claims is applied to a specific technical field, and solves a specific technical problem, then it is held that the algorithm features and the technical features functionally support each other, and have an interactive relationship.

 

2) If execution of the business rules and method features in the claims requires adjustment or improvement of technical means, then it is possible to be considered that the algorithm features and the technical features functionally support each other, and have an interactive relationship.

 

In the new Guidelines, the solution of the claims is no longer limited to a specific technical field, but, beside the above two provisions of the old Guidelines, the following provision has been added on how to determine "functionally mutually supportive and interactive":

 

3) If the algorithm in the claims has a specific technical relationship with the internal structure of the computer system, realizes the improvement of the internal performance of the computer system, raises the computing efficiency or execution effect of the hardware, in the respects including reduced amount of data storage, reduced amount of data transmission, and improved processing speed of the hardware, then it is possible to be held that the algorithm features and the technical features support each other, and have an interactive relationship with each other.

 

For example, a present patent application discloses:

A method for adapting neural network parameters, comprising:

selecting a plurality of dimensions for weight parameters of each layer of at least one layer of the neural network;

determining a size of the weight parameters in each dimension in the plurality of dimensions;

determining a set of candidate values for a target size of the weight parameters in the each dimension in the plurality of dimensions, based on usage of hardware supporting neural network computation;

selecting a subset of candidate values greater than or equal to the size on corresponding dimension in the set of candidate values, with minimum value in the subset of candidate values determined as the target size on the corresponding dimension; and

filing the weight parameter in the dimension if the size of the weight parameter in at least one dimension of the plurality of dimensions is smaller than the target size on the corresponding dimension, such that the size of the weight parameter obtained after filling is equal to the target size in the corresponding dimension.

 

And the reference document discloses: a design method for neural network processors, which searches for unit libraries from constructed neural network component libraries according to neural network topology, weight parameters and dimension parameters of each layer in neural network layers, and hardware resource constraint parameters, generates hardware description language codes of neural network processors corresponding to neural network models according to the unit libraries, and then converts the hardware description language codes into hardware circuits of neural network processors, wherein the neural network feature data and weight data are divided into appropriate data blocks for centralized storage and access.

 

Analysis: The claimed solution relates to a method for adapting neural network parameters. By obtaining neural network parameters with a canonical form, it maps the operations in the neural network to operations supported by the computing architecture, and simplifies the design and execution of the neural network-related hardware.

 

The solution of the above claim differs from the reference document in determining the size of each layer of the weight parameter in each dimension of the neural network, determining the set of candidate values for the target size of the weight parameter in each dimension based on the hardware utilization rate, selecting a subset of candidate values on the corresponding dimension, and determining the minimum value is the target size, and filling the weight parameter on the dimension if the size of the weight parameter on at least one dimension is less than the target size.

 

The application documents corresponding to the above claim show that the solution fills the size of the weight parameter to make it equal to the target size, and when the hardware supporting the neural network computes the data of the neural network, the hardware can efficiently process the data, and the algorithm in the solution improves the computing efficiency of the hardware. Therefore, the above-mentioned algorithm features and technical features used to adapt neural network parameters functionally support and interact with each other. Relative to the reference document, it is determined that the technical problem that the invention actually solves is how to make the hardware efficiently perform operations in a neural network. The above-mentioned content of improving the hardware computing efficiency by adapting neural network parameters has not been disclosed by other references, nor does it fall into the common knowledge in the art, and the prior art as a whole does not inspire the improvement of the technical solution of the above-mentioned reference document to obtain the present patent application, so the claimed technical solution of the patent application possesses inventiveness.

 

Therefore, in the process of drafting or prosecuting a patent application relating to AI and big data, which contains both technical features and algorithmic features or business rules/methods features, the applicant may consider any of the following aspects to show the functional mutual support and interaction between the algorithmic features or business rules/methods features and the technical features, so as to better support the inventiveness:

 

* The algorithm of the claim is applied to a specific technical field, and solves a specific technical problem.

* The claimed algorithm achieves an improvement in the internal performance of the computer system.

* The execution of the business rules and method features in the claims requires adjustment or improvement of the technical means.

 

In addition, considering that in the fields of AI and big data, the starting point of a considerable number of inventions is to improve the user experience, the new Guidelines also clearly add the inventiveness examination standards related to the improvement of user experience, that is, if the solution of an invention patent application improves the user experience, and the improvement is brought about or generated by the technical features and their functions, where there is an interactive relationship between the algorithm features or the business rules/methods features, they shall be taken into account in the inventiveness examination.

 

For example, a present patent application discloses:

A logistics distribution method that improves logistics delivery by notifying a user to pick up parcels in batches, comprising:

sending a arrival notification by a courier to a server that a parcel has arrived through a handheld logistics terminal, where the courier needs to notify a user to pick up the parcel;

sending notifications in batches from the server to all ordering users within delivery range of the courier;

completing pickup by the ordering users according to the notifications;

wherein, the server carries out the notifications in batch in such a way that, according to a delivery person ID, a current position of the logistics terminal and a corresponding distribution range carried in the arrival notification sent by the logistics terminal, the server determines all target order information within the delivery range centered on the current position of the logistics terminal corresponding to the delivery person ID, and then pushes the notifications to terminals of the ordering users corresponding to accounts of the ordering users in all target order information.

 

And the reference document discloses a logistics distribution method, wherein the barcode on the delivery order is scanned by the logistics terminal, and the scanning information is sent to the server to notify the server that the goods have arrived; the server obtains the information of the ordering user in the scanned information, and sends a notification to the ordering user; and the ordering user who receives the notification completes the pickup according to the notification information.

 

Analysis: The solution of the above claim differs from the reference document in that the user is notified in batches that the order has arrived, to realize batch notification, the data architecture and data communication mode between the server, logistics terminal and user terminal in this solution have been adjusted accordingly, and the pickup notification rules and specific batch notification implementation methods support each other, and have an interactive relationship in function. Compared with the reference document, it is determined that the technical problem actually solved by the invention is how to improve the efficiency of order arrival notification and thus improve the efficiency of parcels distribution. As a result, the operation of logistics delivery personnel can be more convenient, and the ordering user can receive pickup notifications in a more timely manner, which improves the user experience in both pick-up and delivery. The solution of the present application can obtain the technical effect of improving the efficiency of order arrival notification and then improving the efficiency of parcels distribution and the user experience, and the improvement of the user experience is brought about by the adjustment of the data mechanism and data communication mode that support each other, and have an interactive relationship in function, as well as the pickup notification rule and the specific batch notification execution mode. As long as no prior art has the technical inspiration to improve the solution of the above-mentioned reference document to obtain the technical solution of present patent application, the technical solution of present patent application will be considered inventive.

 

The new Guidelines have made it clear that in the inventiveness examination of patent applications relating to AI and big data, it is possible to consider user experience improvement as a technical effect. However, it should be noted that the improvement of user experience should be based on the feelings of group users, rather than individual differences, and the improvement should be improvement of user perception objectively brought about by the technical effects produced by technological improvements.

 

Conclusion

 

As is shown from the latest amendments to the Patent Examination Guidelines, the CNIPA has given clearer and more standardized guidance on protection of patents relating to algorithms and business rules in new formats and fields, such as AI, "Internet+", big data, and blockchain, in an effort to meet the demands for protection of all sorts of innovations.

 

 

 

 

 

References:

- Patent Examination Guidelines 2023 (Note: The examples cited in this article are all taken from the examination examples in section 6.2 of Chapter Nine in Part II of the newly amended Patent Examination Guidelines.)

- The Interpretation of the Amendments to Patent Examination Guidelines released on January 18, 2024 by the CNIPA

 

Author's Profile:

Ms. Xiaoli SU

Ms. Su received her degree of Master of Science from Shandong Normal University in 2003 and in the same year she entered the Institute of Computing Technology of Chinese Academy of Sciences to study for a doctorate. She has a lot of experiences in computer and communication patent applications, especially network computing and virtual computing. Ms. Su joined Panawell in 2010, and specializes in patent search, drafting, prosecution, reexamination and counseling in the fields of computer and communication.