How to obtain patent protection for software inventions has always been a challenge. The China National Intellectual Property Administration (CNIPA) added one special section in its newly issued Guidelines for Patent Examination (2023), providing the following clarifications regarding the subject matter eligibility of patent applications involving software algorithms:

If the steps of the algorithm described in the claims are closely related to the technical problem to be solved, for instance, if the data processed by the algorithm carries a precise technical meaning within a technical field, and the execution of the algorithm directly embodies the process of utilizing natural laws to solve a technical problem, resulting in a technical effect, then generally, the solution defined by such claims falls within the scope of "technical solution" as mentioned in Article 2, Paragraph 2 of the Patent Law.

If the solution claimed involves improvements to artificial intelligence or big data algorithms such as deep learning, classification, clustering, and there exists a specific technical correlation between this algorithm and the internal structure of the computer system, enabling the resolution of technical issues related to enhancing hardware computing efficiency or execution performance, including reducing data storage volume, minimizing data transmission volume, and increasing hardware processing speed, thereby achieving technical effects that conform to the laws of nature and improve the internal performance of the computer system, then the solution defined by such claims falls within the scope of "technical solution" as mentioned in Article 2, Paragraph 2 of the Patent Law.

Furthermore, if the solution claimed deals with big data in specific application fields, utilizing classification, clustering, regression analysis, neural networks, and other methods to uncover inherent relationships within the data that conform to the laws of nature, thereby solving technical issues related to enhancing the reliability or accuracy of big data analysis in those specific application fields, and achieving corresponding technical effects, then the solution defined by such claims falls within the scope of "technical solution" as mentioned in Article 2, Paragraph 2 of the Patent Law.


To avoid subject matter eligibility issues in patent applications involving algorithms, the following steps and considerations are crucial:

1. Clarify the Technical Nature of the Technical Solution

  • Tie the Algorithm to a Specific Technical Scenario: Ensure that the algorithm relates to a technical problem and a technical scheme solving such problem, and the execution of such algorithm should be in compliance with natural laws.
  • Demonstrate Technical Effects: Show how the algorithm's output has a tangible and verifiable technical effect.

2. Craft a Thorough Claims Section

  • Specify Application Scenarios: In the claims or at least in the embodiments, describe the specific technical domains or industrial applications where the algorithm is utilized.
  • Define the Technical Solution Clearly: In the claims, combined with the technical features, the realization process of the algorithm is described, in which the abstract algorithm features or algorithm means should be embodied as technical features or technical means.
  • Highlight Objective Technical Effects: In the embodiments, substantiate technical effects with experimental data, comparative tests, or other objective measures are described. The technical features corresponding to the technical effects in the embodiments need to be recorded in the claims.
  • Align with Patent Examination Guidelines: Ensure that the claims comply with relevant patent examination guidelines, particularly those addressing algorithms.


Examples of claim drafting:

Example 1: A knowledge map reasoning method

A knowledge map reasoning method based on relational attention, comprising:

Acquiring an initial embedded representation of a node in a knowledge map, and converting the initial embedded representation into a high-dimensional space to obtain a high-dimensional embedded representation, wherein the node is……; acquiring a neighbor node set of a target node in the knowledge map, and constructing a neighbor subgraph according to the relationship type between the target node and the neighbor nodes in the neighbor node set; aggregating the high-dimensional embedded representation of the target node and the neighbor embedded representation to obtain an aggregated embedded representation of the target node; fusing the aggregated embedded representation according to the first attention score of each neighbor subgraph to obtain a fused embedded representation of the target node; and calculating, according to the fusion embedding representation, the score of the triple corresponding to the target node.

Analysis:

This solution is a knowledge map reasoning method based on relational attention. The data processed in each step of this method is technical data such as text data or semantic information in natural language. The knowledge map is constructed by entity recognition and relationship extraction of related knowledge in question answering system and semantic search, so as to carry out knowledge map reasoning. This solution solves the technical problem of how to enrich semantic information and improve the accuracy of reasoning in the process of text embedding and semantic search. It uses the technical means of natural laws and obtains corresponding technical effects. Therefore, the solution of the invention patent application belongs to patentable subject matter stipulated in the second paragraph of Article 2 of the Patent Law.

Example 2: A training method of deep neural network model

A training method of a deep neural network model, comprising:

Calculating the training time consumption of training data in a preset candidate training scheme; selecting a training scheme with minimum training time consumption from preset candidate training schemes, wherein the candidate training scheme comprises a single-processor training scheme and a multi-processor training scheme based on data parallelism; and carrying out model training on the changed training data in the optimal training scheme.

Analysis:

The solution is a training method of deep neural network model. In order to solve the problem of slow training speed, the model training method selects a single-processor training scheme or a multi-processor training scheme with different processing efficiency for training data of different sizes. The model training method has a specific technical connection with the internal structure of the computer system, which improves the execution effect of hardware in the training process, thus obtaining the technical effect of improving the internal performance of the computer system in line with natural laws. Therefore, the solution of the invention patent application belongs to patentable subject matter.

Example 3: An Analysis Method of Electronic Voucher Usage Tendency

A method for analyzing the usage tendency of electronic vouchers, comprising:

classifying electronic vouchers according to the information of electronic vouchers to obtain the types of electronic vouchers; acquiring user sample data according to the application scene of the electronic ticket; extracting user behavior characteristics from the user sample data according to user behaviors, wherein the user behaviors include: browsing web pages, searching keywords, paying attention, joining shopping carts, purchasing and using electronic coupons; Taking user sample data as training samples and user behavior characteristics as attribute tags, the identification model of e-ticket usage tendency is trained for different types of e-tickets; and predicting, through the trained identification model of e-voucher usage tendency, the usage probability of e-vouchers, and obtaining the user's usage tendency for different types of e-vouchers.

Analysis and conclusion

The solution relates to a method for analyzing the usage tendency of electronic vouchers, which deals with the big data related to electronic vouchers. By classifying electronic vouchers, obtaining sample data, determining behavior characteristics and carrying out model training, the internal relationship between user behavior characteristics and usage tendency of electronic vouchers is excavated, and the behavior characteristics such as long browsing time, multiple searches, frequent use of electronic vouchers and the like indicate a high usage tendency of corresponding types of electronic vouchers. This internal relationship conforms to the laws of nature, thus solving the technical problem of how to improve the accuracy of analyzing users' inclination to use electronic vouchers, and obtaining corresponding technical effects. Therefore, the solution of the invention patent application belongs to patentable subject matter.

Example 4:

A recommendation method, comprising:

collecting the behavior data of the first sample user; conducting marketing activities in different ways to the first sample user; collecting feedback information and income information, training and obtaining a personalized marketing strategy recommendation model according to the feedback information and income information; and determining a target marketing mode corresponding to a target user according to the personalized marketing strategy recommendation model, and conducting marketing activities to the target user.

This method can improve the marketing effect and user experience, and increase the income of marketing activities.

Suggestions:

(1) "direct economic benefits" should not be considered as the beneficial effects, which is not a technical effect;

We can try modifying the beneficial effect in the direction of improving the user experience to form a technical effect. For example, improve the user's pertinence of pushed content.

(2) The "revenue information" in the training data can be changed to "user purchase information", which is composed of technical features.

By adhering to these guidelines and adopting a proactive approach, applicants can increase the chances of overcoming subject matter eligibility challenges and securing patent protection for their algorithm-based inventions. Please don't hesitate to contact Jiaquan IP Law if you have any further questions.