An intelligent recommendation system for patent transformation and application objects

Abstract

The present invention relates to an intelligent recommendation system for patent conversion and application objects, which belongs to the field of intelligent recommendation technology and solves the problem of low accuracy of existing patent conversion and application object recommendation methods. The system includes: a multi-dimensional technical summary module, which uses a large model to obtain technical summaries of patents to be converted in different dimensions, and obtains technical summary results of patents to be converted in corresponding dimensions after reviewing the technical summaries; a multi-dimensional candidate patent retrieval module, which extracts features from the technical summary results of patents to be converted in different dimensions, and performs intelligent retrieval based on the extracted technical features to obtain candidate patent sets of patents to be converted in corresponding dimensions; a patent conversion and application object intelligent recommendation module, which obtains the co-occurrence degree, average similarity and operation degree of co-occurring patents between candidate patent sets of different dimensions of patents to be converted to determine the recommendation degree of each co-occurring patent; and the co-occurring patent with the largest recommendation degree is used as the patent conversion and application object.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to an intelligent recommendation system for patent conversion and application objects. Background Art

[0002] Patent transformation and application are crucial for promoting the practical application of technological innovation, optimizing overall societal resource allocation, and driving industrial upgrading. However, information barriers between patent holders and potential adopters hinder the discovery and effective application of numerous valuable technological achievements. Therefore, identifying potential candidates for patent transformation and application is crucial.

[0003] At present, there are some scientific and technological achievements or patent transformation service systems and patent recommendation systems on the market, and the implementation methods mainly include the following:

[0004] (1) Search for partners with relevant technical capabilities and resources through market research, visits, and negotiations;

[0005] (2) Establishing publishing channels and communication connections between the receiving entity and the supply entity;

[0006] (3) Recommendations based on the same IPC code; however, there are a large number of patents recommended based on IPC, and it is difficult for users to quickly find highly relevant target patents from the massive amount of patent information;

[0007] (4) After obtaining the pre-recommended patent data set based on the data features of bibliographic item dimensions such as international classification numbers, keyword sets, applicant sets, and inventor sets, the feature vectors of each patent to be recommended and the pre-recommended patent data set are compared, and the recommended patent is determined based on the relevance;

[0008] (5) Directly calculate the similarity between patent texts through methods such as keyword extraction and word shift distance, and make recommendations based on this.

[0009] Among these existing technologies, the trend is to identify patent conversion targets through intelligent matching using information technology. However, intelligent matching requires obtaining relatively accurate candidate patents and, based on this, screening out more suitable recommended patents through various control strategies. Currently, methods for obtaining candidate patents through feature matching of bibliographic data or calculating semantic similarity based on full-text texts suffer from accuracy issues, severely limiting the scope of subsequent discovery of patent conversion targets. Therefore, how to effectively improve the accuracy of patent conversion and application target recommendations is a technical problem that urgently needs to be addressed. Summary of the Invention

[0010] In view of the above analysis, an embodiment of the present invention aims to provide an intelligent recommendation system for patent conversion and application objects, so as to solve the problem of low accuracy of existing patent conversion and application object recommendation methods.

[0011] The present invention discloses an intelligent recommendation system for patent conversion and application objects, the system comprising:

[0012] The multi-dimensional technical summary module uses a large model to obtain technical summaries of the patents to be converted in different dimensions, and then reviews the technical summaries to obtain technical summary results of the patents to be converted in the corresponding dimensions;

[0013] The multi-dimensional candidate patent search module extracts features from the technical summary results of the patents to be converted in different dimensions, and performs intelligent searches based on the extracted technical features to obtain a set of candidate patents in the corresponding dimensions of the patents to be converted;

[0014] The intelligent recommendation module for patent conversion and application objects obtains the co-occurrence degree, average similarity and operation degree of co-occurring patents among the candidate patent sets of different dimensions of the patent to be converted to determine the recommendation degree of each co-occurring patent; and selects the co-occurring patent with the largest recommendation degree as the patent conversion and application object.

[0015] On the basis of the above solution, the present invention also makes the following improvements:

[0016] Furthermore, the multi-dimensional candidate patent search module includes: a technical feature extraction unit for each dimension, a search expression generation unit, and an intelligent search unit; wherein,

[0017] The technical feature extraction unit of each dimension extracts features from the technical summary results of the patent to be converted in the corresponding dimension, and obtains the technical features of the patent to be converted in the corresponding dimension;

[0018] The search expression generation unit comprehensively analyzes the technical features of the patent to be converted in different dimensions and obtains multiple intelligent search expressions for the patent to be converted in different dimensions;

[0019] The intelligent retrieval unit uses a variety of intelligent retrieval expressions of the patent to be converted in different dimensions to perform patent retrieval in the intelligent retrieval system to obtain a candidate patent set of the patent to be converted in different dimensions.

[0020] Furthermore, in the search expression generation unit, execute:

[0021] The technical features of the patent to be converted that appear simultaneously in the technical problem dimension, technical means dimension, and technical effect dimension are regarded as key technical features;

[0022] Determine the similarity of multiple technical features in the technical problem dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical problem dimension, and connect them with an OR relationship;

[0023] Determine the similarity of multiple technical features in the technical means dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical means dimension, and connect them with an OR relationship;

[0024] Determine the similarity of multiple technical features in the technical effect dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical effect dimension, and connect them with the relationship of "or";

[0025] Key technical features are used as necessary keywords for retrieval, and similar features in the technical problem dimension, connected by an OR relationship, and the remaining technical features in the technical problem dimension, are connected using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical problem dimension for the patent to be converted;

[0026] Use key technical features as necessary keywords for retrieval, and connect similar features in the technical means dimension connected by the relationship of "or" and the remaining technical features in the technical means dimension using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical means dimension for the patent to be converted;

[0027] The key technical features are used as necessary keywords for retrieval, and similar features of the technical effect dimension connected by the relationship of "or" and the remaining technical features of the technical effect dimension are connected using a variety of different forms of logical relationships to form a variety of intelligent retrieval expressions for the technical effect dimension of the patent to be converted.

[0028] Furthermore, in the intelligent retrieval unit, execute:

[0029] Use each intelligent search expression of the patent to be converted in the corresponding dimension to conduct patent search in the intelligent search system respectively, and obtain the top M related patents with the highest similarity corresponding to each intelligent search expression;

[0030] The related patents that appear repeatedly in the top M related patents with the highest similarity corresponding to different types of intelligent search expressions are used as preliminary candidate patents for the corresponding dimension, and N preliminary candidate patents with the highest similarity in technical content with the patent to be converted in the corresponding dimension are selected from them as the candidate patent set for the patent to be converted in the corresponding dimension.

[0031] Furthermore, in the intelligent recommendation module for patent conversion and application objects, the following are executed:

[0032] Arrange the similarity of each candidate patent in the candidate patent set of different dimensions of the patent to be converted in descending order and assign corresponding similarity scores;

[0033] Candidate patents that appear in at least two candidate patent sets of different dimensions are selected as co-occurring patents. The similarity scores of each co-occurring patent in each candidate patent set are summed up and normalized to obtain the co-occurrence degree of each co-occurring patent.

[0034] The similarity scores of each co-occurring patent in each candidate patent set are averaged to obtain the average similarity of each co-occurring patent;

[0035] According to the number of different types of operational behaviors implemented by the patentee in each co-occurring patent, the operational degree of the corresponding co-occurring patent is obtained.

[0036] Furthermore, in the intelligent recommendation module for patent transformation and application objects, it is assumed that the weights of co-occurrence, average similarity, and operation degree are α, β, and γ respectively, and α+β+γ=1;

[0037] The recommendation degree R of each co-occurring patent is calculated according to the formula: R = α*DC + β*DS + γ*DO;

[0038] Among them, DC, DS, and DO represent the co-occurrence degree, average similarity, and operation degree of co-occurring patents, respectively.

[0039] Furthermore, different types of operational activities implemented by patent holders include: transfer, licensing, and pledge.

[0040] Furthermore, the weights of the number of transfers, the number of licenses, and the number of pledges are set to a, b, and c respectively, and a+b+c=1;

[0041] For each co-occurring patent, assuming the number of transfers is N1, the number of licenses is N2, and the number of pledges is N3, then a*N1+b*N2+c*N3 is normalized to a value of [1,100] as the operational degree of the current co-occurring patent.

[0042] Furthermore, the multi-dimensional technical summary module includes a preliminary technical summary unit and an output review unit for each dimension; wherein,

[0043] The preliminary technical summary unit of each dimension is used to summarize the technical content of the corresponding dimension in the patent to be converted based on the macro model of the corresponding dimension, and obtain the preliminary technical summary results of the patent to be converted in the corresponding dimension;

[0044] The output review unit of each dimension is used to output and review the preliminary technical summary results of the patent to be converted in the corresponding dimension, and obtain the technical summary results of the patent to be converted in the corresponding dimension.

[0045] Furthermore, the dimensions include technical problem dimension, technical means dimension and technical effect dimension.

[0046] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0047] The intelligent recommendation system for patent transformation and application targets provided by the present invention is implemented by the collaborative efforts of a multi-dimensional technology summary module, a multi-dimensional candidate patent search module, and a patent transformation and application target intelligent recommendation module. In the multi-dimensional technology summary module, a large model is used to obtain technical summaries of the patents to be transformed in different dimensions. After reviewing these technical summaries, technical summaries for the corresponding dimensions are obtained for the patents to be transformed, resulting in highly accurate technical summary results for each dimension. Subsequently, in the multi-dimensional candidate patent search module, feature extraction is performed on the technical summaries for the patents to be transformed in different dimensions. Based on the extracted technical features, intelligent searches are performed to obtain candidate patent sets for the patents to be transformed in the corresponding dimensions, effectively improving the accuracy of the selected candidate patents. Finally, in the patent transformation and application target intelligent recommendation module, quantitative data from different perspectives on co-occurring patents across different dimensions is comprehensively considered to quantify the recommendation level of each co-occurring patent. Ultimately, the co-occurring patent with the highest recommendation level is selected as the patent transformation and application target. In summary, by improving the multi-dimensional technology summary and multi-dimensional patent search processes, the present invention can effectively improve the accuracy of the obtained candidate patent set. On this basis, we creatively proposed a quantitative calculation method for the recommendation degree of co-occurring patents, and ultimately successfully selected the objects of patent transformation and application, effectively solving the problem of low accuracy of the existing method of recommending objects of patent transformation and application.

[0048] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. Throughout the drawings, the same reference symbols denote the same components.

[0050] Figure 1 A schematic diagram of the structure of the intelligent recommendation system for patent conversion and application objects provided by an embodiment of the present invention;

[0051] Figure 2 A structural diagram of another patent conversion application object intelligent recommendation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0053] A specific embodiment of the present invention discloses an intelligent recommendation system for patent transformation and application objects, the structural diagram of which is shown in FIG. Figure 1 and 2 shown.

[0054] (1) Multi-dimensional technology summary module

[0055] The multi-dimensional technical summary module uses a large model to obtain the technical summary of the patent to be converted in different dimensions, and obtains the technical summary results of the patent to be converted in the corresponding dimensions after reviewing the technical summary.

[0056] Specifically, in this embodiment, the dimensions involved include the technical problem dimension, the technical means dimension, and the technical effect dimension. Therefore, in the multi-dimensional technical summary module, technical summaries are performed on the patent to be converted in the technical problem dimension, the technical means dimension, and the technical effect dimension, thereby obtaining technical summary results for the patent to be converted in the technical problem dimension, the technical means dimension, and the technical effect dimension.

[0057] Preferably, in this embodiment, the multi-dimensional technology summary module includes a preliminary technology summary unit for each dimension and an output review unit. The functions of each unit are described in detail below.

[0058] 1) Preliminary technical summary unit for each dimension

[0059] The preliminary technical summary unit of each dimension is used to perform a technical summary of the technical content of the corresponding dimension in the patent to be converted based on the large model of the corresponding dimension, and obtain the preliminary technical summary results of the patent to be converted in the corresponding dimension.

[0060] Preferably, in this embodiment, the technical content of different dimensions of the patent to be converted is different. Specifically, the technical content of the technical problem dimension includes the background technology of the patent to be converted, the technical content of the technical means dimension includes the claims of the patent to be converted, and the technical content of the technical effect dimension includes the beneficial effects of the patent to be converted.

[0061] During the specific implementation process, large models of different dimensions may be implemented in the form of models of the same or different categories. In this embodiment, the large model is preferably implemented by a generative artificial intelligence interaction system. Currently, the more common generative artificial intelligence interaction systems include ChatGPT, Wenxin Yiyan or chatGLM. Before using the large model to predict the output of the actual preliminary technical summary results, it is necessary to train the parameters of the large model through a large amount of training corpus to fit the correspondence between input and output. During the training process of large models of different dimensions, it is necessary to select training expectations of the corresponding dimensions for model training. For example, for a large model of the technical problem dimension, it is necessary to select background technologies in multiple patent applications as training expectations for training; for a large model of the technical means dimension, it is necessary to select claims in multiple patent applications as training expectations for training; for a large model of the technical effect dimension, it is necessary to select beneficial effects in multiple patent applications as training expectations for training.

[0062] At the same time, in this embodiment, the large model of each dimension is provided with various forms of interactive input templates of the corresponding dimension; when making a technical summary of the technical content of the corresponding dimension in the patent to be converted, the technical content of the corresponding dimension in the patent to be converted is placed in an interactive input template of the corresponding dimension as the input information of the large model of the corresponding dimension, and the large model of the corresponding dimension processes the input information to obtain a preliminary technical summary result of the corresponding dimension.

[0063] In a certain example, the following method can be used to set the functions of the preliminary technical summary unit for the technical problem dimension, the technical means dimension, and the technical effect dimension respectively.

[0064] a) Preliminary technical summary unit of technical problem dimension

[0065] The preliminary technical summary unit of the technical problem dimension is used to perform a technical summary of the background technology in the patent to be converted based on the large model of the technical problem dimension, and obtain the preliminary technical summary results of the patent to be converted in the technical problem dimension.

[0066] It should be noted that the background technology of the patent to be converted contains technical problems. In order to ensure the accuracy of subsequent feature extraction and intelligent retrieval and reduce the workload of subsequent processes, it is necessary to summarize the technical problems in the background technology.

[0067] Preferably, in the preliminary technical summary unit of the technical problem dimension, a large model of the technical problem dimension is built in. Various forms of interactive input templates of the technical problem dimension are provided in the large model of the technical problem dimension. By setting various forms of interactive input templates of the technical problem dimension, the output results of the large model of the technical problem dimension can be controlled to a certain extent. Exemplarily, an interactive input template of the technical problem dimension can be expressed as: summarizing the technical problems of the "background technology". Here, "background technology" refers to the background technology in the patent to be converted. In addition, in other interactive input templates of the technical problem dimension, further limitations can be made on the basis of the above-mentioned interactive input templates in terms of the number of words in the technical problem summary, the included keywords, etc., so as to obtain various forms of interactive input templates of the technical problem dimension.

[0068] During the specific implementation process, the background technology of the patent to be converted is placed in an interactive input template in the form of the technical problem dimension as the input information of the big model of the technical problem dimension. After the big model of the technical problem dimension processes the input information, a preliminary technical summary result of the technical problem dimension is obtained.

[0069] b) Preliminary technical summary unit of technical means dimension

[0070] The preliminary technical summary unit of the technical means dimension is used to perform a technical summary of the claims in the patent to be converted based on the large model of the technical means dimension, and obtain the preliminary technical summary results of the patent to be converted in the technical means dimension.

[0071] Preferably, in the preliminary technical summary unit of the technical means dimension, a large model of the technical means dimension is built in. Various forms of interactive input templates of the technical means dimension are provided in the large model of the technical means dimension. By setting various forms of interactive input templates of the technical means dimension, the output results of the large model of the technical means dimension can be controlled to a certain extent. Exemplarily, an interactive input template of the technical means dimension can be expressed as: summarizing the technical means of the "claims". Here, "claims" refers to the claims in the patent to be converted. In addition, in other interactive input templates of the technical means dimension, further limitations can be made on the basis of the above-mentioned interactive input templates in terms of the number of words in the technical means summary, the included keywords, etc., so as to obtain various forms of interactive input templates of the technical means dimension.

[0072] During the specific implementation process, the claims of the patent to be converted are placed in an interactive input template in the technical means dimension as input information of the big model of the technical means dimension. After the big model of the technical means dimension processes the input information, a preliminary technical summary result of the technical means dimension is obtained.

[0073] c) Preliminary technical summary unit of technical effect dimension

[0074] The preliminary technical summary unit of the technical effect dimension is used to conduct a technical summary of the beneficial effects in the patent to be transformed based on the large model of the technical effect dimension, and obtain the preliminary technical summary results of the patent to be transformed in the technical effect dimension.

[0075] It should be noted that, during the specific implementation process, a technical effect summary can be conducted on the beneficial effects of the patent to be transformed (and / or specific phrases containing vocabulary representing technical effects) to obtain a preliminary technical summary result of the patent to be transformed in the technical effect dimension.

[0076] Preferably, in the preliminary technical summary unit of the technical effect dimension, a large model of the technical effect dimension is built in. Various forms of interactive input templates of the technical effect dimension are provided in the large model of the technical effect dimension. By setting various forms of interactive input templates of the technical effect dimension, the output results of the large model of the technical effect dimension can be controlled to a certain extent. Exemplarily, an interactive input template of the technical effect dimension can be expressed as: summarizing the technical effects of "beneficial effects", where "beneficial effects" refer to the beneficial effects in the patent to be converted. In addition, in other interactive input templates of the technical effect dimension, further limitations can be made on the basis of the above-mentioned interactive input templates in terms of the number of words in the technical effect summary, the included keywords, etc., so as to obtain various forms of interactive input templates of the technical effect dimension.

[0077] During the specific implementation process, the beneficial effects of the patent to be converted are placed in an interactive input template in the form of the technical effect dimension as input information of the large model of the technical effect dimension. After the large model of the technical effect dimension processes the input information, a preliminary technical summary result of the technical effect dimension is obtained.

[0078] 2) Output review unit for each dimension

[0079] The output review unit of each dimension is used to output and review the preliminary technical summary results of the patent to be converted in the corresponding dimension, and obtain the technical summary results of the patent to be converted in the corresponding dimension.

[0080] Preferably, in this embodiment, the output review unit of each dimension is used to review whether the technical summary result of the corresponding dimension is technical-related information of the patent to be converted. If so, the output review is passed, and the preliminary technical summary result of the corresponding dimension is used as the technical summary result of the patent to be converted in the corresponding dimension; otherwise, the output review fails, and the preliminary technical summary unit of the corresponding dimension is returned, and the interactive input template of the corresponding dimension is replaced, and the technical content of the corresponding dimension in the patent to be converted is placed in the replaced interactive input template of the corresponding dimension, and again used as input information of the large model of the corresponding dimension. The large model of the corresponding dimension processes the input information to obtain the preliminary technical summary result of the corresponding dimension, and then the output review unit of the corresponding dimension performs output review until the output review is passed.

[0081] During the specific implementation process, the output review unit of each dimension has a built-in output review controller and multiple output reviewers of the corresponding dimension; wherein, the output review controller is used to control whether each output reviewer is effective, and to control the logical relationship between the multiple effective output reviewers. Each output reviewer, from its own review perspective, reviews whether the preliminary technical summary results of the corresponding dimension are technical-related information of the patent to be converted. Exemplarily, the output reviewer can be one of a technology classifier, a tendency classifier, an output approval keyword whitelist, and an output approval keyword blacklist. During the specific implementation process, the output review controller adjusts the review intensity of the output review unit of the corresponding dimension by controlling the effectiveness of different output reviewers and controlling the logical relationship between the effective output reviewers.

[0082] The following is a detailed description of the technical classifier, tendency classifier, output approval keyword whitelist, and output approval keyword blacklist.

[0083] Preferably, in this embodiment, both the technology classifier and the tendency classifier are obtained based on neural network training. The technology classifier performs a binary classification judgment on the preliminary technology summary results, determining whether they are technical statements or not. The tendency classifier performs a binary classification judgment on the preliminary technology summary results, determining whether they contain undesirable tendencies or not. Exemplarily, the technology classifier and the tendency classifier can be obtained based on a general-purpose CNN neural network model. During implementation, for the technology classifier, a large amount of input information and its technical labels can be generated in advance. Exemplarily, 0 and 1 labels can be designed, where 1 represents the input information as a technical statement and 0 represents the input information as not a technical statement, thereby forming a technology training sample set. Based on the technology training sample set, the neural network corresponding to the technology classifier undergoes multiple rounds of training. After successful training, a technology classifier is obtained that can be used for binary classification judgments on actual technical statements. Similar to the technology classifier, the tendency classifier can also be trained in advance with a large amount of input information and its tendency labels. Exemplarily, 0 and 1 labels can be designed, where 1 represents the input information as not containing undesirable tendencies and 0 represents the input information as containing undesirable tendencies, thereby forming a tendency training sample set. Based on the propensity training sample set, the neural network corresponding to the propensity classifier undergoes multiple rounds of training. After passing the training, a propensity classifier is obtained that can be used for actual propensity binary classification. During the specific implementation process, to improve the efficiency of training and model application, it is possible to consider using the same technical classifier and propensity classifier for output review units of different dimensions.

[0084] Preferably, considering that the output approval keyword whitelist and output approval keyword blacklist of different dimensions have different focuses, it is necessary to specifically set the output approval keyword whitelist and output approval keyword blacklist of different dimensions according to the content of the patent to be converted. For example, the output approval keyword whitelist of the technical problem dimension may include the core vocabulary of the technical problem solved by the patent to be converted. The output approval keyword blacklist of the technical problem dimension may include the relevant vocabulary of common technical problems that the patent to be converted does not involve in the relevant technical field. The output approval keyword whitelist of the technical means dimension may include the core vocabulary of the technical means of the patent to be converted. The output approval keyword blacklist of the technical means dimension may include the relevant vocabulary of common technical means that the patent to be converted does not involve in the relevant technical field. In addition, the output approval keyword whitelist of the technical effect dimension may include the core vocabulary of the technical effect of the patent to be converted. The output approval keyword blacklist of the technical effect dimension may include the relevant vocabulary of common technical effects that the patent to be converted does not involve in the relevant technical field.

[0085] Based on the above description, it can be seen that for each output reviewer, it is possible to review whether the technical summary results of the corresponding dimension are technical-related information of the patent to be converted from the corresponding review perspective. That is, for the technical classifier, it can be judged whether the preliminary technical summary result is a technical statement. For example, 1 is a technical statement, that is, technical-related information of the patent to be converted; 0 is not a technical statement, that is, it is not technical-related information of the patent to be converted. For the tendency classifier, it can be judged whether the preliminary technical summary result contains bad tendencies. 1 means it does not contain bad tendencies, that is, it is technical-related information of the patent to be converted; 0 means it contains bad tendencies, that is, it is not technical-related information of the patent to be converted. For the output approval keyword whitelist, it can be judged whether the preliminary technical summary result contains words in the output approval keyword whitelist. For example, 1 means it contains, that is, it is technical-related information of the patent to be converted; 0 means it does not contain, that is, it is not technical-related information of the patent to be converted. For the output approval keyword blacklist, it can be judged whether the preliminary technical summary result contains words in the output approval keyword blacklist. For example, 1 means it does not contain, that is, it is technical-related information of the patent to be converted; 0 means it contains, that is, it is not technical-related information of the patent to be converted.

[0086] The output audit controller can control different output auditors to be effective, and only the effective output auditors can audit whether the technical summary results of the corresponding dimension are technical related information of the patent to be converted. In addition, the output audit controller can also adjust the audit strength of the output audit unit of the corresponding dimension by controlling the logical relationship between the effective output auditors. For example, when the output audit controller controls the above four output auditors to be effective and the logical relationship is "and", only when the judgment results of the above four output auditors are all 1, the output audit passes; otherwise, the output audit fails. When the output audit controller controls the above four output auditors to be effective and the logical relationship is "or", the judgment result of any output auditor is 1, the output audit passes; only when the judgment results of the four output auditors are all 0, the output audit fails. Therefore, by controlling the effectiveness and logical relationship of the output auditors by the output audit controller, the purpose of adjusting the audit strength of the output audit unit can be achieved.

[0087] (2) Multi-dimensional candidate patent search module

[0088] The multi-dimensional candidate patent search module extracts features from the technical summary results of the patents to be converted in different dimensions, and performs intelligent searches based on the extracted technical features to obtain a set of candidate patents in the corresponding dimensions of the patents to be converted.

[0089] That is, the multi-dimensional candidate patent search module extracts features from the technical summary results of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension, and performs intelligent retrieval based on the extracted technical features to obtain a set of candidate patents for the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension.

[0090] Preferably, the multi-dimensional candidate patent search module includes: a technical feature extraction unit for each dimension, a search expression generation unit, and an intelligent search unit. The specific functions are described as follows.

[0091] 1) Technical feature extraction unit for each dimension

[0092] The technical feature extraction unit of each dimension extracts features from the technical summary results of the patent to be transformed in the corresponding dimension, and obtains the technical features of the patent to be transformed in the corresponding dimension.

[0093] Specifically, in the technical feature extraction unit of the technical problem dimension, feature extraction is performed on the technical summary results of the patent to be converted in the technical problem dimension to obtain the technical features of the patent to be converted in the technical problem dimension. In the technical feature extraction unit of the technical means dimension, feature extraction is performed on the technical summary results of the patent to be converted in the technical means dimension to obtain the technical features of the patent to be converted in the technical means dimension. In the technical feature extraction unit of the technical effect dimension, feature extraction is performed on the technical summary results of the patent to be converted in the technical effect dimension to obtain the technical features of the patent to be converted in the technical effect dimension.

[0094] Preferably, in this embodiment, the TF-IDF algorithm can be combined with a key dictionary to perform feature extraction on the technical summary results of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension, respectively, so as to obtain the technical features of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension; or, feature extraction is performed on the technical summary results of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension, respectively, and after obtaining the corresponding feature extraction results, the corresponding feature extraction results are further associated and inferred by hypernym elements to summarize the technical feature hypernyms of the corresponding feature extraction results, and the feature extraction results and technical feature hypernyms of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension are respectively summarized to form the technical features of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension. It should be noted that in this embodiment, the technical features of the patent to be converted in the technical problem dimension, technical means dimension, and technical effect dimension are embodied in the form of feature words.

[0095] During specific implementations, deep learning models, such as convolutional neural networks (CNNs), can be used to associate and infer hypernyms of each technical feature extraction result to obtain hypernyms of the corresponding feature extraction results. Through intelligent retrieval based on the summarized hypernyms of technical features using these deep learning models, it is possible to effectively identify patents that have different patent expressions but the same hypernym (i.e., the same essential content).

[0096] 2) Search expression generation unit

[0097] The search expression generation unit comprehensively analyzes the technical features of the patent to be converted in different dimensions and obtains multiple intelligent search expressions of the patent to be converted in different dimensions.

[0098] Preferably, the search expression generation unit can combine the format requirements of the intelligent search engine for the search expression and adopt the following method to obtain the intelligent search expression of the patent to be converted in the technical problem dimension, technical means dimension and technical effect dimension:

[0099] The technical features of the patent to be converted that appear simultaneously in the technical problem dimension, technical means dimension, and technical effect dimension are regarded as key technical features;

[0100] Determine the similarity of multiple technical features in the technical problem dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical problem dimension, and connect them with an OR relationship;

[0101] Determine the similarity of multiple technical features in the technical means dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical means dimension, and connect them with an OR relationship;

[0102] Determine the similarity of multiple technical features in the technical effect dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical effect dimension, and connect them with the relationship of "or";

[0103] Key technical features are used as necessary keywords for retrieval, and similar features in the technical problem dimension, connected by an OR relationship, and the remaining technical features in the technical problem dimension, are connected using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical problem dimension for the patent to be converted;

[0104] Use key technical features as necessary keywords for retrieval, and connect similar features in the technical means dimension connected by the relationship of "or" and the remaining technical features in the technical means dimension using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical means dimension for the patent to be converted;

[0105] The key technical features are used as necessary keywords for retrieval, and similar features of the technical effect dimension connected by the relationship of "or" and the remaining technical features of the technical effect dimension are connected using a variety of different forms of logical relationships to form a variety of intelligent retrieval expressions for the technical effect dimension of the patent to be converted.

[0106] For example, the various logical connections employed here can include logical relationships such as "and" and "or," thereby enriching the forms of intelligent search expressions and more closely resembling the actual search habits of technical personnel. This approach enables the formation of multiple intelligent search expressions for patents to be converted in the dimensions of technical issues, technical means, and technical effects, ensuring that the intelligent search expressions encompass key technical features while also ensuring a richness of the retrieved patents.

[0107] 3) Intelligent retrieval unit

[0108] The intelligent retrieval unit uses a variety of intelligent retrieval expressions of the patent to be converted in different dimensions to perform patent retrieval in the intelligent retrieval system to obtain a candidate patent set of the patent to be converted in different dimensions.

[0109] Preferably, in the intelligent retrieval unit, a plurality of intelligent retrieval expressions of the patent to be converted in the technical problem dimension, technical means dimension and technical effect dimension are used to perform patent retrieval in the intelligent retrieval system, and the similar patents of the patent to be converted in the technical problem dimension, technical means dimension and technical effect dimension are respectively used as the candidate patent sets of the patent to be converted in the technical problem dimension, technical means dimension and technical effect dimension.

[0110] Specifically, in the intelligent retrieval unit, for each dimension, the candidate patent set of the patent to be converted in the corresponding dimension is obtained in the following way:

[0111] Use each intelligent search expression of the patent to be converted in the corresponding dimension to perform patent search in the intelligent search system respectively, and obtain the top M related patents with the highest similarity corresponding to each intelligent search expression.

[0112] The related patents that appear repeatedly in the top M related patents with the highest similarity corresponding to different types of intelligent search expressions are used as preliminary candidate patents for the corresponding dimension, and N preliminary candidate patents with the highest similarity in technical content with the patent to be converted in the corresponding dimension are selected from them as the candidate patent set for the patent to be converted in the corresponding dimension.

[0113] During implementation, by performing the above operations for the technical problem dimension, technical means dimension, and technical effect dimension, a candidate patent set for the patent to be converted in these dimensions can be obtained. For example, M is greater than N. During implementation, considering that the number of patent conversion and application targets ultimately selected is relatively small, or even only one, the value of N should not be too large. For example, N is set to 50. The value of N can also be adaptively set based on the number of preliminary candidate patents.

[0114] (3) Intelligent recommendation module for patent transformation and application objects

[0115] The intelligent recommendation module for patent conversion and application objects obtains the co-occurrence degree, average similarity and operation degree of co-occurring patents among the candidate patent sets of different dimensions of the patent to be converted to determine the recommendation degree of each co-occurring patent; and selects the co-occurring patent with the largest recommendation degree as the patent conversion and application object.

[0116] In the patent conversion application object intelligent recommendation module, the following processing is performed on the candidate patent sets of different dimensions of the patents to be converted:

[0117] 1) Calculate the co-occurrence degree DC of co-occurring patents

[0118] The similarity of each candidate patent in the candidate patent set of different dimensions (i.e., the similarity of each candidate patent with the patent to be converted in the corresponding dimension) is arranged in descending order and assigned a corresponding similarity score. For example, in each candidate patent set, the candidate patent with the highest similarity score is 50 points, and so on, the candidate patent with the lowest similarity score is 1 point.

[0119] Candidate patents that appear in at least two candidate patent sets in different dimensions are selected as co-occurring patents. The similarity scores of each co-occurring patent in each candidate patent set are summed up and normalized to obtain the co-occurrence degree DC of each co-occurring patent.

[0120] For example, the maximum value of the sum of similarity scores of co-occurring patents in each candidate patent set is 150, and is normalized to a value of [1, 100], thereby obtaining the co-occurrence degree of each co-occurring patent.

[0121] 2) Calculate the average similarity DS of co-occurring patents

[0122] The similarity scores of each co-occurring patent in each candidate patent set are averaged to obtain the average similarity of each co-occurring patent.

[0123] It should be noted that only co-occurring patents are included in the calculation. That is, if a candidate patent only appears in two candidate patent sets, only the average similarity of the two candidate patent sets is calculated.

[0124] The maximum value of the similarity score is 100, and the maximum value of the average similarity DS is also 100.

[0125] 3) Calculate the operational degree (DO) of co-occurring patents

[0126] According to the number of different types of operational behaviors implemented by the patentee in each co-occurring patent, the operational degree of the corresponding co-occurring patent is obtained.

[0127] Specifically, different types of operational activities performed by patent holders include transfers, licenses, and pledges. The weights for the number of transfers, licenses, and pledges are set to a, b, and c, respectively, where a + b + c = 1. For each co-occurring patent, assuming the number of transfers is N1, the number of licenses is N2, and the number of pledges is N3, the operational degree (DO) of the current co-occurring patent is obtained by normalizing a*N1+b*N2+c*N3 to a value between 1 and 100.

[0128] 4) Calculate the recommendation degree R of co-occurring patents

[0129] Assume that the weights of co-occurrence, average similarity, and operation degree are α, β, and γ respectively, and satisfy α+β+γ=1.

[0130] The recommendation degree R of each co-occurring patent is calculated according to the formula: R = α*DC+β*DS+γ*DO.

[0131] Afterwards, the co-occurring patents with the highest recommendation degree are selected as the objects of patent transformation and application.

[0132] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0133] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. An intelligent recommendation system for patent conversion and application objects, characterized by: The system comprises: A multi-dimensional technical summary module uses a large model to obtain technical summaries of the patent to be converted in different dimensions, and after reviewing the technical summaries, obtains technical summary results of the patent to be converted in the corresponding dimensions; the different dimensions include the technical problem dimension, the technical means dimension, and the technical effect dimension; the multi-dimensional technical summary module includes an output review unit for each dimension, a built-in output review controller and multiple output reviewers, and the output review controller adjusts the review intensity of the output review unit of the corresponding dimension by controlling the effectiveness of different output reviewers and controlling the logical relationship between the effective output reviewers; The multi-dimensional candidate patent search module extracts features from the technical summary results of the patents to be converted in different dimensions, and performs intelligent searches based on the extracted technical features to obtain a set of candidate patents in the corresponding dimensions of the patents to be converted; The patent conversion and application object intelligent recommendation module obtains the co-occurrence degree, average similarity and operation degree of the co-occurring patents among the candidate patent sets of different dimensions of the patent to be converted to determine the recommendation degree of each co-occurring patent; and selects the co-occurring patent with the highest recommendation degree as the patent conversion and application object. In the patent conversion and application object intelligent recommendation module, the following are executed: Arrange the similarity of each candidate patent in the candidate patent set of different dimensions of the patent to be converted in descending order and assign corresponding similarity scores; Candidate patents that appear in at least two candidate patent sets of different dimensions are selected as co-occurring patents. The similarity scores of each co-occurring patent in each candidate patent set are summed up and normalized to obtain the co-occurrence degree of each co-occurring patent. The multi-dimensional candidate patent search module includes: a technical feature extraction unit for each dimension, a search expression generation unit, and an intelligent search unit; In the Smart Search Unit, execute: Use each intelligent search expression of the patent to be converted in the corresponding dimension to conduct patent search in the intelligent search system respectively, and obtain the top M related patents with the highest similarity corresponding to each intelligent search expression; The related patents that appear repeatedly in the top M related patents with the highest similarity corresponding to different types of intelligent search expressions are used as preliminary candidate patents for the corresponding dimension, and N preliminary candidate patents with the highest similarity in technical content with the patent to be converted in the corresponding dimension are selected from them as the candidate patent set for the patent to be converted in the corresponding dimension.

2. The patent conversion and application object intelligent recommendation system according to claim 1 is characterized in that: The technical feature extraction unit of each dimension extracts features from the technical summary results of the patent to be converted in the corresponding dimension, and obtains the technical features of the patent to be converted in the corresponding dimension; The search expression generation unit comprehensively analyzes the technical features of the patent to be converted in different dimensions and obtains multiple intelligent search expressions for the patent to be converted in different dimensions; The intelligent retrieval unit uses a variety of intelligent retrieval expressions of the patent to be converted in different dimensions to perform patent retrieval in the intelligent retrieval system to obtain a candidate patent set of the patent to be converted in different dimensions.

3. The patent conversion and application object intelligent recommendation system according to claim 2 is characterized in that: In the search expression generation unit, execute: The technical features of the patent to be converted that appear simultaneously in the technical problem dimension, technical means dimension, and technical effect dimension are regarded as key technical features; Determine the similarity of multiple technical features in the technical problem dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical problem dimension, and connect them with an OR relationship; Determine the similarity of multiple technical features in the technical means dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical means dimension, and connect them with an OR relationship; Determine the similarity of multiple technical features in the technical effect dimension of the patent to be converted, and take multiple technical features whose similarity exceeds a certain similarity threshold as similar features in the technical effect dimension, and connect them with the relationship of "or"; Key technical features are used as necessary keywords for retrieval, and similar features in the technical problem dimension, connected by an OR relationship, and the remaining technical features in the technical problem dimension, are connected using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical problem dimension for the patent to be converted; Use key technical features as necessary keywords for retrieval, and connect similar features in the technical means dimension connected by the relationship of "or" and the remaining technical features in the technical means dimension using various forms of logical relationships to form multiple intelligent retrieval expressions in the technical means dimension for the patent to be converted; The key technical features are used as necessary keywords for retrieval, and similar features of the technical effect dimension connected by the relationship of "or" and the remaining technical features of the technical effect dimension are connected using a variety of different forms of logical relationships to form a variety of intelligent retrieval expressions for the technical effect dimension of the patent to be converted.

4. The intelligent recommendation system for patent transformation and application objects according to any one of claims 1 to 3, characterized in that: In the patent transformation and application object intelligent recommendation module, the following are also performed: The similarity scores of each co-occurring patent in each candidate patent set are averaged to obtain the average similarity of each co-occurring patent; According to the number of different types of operational behaviors implemented by the patentee in each co-occurring patent, the operational degree of the corresponding co-occurring patent is obtained.

5. The patent conversion and application object intelligent recommendation system according to claim 4 is characterized in that: In the patent transformation and application object intelligent recommendation module, it is assumed that the weights of co-occurrence, average similarity, and operation degree are α, β, and γ respectively, and α+β+γ=1; The recommendation degree R of each co-occurring patent is calculated according to the following formula: ; Among them, DC, DS, and DO represent the co-occurrence degree, average similarity, and operation degree of co-occurring patents, respectively.

6. The patent conversion and application object intelligent recommendation system according to claim 4 is characterized in that: Different types of operations implemented by patent holders include: transfer, licensing, and pledge.

7. The patent conversion and application object intelligent recommendation system according to claim 6 is characterized in that: Set the weights of the number of transfers, the number of permissions, and the number of pledges to be a, b, and c respectively, and satisfy a+b+c=1; For each co-occurring patent, assuming the number of transfers is N1, the number of licenses is N2, and the number of pledges is N3, then The value normalized to [1,100] is used as the operational degree of the current co-occurring patent.

8. The patent conversion and application object intelligent recommendation system according to claim 4 is characterized in that: The multi-dimensional technology summary module also includes a preliminary technology summary unit for each dimension; wherein, The preliminary technical summary unit of each dimension is used to perform a technical summary of the technical content of the corresponding dimension in the patent to be converted based on the large model of the corresponding dimension, and obtain the preliminary technical summary results of the patent to be converted in the corresponding dimension.

Images