A patent value assessment method and system based on deep learning

Abstract

The present invention is a patent value assessment method and system based on deep learning, which relates to the field of patent assessment technology, wherein the method comprises the following steps: obtaining a patent indicator data set; extracting semantic features of the patent indicator data set using a pre-trained large language model to generate a text vector; completing the indicators of an unlabeled patent data set based on a semi-supervised learning paradigm, and constructing a training set and a test set; constructing an evaluation model based on deep learning, iteratively training the evaluation model using the training set, and evaluating the model based on the test set and preset performance evaluation indicators to obtain a trained evaluation model; obtaining patent indicator data to be evaluated, inputting the patent indicator data to be evaluated into the trained evaluation model, and obtaining a predicted score for the patent indicator data to be evaluated; and obtaining an evaluation result of the patent indicator data to be evaluated based on the predicted score.

Description

Technical Field

[0001] This application relates to the field of patent evaluation technology, and mainly to a patent value evaluation method and system based on deep learning. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, patent valuation methods are also evolving. Patent valuation is of great significance to a company's technological layout, market competition strategies, and the government's formulation of scientific and technological innovation policies. However, traditional patent valuation methods, which rely primarily on expert review, citation analysis, or simple statistical indicators, are subject to high subjectivity, low efficiency, and difficulty processing large amounts of data. Especially in the context of the rapid growth in the number of patents worldwide, manual valuation methods are no longer able to meet the demand for efficient and accurate evaluation.

[0003] Existing methods are currently struggling to adapt to the needs of actual enterprise patent valuation scenarios. The main reasons are: existing methods mostly focus on small-scale models of a single technical dimension, such as decision trees, support vector machines, and random forests. However, these methods rely on manually pre-processed structured data and are unable to directly mine deep semantic information from patent texts (such as abstracts and claims). Furthermore, in real-world scenarios, a large amount of patent data lacks manual annotation, resulting in the generalization ability of supervised learning models being limited due to insufficient training data. Furthermore, in real-world scenarios, due to the high cost of manual annotation, some patent data lacks labels and cannot be directly used for training and prediction.

[0004] Chinese patent publication number CN108416535A discloses a method for patent value assessment based on deep learning, including: "obtaining the attribute characteristics and citation relationships of the patent, and obtaining the corresponding attribute network representation of the patent through the attribute network representation method; obtaining the text information of the patent, and obtaining the corresponding text representation of the patent through a multi-layer convolutional neural network; using the corresponding attribute network representation and text representation of the patent to evaluate and predict the value of the patent." However, this method only obtains the attribute characteristics, citation relationships and text information of the patent, and does not fully utilize this key information that may affect the value of the patent, resulting in an incomplete evaluation result. In addition, the method obtains the corresponding attribute network representation of the patent through the attribute network representation method. This method has difficulty capturing the complex nonlinear relationships between patent attributes, affecting the accuracy of the evaluation. At the same time, the method does not consider the situation in which some input data labels are missing in real scenarios, and the comprehensive utilization of labeled and unlabeled data is not efficient enough. In addition, the method uses a relatively simple neural network model to extract text information, rather than a large language model with powerful expression capabilities, which results in insufficient and inaccurate information mining of the patent text content. Summary of the Invention

[0005] In order to solve the above-mentioned problems existing in the prior art, this application provides a patent value assessment method and system based on deep learning.

[0006] The technical solution of this application is as follows:

[0007] In one aspect, the present invention proposes a patent value assessment method based on deep learning, the method comprising:

[0008] Obtaining a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically refers to patents that include both manual indicators and machine indicators, and the unannotated patent dataset specifically refers to patents that include machine indicators but lack manual indicators;

[0009] The pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate text vectors. The unlabeled patent dataset is supplemented with indicators based on the semi-supervised learning paradigm, specifically:

[0010] The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows:

[0011] ;

[0012] Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data; Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data;

[0013] Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets;

[0014] Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on the supplemented artificial indicators, which can be expressed as follows:

[0015] ;

[0016] ;

[0017] ;

[0018] Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on completed manual indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on the completed manual indicators;

[0019] The machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data is calculated and expressed as follows:

[0020] ;

[0021] Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data;

[0022] Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ;

[0023] Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows:

[0024] ;

[0025] ;

[0026] Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators;

[0027] Perform weighted calculations on the pseudo labels based on manual and machine indicators and their corresponding confidence levels to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, and construct training and test sets;

[0028] Constructing an evaluation model based on deep learning, iteratively training the evaluation model using the training set, and evaluating the evaluation model based on the test set and preset performance evaluation indicators to obtain a trained evaluation model;

[0029] Obtain patent indicator data to be evaluated, input the patent indicator data to be evaluated into a trained evaluation model to obtain a predicted score for the patent indicator data to be evaluated; and obtain an evaluation result of the patent indicator data to be evaluated based on the predicted score.

[0030] Preferably, the method further comprises performing data cleaning on the patent indicator dataset, wherein the data cleaning comprises processing missing values, outliers and standardizing the data format.

[0031] Preferably, a pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as follows:

[0032] ;

[0033] ;

[0034] Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the The index value of unlabeled patent data.

[0035] Preferably, the weighted calculation of the pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels is expressed as follows:

[0036] ;

[0037] ;

[0038] Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight.

[0039] Preferably, the specific steps of constructing the training set and the test set are:

[0040] Filter high confidence sets based on the preset confidence threshold ,in Indicates the preset confidence threshold;

[0041] The training set is specifically a combination of a high confidence set and an annotated patent data set, which is expressed as ,in represents a collection of annotated patent data;

[0042] The test set is specifically an unlabeled patent dataset after index completion, which is expressed as .

[0043] On the other hand, the present invention also proposes a patent value assessment system based on deep learning, which includes a data acquisition module, an indicator completion module, a patent assessment module, and a result output module, wherein:

[0044] The data acquisition module is used to acquire a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically includes patents that include manual indicators and machine indicators, and the unannotated patent dataset specifically includes patents that include machine indicators but lack manual indicators; and transmit the patent indicator dataset to the indicator completion module;

[0045] The indicator completion module is used to extract the semantic features of the patent indicator dataset using a pre-trained large language model and generate text vectors; it completes the indicators of the unlabeled patent dataset based on a semi-supervised learning paradigm, specifically:

[0046] The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows:

[0047] ;

[0048] Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data; Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data;

[0049] Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets;

[0050] Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on the supplemented artificial indicators, which can be expressed as follows:

[0051] ;

[0052] ;

[0053] ;

[0054] Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on completed manual indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on the completed manual indicators;

[0055] The machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data is calculated and expressed as follows:

[0056] ;

[0057] Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data,

[0058] Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ;

[0059] Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows:

[0060] ;

[0061] ;

[0062] Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators;

[0063] Perform weighted calculations on the pseudo labels based on manual and machine indicators and their corresponding confidence levels to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, and construct training and test sets;

[0064] The patent evaluation module is used to construct an evaluation model based on deep learning, iteratively train the evaluation model using the training set, and evaluate it based on the test set and preset performance evaluation indicators to obtain a trained evaluation model;

[0065] Obtaining patent indicator data to be evaluated, inputting the patent indicator data to be evaluated into a trained evaluation model to obtain a predicted score for the patent indicator data to be evaluated; and obtaining an evaluation result for the patent indicator data to be evaluated based on the predicted score;

[0066] The result output module is used to display the evaluation results of the patent indicator data to be evaluated.

[0067] Preferably, a pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as follows:

[0068] ;

[0069] ;

[0070] Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the The index value of unlabeled patent data.

[0071] Preferably, the weighted calculation of the pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels is expressed as follows:

[0072] ;

[0073] ;

[0074] Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight;

[0075] Filter high confidence sets based on the preset confidence threshold ,in Indicates the preset confidence threshold;

[0076] The training set is specifically a high confidence set and an annotated patent data set, and the combination is expressed as ,in represents a collection of annotated patent data;

[0077] The test set is specifically an unlabeled patent dataset after index completion, which is expressed as .

[0078] On the other hand, the present invention also proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a patent value assessment method based on deep learning as described in an embodiment of the present invention.

[0079] On the other hand, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a patent value assessment method based on deep learning as described in an embodiment of the present invention.

[0080] Compared with the prior art, the present invention has the following beneficial effects:

[0081] 1) This invention provides a patent value assessment method and system based on deep learning. It uses a pre-trained large language model to extract semantic features from patent index datasets and generate text vectors, thereby enhancing the depth of mining patent text information and more accurately capturing the deep semantic information in the text.

[0082] 2) This invention provides a patent value assessment method and system based on deep learning. Through a semi-supervised learning paradigm, it improves the overall utilization efficiency of patent data, and supplements the indicators of unlabeled patent datasets, increasing the scale and diversity of training data and improving the generalization ability of the model.

[0083] 3) The present invention provides a patent value assessment method and system based on deep learning, constructs an assessment model based on deep learning, enhances the model's ability to handle complex nonlinear relationships, and improves the accuracy and reliability of the assessment results; through iterative training and a test set-based assessment mechanism, the model parameters are continuously optimized, thereby enhancing the model's stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0085] Figure 2 Schematic diagram of the impact of indicator weights on the expert mixture model according to an embodiment of the present invention;

[0086] Figure 3 Schematic diagram of the influence of the confidence indicator weight on the expert mixture model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0087] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0088] The present invention provides the following technical solution: a patent value assessment method and system based on deep learning.

[0089] Example 1:

[0090] See Figure 1 This embodiment provides a patent value assessment method based on deep learning, and the specific steps include:

[0091] S1. Obtain a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically includes patents that include manual indicators and machine indicators, and the unannotated patent dataset specifically includes patents that include machine indicators but lack manual indicators;

[0092] In this embodiment, the patent indicator dataset includes 10,000 patent data items, of which 5,000 items are labeled patent data items and 5,000 items are unlabeled patent data items. Each piece of patent data consists of 20 indicators, including 9 manual indicators and 11 machine indicators.

[0093] The artificial indicators include innovation, social benefits, application prospects, technological maturity, industry influence, policy compliance, technological compatibility, international competitiveness and industrial chain driving force;

[0094] The machine indicators include patent life, number of right holders, number of patent families, number of patent lawsuits, patent authorization time, number of patent citations, number of patent conversions, patent ranking, number of patent-related projects, patent authorization time and market share of patent technology applications;

[0095] The method further includes performing data cleaning on the patent indicator dataset, wherein the data cleaning includes processing missing values, outliers and standardizing the data format;

[0096] S2. Use the pre-trained large language model to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as:

[0097] ;

[0098] ;

[0099] Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the Index value of unannotated patent data;

[0100] In this embodiment, the pre-trained large language model For the ChatGLM3-6B model, the preset sentence embedding model It is the text2vec-base-chinese model;

[0101] S3, completing indicators of unlabeled patent datasets based on semi-supervised learning paradigm;

[0102] The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows:

[0103] ;

[0104] Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data;

[0105] Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets;

[0106] Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on artificial indicators, which can be expressed as follows:

[0107] ;

[0108] ;

[0109] ;

[0110] Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on artificial indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on manual indicators;

[0111] Calculate the machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data, which can be expressed as

[0112] ;

[0113] Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data;

[0114] Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ;

[0115] Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows:

[0116] ;

[0117] ;

[0118] Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators;

[0119] The pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels are weighted to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, which can be expressed as follows:

[0120] ;

[0121] ;

[0122] Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight;

[0123] In this embodiment, the preset indicator weights Set in [0, 1], increase the step size to 0.1; the preset number of selected is 4;

[0124] S4, and build training sets and test sets, and filter high confidence sets according to the preset confidence threshold ,in Indicates the preset confidence threshold;

[0125] The training set is specifically a combination of a high confidence set and an annotated patent data set, which is expressed as ,in represents a collection of annotated patent data;

[0126] The test set is specifically an unlabeled patent dataset after index completion, which is expressed as ;

[0127] In this embodiment, the preset confidence threshold Set it in [0.9, 0.95], increasing the step size by 0.01;

[0128] S5. Construct an evaluation model based on deep learning, iteratively train the evaluation model using the training set, and evaluate it based on the test set and preset performance evaluation indicators to obtain a trained evaluation model;

[0129] In this embodiment, the evaluation model based on deep learning is an expert mixture model; the expert mixture model includes An expert network and a gating network, where each expert network consists of two fully connected layers;

[0130] S51. Iteratively train the expert mixture model using the training set; input the training set into the expert mixture model to obtain a predicted score of the patent value, which is expressed as:

[0131] , ;

[0132] ;

[0133] ;

[0134] ;

[0135] Where, A predicted score representing the patent value; Indicates the The weight of the expert network; Indicates the a network of experts; represents the number of expert networks; Indicates the The index value of the expert network; represents the activation function; Represents the data of the input training set; Indicates the The weights of the first fully connected layer in the expert network; Indicates the The weights of the second fully connected layer in the expert network; Indicates the gate function and The transpose of the weights corresponding to the expert networks; represents the exponential function;

[0136] S52, optimizing the expert mixture model using a mean square error loss function, which is expressed as:

[0137] ;

[0138] Where, represents the mean square error loss function; Indicates the Pseudo labels for training set data; Indicates the number of data in the training set; Indicates the The index value of the training set data;

[0139] S6, and performing evaluation based on the test set and preset performance evaluation indicators to obtain a trained expert mixture model;

[0140] The preset performance evaluation indicators are specifically mean absolute error, mean absolute percentage error and root mean square error, which are expressed as follows:

[0141] ;

[0142] ;

[0143] ;

[0144] Where, represents the mean absolute error; represents the mean absolute percentage error; represents the root mean square error; Indicates the Predicted score of patent value based on training set data; Indicates the The true score of the patent value of the training set data;

[0145] Please refer to Table 1. In this embodiment, the multilayer perceptron is used as the baseline method to compare the error comparison results under the preset performance evaluation indicators of the baseline method, the method combining the baseline method with the semi-supervised learning paradigm, the method combining the baseline method with the expert mixture model, and the method combining the baseline method, the semi-supervised learning paradigm, and the expert mixture model.

[0146] Table 1 Error comparison results

[0147]

[0148] Furthermore, in this embodiment, the effects of different indicator weights and confidence thresholds on the performance evaluation indicators of the expert mixture model are also analyzed; see Figure 2 , the indicator weight When <0.4, the performance evaluation index of the evaluation model is stable; see Figure 3 , the confidence threshold =0.93, the performance evaluation index of the evaluation model is stable;

[0149] S7. Obtain the patent indicator data to be evaluated, input the patent indicator data to be evaluated into the trained evaluation model, and obtain a predicted score for the patent indicator data to be evaluated; and obtain an evaluation result of the patent indicator data to be evaluated based on the predicted score.

[0150] Example 2:

[0151] This embodiment provides a patent value assessment system based on deep learning, which includes a data acquisition module, an indicator completion module, a patent assessment module, and a result output module, wherein:

[0152] The data acquisition module is used to acquire a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically includes patents that include manual indicators and machine indicators, and the unannotated patent dataset specifically includes patents that include machine indicators but lack manual indicators; and transmit the patent indicator dataset to the indicator completion module;

[0153] The indicator completion module is used to extract the semantic features of the patent indicator dataset using a pre-trained large language model and generate text vectors; based on the semi-supervised learning paradigm, the indicator completion module performs indicator completion on the unlabeled patent dataset and constructs a training set and a test set;

[0154] The pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as follows:

[0155] ;

[0156] ;

[0157] Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the Index value of unannotated patent data;

[0158] Based on the semi-supervised learning paradigm, the unlabeled patent dataset is supplemented with indicators, and the training and test sets are constructed. Specifically:

[0159] The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows:

[0160] ;

[0161] Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data;

[0162] Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets;

[0163] Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on artificial indicators, which can be expressed as follows:

[0164] ;

[0165] ;

[0166] ;

[0167] Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on artificial indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on manual indicators;

[0168] The machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data is calculated and expressed as follows:

[0169] ;

[0170] Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data;

[0171] Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ;

[0172] Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows:

[0173] ;

[0174] ;

[0175] Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators;

[0176] The pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels are weighted to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, which can be expressed as follows:

[0177] ;

[0178] ;

[0179] Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight;

[0180] Filter high confidence sets based on the preset confidence threshold ,in Indicates the preset confidence threshold;

[0181] The training set is specifically a combination of a high confidence set and an annotated patent data set, which is expressed as ,in represents a collection of annotated patent data;

[0182] The test set is specifically an unlabeled patent dataset after index completion, which is expressed as ;

[0183] The patent evaluation module is used to construct an evaluation model based on deep learning, iteratively train the evaluation model using the training set, and evaluate it based on the test set and preset performance evaluation indicators to obtain a trained evaluation model;

[0184] Obtaining patent indicator data to be evaluated, inputting the patent indicator data to be evaluated into a trained evaluation model to obtain a predicted score for the patent indicator data to be evaluated; and obtaining an evaluation result for the patent indicator data to be evaluated based on the predicted evaluation;

[0185] The result output module is used to display the evaluation results of the patent indicator data to be evaluated.

[0186] Example 3:

[0187] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a patent value assessment method based on deep learning as described in any embodiment of the present invention is implemented.

[0188] Example 4:

[0189] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, a patent value assessment method based on deep learning as described in any embodiment of the present invention is implemented.

[0190] It is worth noting that the system, electronic device and computer-readable storage medium described in the present invention are all based on the same principles as the method described in Example 1, and will not be repeated here.

[0191] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A patent value assessment method based on deep learning, characterized in that: The method comprises: Obtaining a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically refers to patents that include both manual indicators and machine indicators, and the unannotated patent dataset specifically refers to patents that include machine indicators but lack manual indicators; The pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate text vectors. The unlabeled patent dataset is supplemented with indicators based on the semi-supervised learning paradigm, specifically: The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows: ； Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data; Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets; Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on the supplemented artificial indicators, which can be expressed as follows: ； ； ； Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on completed manual indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on the completed manual indicators; The machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data is calculated and expressed as follows: ； Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data; Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ; Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows: ； ； Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators; Perform weighted calculations on the pseudo labels based on manual and machine indicators and their corresponding confidence levels to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, and construct training and test sets; Constructing an evaluation model based on deep learning, iteratively training the evaluation model using the training set, and evaluating the evaluation model based on the test set and preset performance evaluation indicators to obtain a trained evaluation model; Obtain patent indicator data to be evaluated, input the patent indicator data to be evaluated into a trained evaluation model to obtain a predicted score for the patent indicator data to be evaluated; and obtain an evaluation result of the patent indicator data to be evaluated based on the predicted score.

2. The patent value assessment method based on deep learning according to claim 1, characterized in that: The method further includes performing data cleaning on the patent indicator dataset, wherein the data cleaning includes processing missing values, outliers and standardizing the data format.

3. The patent value assessment method based on deep learning according to claim 1, characterized in that: The pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as follows: ； ； Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the The index value of unlabeled patent data.

4. A patent value assessment method based on deep learning according to claim 3, characterized in that: The weighted calculation of the pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels is expressed as follows: ； ； Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight.

5. The patent value assessment method based on deep learning according to claim 4 is characterized in that: The specific steps to construct the training set and test set are: Filter high confidence sets based on the preset confidence threshold ,in Indicates the preset confidence threshold; The training set is specifically a combination of a high confidence set and an annotated patent data set, which is expressed as ,in represents a collection of annotated patent data; The test set is specifically an unlabeled patent dataset after index completion, which is expressed as .

6. A patent value assessment system based on deep learning, characterized in that: The system includes a data acquisition module, an indicator completion module, a patent evaluation module and a result output module, wherein: The data acquisition module is used to acquire a patent indicator dataset, including an annotated patent dataset and an unannotated patent dataset, wherein patent indicators include manual indicators and machine indicators; the annotated patent dataset specifically includes patents that include manual indicators and machine indicators, and the unannotated patent dataset specifically includes patents that include machine indicators but lack manual indicators; and transmit the patent indicator dataset to the indicator completion module; The indicator completion module is used to extract the semantic features of the patent indicator dataset using a pre-trained large language model and generate text vectors; it completes the indicators of the unlabeled patent dataset based on a semi-supervised learning paradigm, specifically: The text similarity between the text vectors of the annotated patent data and the text vectors of the unannotated patent data is calculated and expressed as follows: ； Where, Indicates the The text vector of the annotated patent data is The text similarity between the text vectors of unlabeled patent data; Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Arrange the generated text similarities in descending order and select the first The artificial indicators of the annotated patent data constitute the artificial indicator set represented as ,in Indicates the number of selected presets; Based on the artificial indicator set, the artificial indicators of the unlabeled patent data are supplemented to generate pseudo labels of the unlabeled patent data based on the supplemented artificial indicators, which can be expressed as follows: ； ； ； Where, Indicates the Manual indicators to complete unlabeled patent data; Indicates the Manual indicators of annotated patent data; Indicates the Pseudo-labeling of unlabeled patent data based on completed manual indicators; Indicates the Value score of annotated patent data; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on the completed manual indicators; The machine indicator similarity between the machine indicators of the annotated patent data and the machine indicators of the unannotated patent data is calculated and expressed as follows: ； Where, Indicates the The machine indicators of the annotated patent data are compared with the The machine indicator similarity between the machine indicators of unlabeled patent data; Arrange the generated machine index similarities in descending order and select the top The machine indicators of the annotated patent data constitute the machine indicator set expressed as ; Generate pseudo labels based on machine indicators for unlabeled patent data, expressed as follows: ； ； Where, Indicates the Pseudo-labeling of unlabeled patent data based on machine indicators; Indicates the The confidence of the pseudo-labels of unlabeled patent data based on machine indicators; Perform weighted calculations on the pseudo labels based on manual and machine indicators and their corresponding confidence levels to obtain the final pseudo labels and final confidence levels of the unlabeled patent data, and construct training and test sets; The patent evaluation module is used to construct an evaluation model based on deep learning, iteratively train the evaluation model using the training set, and evaluate it based on the test set and preset performance evaluation indicators to obtain a trained evaluation model; Obtaining patent indicator data to be evaluated, inputting the patent indicator data to be evaluated into a trained evaluation model to obtain a predicted score for the patent indicator data to be evaluated; and obtaining an evaluation result for the patent indicator data to be evaluated based on the predicted score; The result output module is used to display the evaluation results of the patent indicator data to be evaluated.

7. A patent value assessment system based on deep learning according to claim 6, characterized in that: The pre-trained large language model is used to extract the semantic features of the patent indicator dataset and generate a text vector, which can be expressed as follows: ； ； Where, Indicates the Text vectors of annotated patent data; Indicates the Text vectors of unlabeled patent data; Indicates the preset text prompt word; Indicates the The text content of the annotated patent data; Indicates the Text content of unannotated patent data; Represents the preset sentence embedding model; Represents a pre-trained large language model; Indicates a connection operation; Indicates the Index value of annotated patent data; Indicates the The index value of unlabeled patent data.

8. The patent value assessment system based on deep learning according to claim 7, characterized in that: The weighted calculation of the pseudo labels based on manual indicators and machine indicators and their corresponding confidence levels is expressed as follows: ； ； Where, Indicates the Final pseudo labels for unlabeled patent data; Indicates the The final confidence level of unlabeled patent data; Indicates the preset indicator weight; Filter high confidence sets based on the preset confidence threshold ,in Indicates the preset confidence threshold; The training set is specifically a combination of a high confidence set and an annotated patent data set, which is expressed as ,in represents a collection of annotated patent data; The test set is specifically an unlabeled patent dataset after index completion, which is expressed as .

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements a patent value assessment method based on deep learning as described in any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements a patent value assessment method based on deep learning as described in any one of claims 1 to 5.

Images