A patent text classification method and device

Through the neural network model combining patent attributes and text word vector characteristics, a patent text classification model was established, which solved the problem of low efficiency in patent text classification and achieved fast and accurate classification results.

CN112015895BActive Publication Date: 2024-11-01GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010870909.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-08-26
Publication Date
2024-11-01
Estimated Expiration
2040-08-26

AI Technical Summary

Technical Problem

The prior art has low efficiency in classification of patent texts, complex and time-consuming, making it difficult to effectively classify.

Method used

The neural network model is used to combine patent attribute features and text word vector features to realize automatic classification of patent text through the patent text classification model, establish technical classification nodes and train classification models to improve classification efficiency.

Benefits of technology

The efficiency of patent text classification has been improved, the problem of low classification work efficiency in the prior art has been solved, and the rapid and accurate patent text classification has been achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112015895B_ABST
    Figure CN112015895B_ABST
Patent Text Reader

Abstract

The present application provides a patent text classification method and device, wherein the method includes: obtaining the patent text to be classified, and extracting the patent attribute features of the patent text to be classified; performing word segmentation processing and word vector construction processing on the patent text to be classified to obtain text word vector features; inputting the patent attribute features and the text word vector features into the patent text classification model to obtain the classification results of the patent text to be classified according to the operation of the patent text classification model; wherein the patent text classification model is a neural network model constructed according to the preset training samples and the classification information set corresponding to the preset technical classification node. The present application realizes the classification of patent texts by using a neural network learning method, establishing technical classification nodes according to the user's technical system, and training a patent text classification model based on the classification information set constructed by the technical classification nodes, thereby solving the technical problem of low efficiency of patent text classification in the prior art.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of text classification, and in particular to a patent text classification method and device. Background Art

[0002] With the current pace of vigorously developing intellectual property rights in my country, in order to better protect their own interests, many companies have begun to pay attention to the operation of intellectual property rights. Moreover, for technical companies, patent texts are an effective way to understand and master industry technologies, and help provide direction for technological research and development. For this purpose, many companies will choose to build their own patent database based on their own technical system for use.

[0003] However, unlike ordinary texts, patent texts have very strong expressions of technical and legal characteristics and have many additional attributes, such as classification numbers, applicants, etc. It is very complicated and time-consuming to classify patents according to one's own system, which leads to the technical problem of low efficiency in the classification of patent texts in the existing technology. Summary of the invention

[0004] The present application provides a patent text classification method and device, which are used to solve the technical problem of low efficiency in the classification of patent texts in the prior art.

[0005] First, the first aspect of the present application provides a patent text classification method, comprising:

[0006] Obtaining a patent text to be classified, and extracting patent attribute features of the patent text to be classified;

[0007] Performing word segmentation and word vector construction on the patent text to be classified to obtain text word vector features;

[0008] Inputting the patent attribute features and the text word vector features into a patent text classification model to obtain a classification result of the patent text to be classified according to the operation of the patent text classification model;

[0009] Among them, the patent text classification model is a neural network model constructed based on preset training samples and a classification information set corresponding to a preset technical classification node. The training samples are specifically patent attribute features and text word vector features obtained based on preset patent text samples.

[0010] Optionally, the step of inputting the patent attribute features and the text word vector features into a patent text classification model to obtain a classification result of the patent text to be classified according to the operation of the patent text classification model specifically includes:

[0011] Inputting the patent attribute features and the text word vector features into a patent text classification model, and obtaining the matching degree between the patent text to be classified and each of the technical classification nodes through the operation of the patent text classification model;

[0012] According to the sizes of the respective matching degrees, the patent text to be classified is classified with the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

[0013] Optionally, before classifying the patent text to be classified according to the sizes of the respective matching degrees and using the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified, the method further includes:

[0014] If each of the matching degrees is less than a preset matching degree threshold, the classification result of the patent text to be classified is output as classification failure.

[0015] Optionally, the configuration process of the patent text classification model specifically includes:

[0016] According to the preset training samples, patent attribute features and text word vector features of the training samples are obtained respectively;

[0017] The patent attribute features of the training samples, the text word vector features of the training samples and the classification information set corresponding to the preset technical classification nodes are input into the initial neural network model for model training, and the patent text classification model is obtained after the training is completed.

[0018] Optionally, the training sample configuration process specifically includes:

[0019] The corresponding patent text is extracted from the patent database using a patent search formula, and information is annotated on the patent text based on the patent search formula to obtain a training sample.

[0020] Secondly, the second aspect of the present application provides a patent text classification device, comprising:

[0021] A patent attribute feature extraction unit, used to obtain a patent text to be classified and extract patent attribute features of the patent text to be classified;

[0022] A word vector feature acquisition unit is used to perform word segmentation processing and word vector construction processing on the patent text to be classified to obtain text word vector features;

[0023] A patent text classification unit, used for inputting the patent attribute features and the text word vector features into a patent text classification model, so as to obtain a classification result of the patent text to be classified according to the operation of the patent text classification model;

[0024] Among them, the patent text classification model is a neural network model constructed based on preset training samples and a classification information set corresponding to a preset technical classification node. The training samples are specifically patent attribute features and text word vector features obtained based on preset patent text samples.

[0025] Optionally, the patent text classification unit specifically includes:

[0026] A technical matching degree calculation subunit is used to input the patent attribute features and the text word vector features into a patent text classification model, and obtain the matching degree between the patent text to be classified and each of the technical classification nodes through the operation of the patent text classification model;

[0027] The classification subunit is used to classify the patent text to be classified according to the size of each matching degree and the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

[0028] Optionally, the patent text classification unit further includes:

[0029] The matching degree determination subunit is used to output the classification result of the patent text to be classified as classification failure if each of the matching degrees is less than a preset matching degree threshold.

[0030] Optionally, it also includes:

[0031] A sample feature acquisition unit, used to obtain patent attribute features and text word vector features of the training samples according to preset training samples;

[0032] The model training unit is used to input the patent attribute features of the training samples, the text word vector features of the training samples and the classification information set corresponding to the preset technical classification nodes into the initial neural network model for model training, and obtain the patent text classification model after the training is completed.

[0033] Optionally, it also includes:

[0034] The sample annotation unit is used to extract corresponding patent texts from the patent database using a patent search formula, and to annotate the patent texts based on the patent search formula to obtain training samples.

[0035] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0036] The present application provides a patent text classification method, comprising: obtaining a patent text to be classified, and extracting patent attribute features of the patent text to be classified; performing word segmentation processing and word vector construction processing on the patent text to be classified to obtain text word vector features; inputting the patent attribute features and the text word vector features into a patent text classification model to obtain a classification result of the patent text to be classified according to the operation of the patent text classification model; wherein the patent text classification model is a neural network model constructed according to a preset training sample and a classification information set corresponding to a preset technical classification node, and the training sample is specifically a patent attribute feature and a text word vector feature obtained according to a preset patent text sample.

[0037] This application establishes several technical classification nodes according to the user's technical system through neural network learning, and implements the classification of patent texts based on the classification information set constructed by the technical classification nodes and the patent text classification model trained with the patent text as training samples, thereby solving the technical problem of low efficiency in the classification of patent texts in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 A flowchart of a first embodiment of a patent text classification method provided in this application;

[0040] Figure 2 A flowchart of a second embodiment of a patent text classification method provided in this application;

[0041] Figure 3 A schematic structural diagram of the first embodiment of a patent text classification device provided in this application. DETAILED DESCRIPTION

[0042] The embodiments of the present application provide a patent text classification method and device, which are used to solve the technical problem of low efficiency in the classification of patent texts in the prior art.

[0043] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] See also Figure 1 The first embodiment of the present application provides a patent text classification method, including:

[0045] Step 101: Obtain patent text to be classified, and extract patent attribute features of the patent text to be classified;

[0046] First, after obtaining the patent text to be classified, special fields in the patent text can be identified and corresponding patent attribute features can be extracted based on these special fields.

[0047] Step 102: perform word segmentation and word vector construction on the patent text to be classified to obtain text word vector features;

[0048] Then, the patent text to be classified is segmented into words, and the segmented words are constructed into corresponding text word vector features through word vector construction.

[0049] Step 103: input the patent attribute features and the text word vector features into the patent text classification model to obtain the classification result of the patent text to be classified according to the operation of the patent text classification model;

[0050] Then, the patent attribute features obtained in step 101 and the text word vector features obtained in step 102 are input into a pre-trained patent text classification model to obtain the classification result of the patent text to be classified according to the operation of the patent text classification model.

[0051] Among them, the patent text classification model mentioned in this embodiment is a neural network model constructed based on preset training samples and a classification information set corresponding to a preset technical classification node. The training samples are specifically based on patent attribute features and text word vector features obtained based on preset patent text samples.

[0052] The embodiment of the present application establishes a number of technical classification nodes according to the user's technical system through neural network learning, and implements the classification of patent texts based on the classification information set constructed by the technical classification nodes and the patent text classification model trained with the patent text as a training sample, thereby solving the technical problem of low efficiency in the classification of patent texts in the prior art.

[0053] The above is a detailed description of the first embodiment of a patent text classification method provided by the present application. The following is a detailed description of the second embodiment of a patent text classification method provided by the present application.

[0054] See also Figure 2 , the second embodiment of the present application provides a patent text classification method.

[0055] On the basis of the first embodiment of the present application, further, step 103 specifically includes:

[0056] Step 1031: input the patent attribute features and the text word vector features into the patent text classification model, and obtain the matching degree between the patent text to be classified and each technical classification node through the operation of the patent text classification model;

[0057] Step 1032: According to the sizes of the respective matching degrees, the patent text to be classified is classified with the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

[0058] It should be noted that after the patent attribute features and text word vector features are input into the patent text classification model, the patent text classification model will compare the input patent attribute features, text word vector features and the classification information set corresponding to each technical classification node, and output the matching degree between the features of the patent text to be classified and each technical classification node. Taking the electric power field as an example, the technical classification nodes can include four major classification nodes: power transmission, power transformation, power consumption and power generation. Each technical node can also be subdivided into multiple sub-classification nodes, which can be set by the user and will not be elaborated here.

[0059] According to the output of the patent text classification model, the technical classification node corresponding to the maximum matching degree is determined, and the patent text to be classified is classified according to the technical classification node corresponding to the maximum matching degree, so as to obtain the classification result of the patent text to be classified. For example, the matching degree of a patent belonging to the power generation classification is 0.80, the matching degree of the substation classification is 0.50, the matching degree of the substation classification is 0.40, and the matching degree of the substation classification is 0.30. The system determines that this patent belongs to the power generation classification.

[0060] Furthermore, before step 1032, this embodiment further includes:

[0061] Step 10311, determine whether each matching degree is less than a preset matching degree threshold. If so, output the classification result of the patent text to be classified as classification failure. If not, continue to execute step 1032.

[0062] It should be noted that when the matching degrees corresponding to each technical classification node are less than the preset matching degree threshold, for example, when the matching degrees output by the model are all lower than 0.5, it means that the possibility that the patent text belongs to these technical classification nodes is very low. At this time, the classification result of classification failure is output.

[0063] Furthermore, the configuration process of the patent text classification model of this embodiment specifically includes:

[0064] Step 1001: According to the preset training samples, patent attribute features and text word vector features of the training samples are obtained respectively.

[0065] Step 1002: input the patent attribute features of the training samples, the text word vector features of the training samples, and the classification information set corresponding to the preset technical classification nodes into the initial neural network model for model training. After the training is completed, a patent text classification model is obtained, wherein the classification information set mentioned in this embodiment may include: technical keywords commonly used by the technical classification node, patent classification numbers, etc.

[0066] Furthermore, the configuration process of the training samples in this embodiment specifically includes:

[0067] Step 1000: extract corresponding patent texts from a patent database using a patent search formula, and annotate the patent texts based on the patent search formula to obtain training samples.

[0068] It should be noted that this embodiment generates training samples by utilizing patent search formula retrieval, extracts corresponding patent texts from the patent database according to the input patent search formula, and while extracting the patent texts according to the key information of the search formula, annotates the patent texts based on the key information in the patent search formula, thereby obtaining training samples and improving the configuration efficiency of training samples.

[0069] The embodiment of the present application establishes a number of technical classification nodes according to the user's technical system through neural network learning, and implements the classification of patent texts based on the classification information set constructed by the technical classification nodes and the patent text classification model trained with the patent text as training samples, and automatically configures the training samples in combination with the retrieval method, which further improves the efficiency of model training and solves the technical problem of low efficiency in the classification of patent texts in the prior art.

[0070] The above is a detailed description of the second embodiment of a patent text classification method provided in the present application. The following is a detailed description of the first embodiment of a patent text classification device provided in the present application.

[0071] See also Figure 3 The second aspect of the present application provides a patent text classification device, comprising:

[0072] The patent attribute feature extraction unit 301 is used to obtain the patent text to be classified and extract the patent attribute features of the patent text to be classified;

[0073] A word vector feature acquisition unit 302 is used to perform word segmentation processing and word vector construction processing on the patent text to be classified to obtain the word vector feature of the text;

[0074] The patent text classification unit 303 is used to input the patent attribute features and the text word vector features into the patent text classification model to obtain the classification result of the patent text to be classified according to the operation of the patent text classification model;

[0075] Among them, the patent text classification model is a neural network model constructed based on preset training samples and the classification information set corresponding to the preset technical classification nodes. The training samples are specifically based on the patent attribute features and text word vector features obtained from the preset patent text samples.

[0076] Furthermore, the patent text classification unit 303 specifically includes:

[0077] The technical matching degree calculation subunit 3031 is used to input the patent attribute features and the text word vector features into the patent text classification model, and obtain the matching degree between the patent text to be classified and each technical classification node through the operation of the patent text classification model;

[0078] The classification subunit 3032 is used to classify the patent text to be classified according to the size of each matching degree and the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

[0079] Furthermore, the patent text classification unit also includes:

[0080] The matching degree determination subunit 30311 is used to output the classification result of the patent text to be classified as classification failure if all matching degrees are less than a preset matching degree threshold.

[0081] Furthermore, it also includes:

[0082] The sample feature acquisition unit 3001 is used to obtain patent attribute features and text word vector features of the training samples according to the preset training samples;

[0083] The model training unit 3002 is used to input the patent attribute features of the training samples, the text word vector features of the training samples and the classification information set corresponding to the preset technical classification nodes into the initial neural network model for model training, and obtain the patent text classification model after the training is completed.

[0084] Furthermore, it also includes:

[0085] The sample annotation unit 3000 is used to extract corresponding patent texts from the patent database using the patent search formula, and annotate the patent texts based on the patent search formula to obtain training samples.

[0086] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0087] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0088] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0089] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0090] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0091] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0092] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A patent text classification method, characterized in that: include: Obtaining a patent text to be classified, and extracting patent attribute features of the patent text to be classified; Performing word segmentation and word vector construction on the patent text to be classified to obtain text word vector features; Inputting the patent attribute features and the text word vector features into a patent text classification model to obtain a classification result of the patent text to be classified according to the operation of the patent text classification model; The configuration process of the patent text classification model specifically includes: obtaining the patent attribute features and text word vector features of the training samples according to the preset training samples; inputting the patent attribute features of the training samples, the text word vector features of the training samples and the classification information set corresponding to the preset technical classification nodes into the initial neural network model for model training, and obtaining the patent text classification model after the training is completed; The step of inputting the patent attribute features and the text word vector features into the patent text classification model to obtain the classification result of the patent text to be classified according to the operation of the patent text classification model specifically includes: Inputting the patent attribute features and the text word vector features into a patent text classification model, and obtaining the matching degree between the patent text to be classified and each of the technical classification nodes through the operation of the patent text classification model; According to the sizes of the respective matching degrees, the patent text to be classified is classified with the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

2. A patent text classification method according to claim 1, characterized in that: According to the sizes of the respective matching degrees, the patent text to be classified is classified with the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified, and the method further includes: If each of the matching degrees is less than a preset matching degree threshold, the classification result of the patent text to be classified is output as classification failure.

3. A patent text classification method according to claim 1, characterized in that: The configuration process of the training samples specifically includes: The corresponding patent text is extracted from the patent database using a patent search formula, and information is annotated on the patent text based on the patent search formula to obtain a training sample.

4. A patent text classification device, characterized in that: include: A patent attribute feature extraction unit, used to obtain a patent text to be classified and extract patent attribute features of the patent text to be classified; A word vector feature acquisition unit is used to perform word segmentation processing and word vector construction processing on the patent text to be classified to obtain text word vector features; A sample feature acquisition unit, used to obtain patent attribute features and text word vector features of the training samples according to preset training samples; A model training unit, used to input the patent attribute features of the training samples, the text word vector features of the training samples and the classification information set corresponding to the preset technical classification nodes into the initial neural network model for model training, and obtain a patent text classification model after the training is completed; A patent text classification unit, used for inputting the patent attribute features and the text word vector features into the patent text classification model, so as to obtain the classification result of the patent text to be classified according to the operation of the patent text classification model; The patent text classification unit specifically includes: A technical matching degree calculation subunit is used to input the patent attribute features and the text word vector features into a patent text classification model, and obtain the matching degree between the patent text to be classified and each of the technical classification nodes through the operation of the patent text classification model; The classification subunit is used to classify the patent text to be classified according to the size of each matching degree and the technical classification node corresponding to the maximum matching degree to obtain the classification result of the patent text to be classified.

5. A patent text classification device according to claim 4, characterized in that: The patent text classification unit also includes: The matching degree determination subunit is used to output the classification result of the patent text to be classified as classification failure if each of the matching degrees is less than a preset matching degree threshold.

6. A patent text classification device according to claim 4, characterized in that: Also includes: The sample annotation unit is used to extract corresponding patent texts from the patent database using a patent search formula, and to annotate the patent texts based on the patent search formula to obtain training samples.