Contract classification method, equipment and medium

Abstract

The invention discloses a contract classification method and device and a medium, and the method comprises the steps: carrying out the feature extraction of a contract text, and obtaining a feature keyword of the contract text; converting the feature keywords to obtain semantic representation of the contract text; when a plurality of contract texts exist, calculating the semantic similarity between the contract texts through a twin neural network technology according to the feature keyword and semantic representation of each contract text; and classifying the plurality of contract texts according to the semantic similarity between the contract texts. And the contract classification accuracy and efficiency are improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a contract classification method, device and medium. Background Art

[0002] In the current digital age, the contract management system has become an indispensable part of corporate office work. In contract management, the classification of contract texts is a very important task.

[0003] At present, contract classification methods usually rely on manual reading and understanding of contract texts, or simple keyword classification, which is easily affected by subjective factors and cannot deeply analyze the internal semantics of the contract text, resulting in low accuracy and efficiency of contract classification results. Summary of the invention

[0004] The embodiments of the present application provide a contract classification method, device and medium for solving the problems of low accuracy and efficiency of contract classification results.

[0005] The present application embodiment adopts the following technical solutions:

[0006] On the one hand, an embodiment of the present application provides a contract classification method, which includes: extracting features from a contract text to obtain feature keywords of the contract text; converting the feature keywords to obtain a semantic representation of the contract text; when there are multiple contract texts, calculating the semantic similarity between each contract text based on the feature keywords and semantic representation of each contract text through a twin neural network technology; and classifying multiple contract texts based on the semantic similarity between each contract text.

[0007] In one example, the feature extraction of the contract text to obtain the feature keywords of the contract text specifically includes: obtaining a sample contract text and sample feature keywords of the sample contract text; using the sample contract text as a sample input and the sample feature keywords as sample labels, training a text extraction recurrent neural network architecture to obtain a text extraction recurrent neural network model; and inputting the contract text into the text extraction recurrent neural network model to obtain the feature keywords of the contract text.

[0008] In one example, the feature keywords are converted to obtain the semantic representation of the contract text, specifically including: based on an unlabeled sample text data set, self-supervising training is performed on an initial bidirectional encoder representation model to obtain a pre-trained bidirectional encoder representation model; based on sample feature keywords of a sample contract text and sample semantic representations of the sample feature keywords, a labeled sample data set is generated; based on the labeled sample data set, fine-tuning training is performed on the pre-trained bidirectional encoder representation model to obtain a bidirectional encoder representation model; and the feature keywords are input into the bidirectional encoder representation model to obtain the semantic representation of the contract text.

[0009] In one example, the semantic similarity between each contract text is calculated through the twin neural network technology based on the characteristic keywords and semantic representation of each contract text, specifically including: obtaining sample characteristic keywords and sample semantic representation of the sample contract text, and the sample semantic similarity between the sample contract texts; using the sample contract text, the sample characteristic keywords and sample semantic representation of the sample contract text as sample input, and the sample semantic similarity between the sample contract texts as sample labels, training the similarity twin neural network model architecture to obtain a similarity twin neural network model; inputting each contract text, the characteristic keywords and semantic representation of each contract text into the similarity twin neural network model to obtain the semantic similarity between each contract text.

[0010] In one example, the classification of multiple contract texts according to the semantic similarity between each contract text specifically includes: obtaining the sample semantic similarity between a sample contract text and a sample contract text, and the sample category of the sample contract text; using the sample semantic similarity between the sample contract text and the sample contract text as a sample input, using the sample category of the sample contract text as a sample label, and training a classification convolutional neural network model architecture to obtain a classification convolutional neural network model; inputting each contract text and the semantic similarity between each contract text into the classification convolutional neural network model to obtain the category of the contract text.

[0011] In one example, after classifying multiple contract texts according to the semantic similarity between each contract text, the method also includes: parsing the user's query request to obtain query keywords; calculating the matching degree between the query keywords and the characteristic keywords of each contract text in the contract database; sorting the contracts in order of matching degree from high to low; selecting the contract to be displayed in the contract sorting in order from high to low, based on a preset number of contracts, and feeding back the contract to be displayed to the user.

[0012] In one example, the calculation of the matching degree between the query keyword and the characteristic keyword of each contract text specifically includes: matching the query keyword in a pre-constructed standard dictionary to obtain the standard keyword of the query keyword; matching the characteristic keyword of each contract text in the standard dictionary to obtain the standard characteristic keyword of each contract text; comparing the standard keyword with the standard characteristic keyword to determine the common keywords; and determining the matching degree of each contract based on the number of common keywords.

[0013] In one example, the method further includes: when there is a contract text, outputting the contract text to a temporary contract text storage area; when other contract texts are received, classifying the contract text with the other contract texts.

[0014] On the other hand, an embodiment of the present application provides a contract classification device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a contract classification method described in any one of the above items.

[0015] On the other hand, an embodiment of the present application provides a contract classification non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions can execute any of the above-described contract classification methods.

[0016] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0017] It can extract the characteristic keywords and semantic representations of the contract text, deeply explore the semantics within the contract text, and further calculate the semantic similarity between each contract text. It can effectively handle the complex semantic relationship between contract texts by utilizing natural language processing and semantic similarity technology, and can more quickly and accurately classify contracts of the same category into corresponding categories, avoiding the problem of strong subjectivity in manual classification and improving the accuracy and efficiency of classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the present application, some embodiments of the present application will be described in detail below in conjunction with the accompanying drawings, in which:

[0019] Figure 1 A flowchart of a contract classification method provided in an embodiment of the present application;

[0020] Figure 2 A schematic diagram of the architecture of a contract classification system provided in an embodiment of the present application;

[0021] Figure 3 A schematic diagram of the structure of a contract classification device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0023] Some embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0024] Figure 1 A flowchart of a contract classification method provided in an embodiment of the present application. The method can be applied to different business fields, such as Internet financial business field, e-commerce business field, instant messaging business field, game business field, official business field, etc. The process can be executed by a computing device in the corresponding field, and some input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.

[0025] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail by taking a server as an example.

[0026] It should be noted that the server may be a single device or a system consisting of multiple devices, that is, a distributed server, and this application does not make any specific limitation on this.

[0027] Figure 1 The process in includes the following steps:

[0028] S101: Extract features from the contract text to obtain feature keywords of the contract text.

[0029] In some embodiments of the present application, the contract text is feature extracted by natural language processing technology. The present application is performed using a recurrent neural network (RNN), which is a type of recurrent neural network that takes sequence data as input, recurses in the direction of sequence evolution, and all nodes (recurrent units) are connected in a chain. The specific manifestation is that the network will memorize the previous information and apply it to the calculation of the current output, that is, the nodes between the hidden layers are no longer disconnected but connected, and the input of the hidden layer includes not only the output of the input layer but also the output of the hidden layer at the previous moment.

[0030] Specifically, it is necessary to pre-build a text extraction recurrent neural network model. The process is as follows: First, obtain the sample contract text and the sample feature keywords of the sample contract text. Then, use the sample contract text as the sample input and the sample feature keywords as the sample label to train the text extraction recurrent neural network architecture and obtain the text extraction recurrent neural network model.

[0031] Based on this, the contract text is input into the text extraction recurrent neural network model to obtain the characteristic keywords of the contract text.

[0032] S102: Convert the characteristic keywords to obtain a semantic representation of the contract text.

[0033] In some embodiments of the present application, using the Bidirectional Encoder Representation (BERT), a deep bidirectional representation of a language can be captured by pre-training on large-scale text data, and then fine-tuned for different natural language processing tasks.

[0034] Specifically, it is necessary to pre-build a bidirectional encoder representation model. The process is as follows: First, the initial bidirectional encoder representation model is self-supervised trained based on the unlabeled sample text dataset to obtain a pre-trained bidirectional encoder representation model. Then, a labeled sample dataset is generated based on the sample feature keywords of the sample contract text and the sample semantic representation of the sample feature keywords. Finally, the pre-trained bidirectional encoder representation model is fine-tuned based on the labeled sample dataset to obtain a bidirectional encoder representation model.

[0035] Based on this, the feature keywords are input into the bidirectional encoder representation model to obtain the semantic representation of the contract text.

[0036] It should be noted that semantic representation refers to converting the characteristic keywords in the contract text into vector representations with semantic information through some conversion methods, such as word vectors, BERT and other models. This representation can better capture the semantic information in the contract text.

[0037] S103: When there are multiple contract texts, the semantic similarity between each contract text is calculated using the twin neural network technology based on the characteristic keywords and semantic representation of each contract text.

[0038] In some embodiments of the present application, the semantic similarity is calculated using the twin neural network technology, which can perform small sample learning and is not easily disturbed by erroneous samples.

[0039] Specifically, it is necessary to pre-build a similarity twin neural network model, and the process is as follows: First, obtain the sample feature keywords and sample semantic representation of the sample contract text, as well as the sample semantic similarity between the sample contract texts. Then, the sample contract text, the sample feature keywords and sample semantic representation of the sample contract text are used as sample inputs, and the sample semantic similarity between the sample contract texts is used as the sample label, and the similarity twin neural network model architecture is trained to obtain the similarity twin neural network model.

[0040] Based on this, each contract text, the characteristic keywords and semantic representation of each contract text are input into the similarity twin neural network model to obtain the semantic similarity between each contract text.

[0041] It can be seen that the semantic similarity calculation method is used to compare the semantic similarities between different contract texts to determine the degree of association between them.

[0042] S104: Classify the multiple contract texts according to the semantic similarity between each contract text.

[0043] In some embodiments of the present application, convolutional neural network technology is used to identify and classify data. Convolutional neural networks can automatically learn and extract data features, and can share parameters, which can reduce model parameters and calculations, and improve model training efficiency.

[0044] Specifically, it is necessary to pre-build a classification convolutional neural network model, and the process is as follows: First, obtain the sample semantic similarity between the sample contract text and the sample contract text, as well as the sample category of the sample contract text. Then, take the sample semantic similarity between the sample contract text and the sample contract text as the sample input, take the sample category of the sample contract text as the sample label, train the classification convolutional neural network model architecture, and obtain the classification convolutional neural network model.

[0045] Based on this, each contract text and the semantic similarity between each contract text are input into the classification convolutional neural network model to obtain the category of the contract text.

[0046] In some embodiments of the present application, graph database technology (Neo4j) can be used to achieve more complex and accurate association retrieval. This technology can improve retrieval efficiency and query flexibility, and has better performance and response speed.

[0047] Specifically, when a user searches, first, the user's query request is parsed to obtain the query keyword. Then, in the contract database, the matching degree between the query keyword and the characteristic keyword of each contract text is calculated. Then, the contracts are sorted in order of matching degree from high to low. Finally, in the order from high to low, with a preset number of contracts as a condition, the contracts to be displayed are selected in the contract sorting, and the contract sorting to be displayed is fed back to the user.

[0048] In the process of calculating the matching degree, since there may be different descriptions for the same word, a standard dictionary is constructed.

[0049] Based on this, the process is as follows: First, match the query keywords in the pre-built standard dictionary to obtain the standard keywords of the query keywords. Then, match the characteristic keywords of each contract text in the standard dictionary to obtain the standard characteristic keywords of each contract text. Then, compare the standard keywords with the standard characteristic keywords to determine the common keywords. Finally, determine the matching degree of each contract based on the number of common keywords. For example, different numbers correspond to different matching degrees.

[0050] It should be noted that when the contract text has one, the contract text is output to the contract text temporary storage area, so that when other contract texts are received, the contract text is classified with other contract texts. That is, the contract classification is performed according to the semantic similarity between the contract text and other contract texts.

[0051] Since the current contract management system cannot achieve in-depth understanding and analysis of contract texts, cannot achieve accurate semantic similarity comparison, has certain limitations in semantic similarity comparison, and cannot achieve accurate matching of complex semantic relationships between contracts, the classification results are inaccurate, affecting the user's trust in the contract management system and user experience. However, this application can automatically classify and retrieve contracts through natural language processing and semantic similarity technology, that is, it can use natural language processing and artificial intelligence technology to semantically represent contracts at the initial stage of the contract, calculate and classify the semantic similarity between different contract texts, and when the user searches, it matches the query conditions in the contract database to obtain the contract with the highest matching degree, thereby improving the efficiency of contract classification and retrieval, and can realize an intelligent contract management system, improve the efficiency and accuracy of contract management, and provide enterprises with more convenient and intelligent contract management solutions.

[0052] It should be noted that although the embodiments of the present application are based on Figure 1Steps S101 to S104 are described in sequence, but this does not mean that steps S101 to S104 must be performed in a strict order. Figure 1 The order shown in the figure is to introduce and explain step S101 to step S104 in sequence, in order to facilitate those skilled in the art to understand the technical solution of the embodiment of the present application. In other words, in the embodiment of the present application, the order between step S101 to step S104 can be appropriately adjusted according to actual needs.

[0053] pass Figure 1 The method can extract the characteristic keywords and semantic representation of the contract text, deeply mine the semantics within the contract text, and further calculate the semantic similarity between each contract text. It can effectively handle the complex semantic relationship between contract texts by utilizing natural language processing and semantic similarity technology, and can classify the same type of contracts into corresponding categories more quickly and accurately, avoiding the problem of strong subjectivity in manual classification and improving the accuracy and efficiency of classification.

[0054] More intuitively, Figure 2 A schematic diagram of the architecture of a contract classification system provided in an embodiment of the present application.

[0055] exist Figure 2 In the classification, the user inputs the contract information, the contract management module receives the contract information, the WEB server processes the contract information, the data model performs natural language processing and semantic classification on the processed contract information, and then accesses the database to store the classified contract text in the database. The database returns the data to the WEB server, and the WEB server displays the corresponding information to the user, that is, the result of the contract classification.

[0056] During the search, the contract management module receives the query conditions entered by the user, and then the WEB server processes the query conditions, and the data model performs intelligent search on the query conditions, thereby querying the database according to the query conditions. The database returns the data to the WEB server, and the WEB server feeds back the data to the user, that is, the search results.

[0057] Based on the same idea, some embodiments of the present application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0058] Figure 3 A schematic diagram of the structure of a contract classification device provided in an embodiment of the present application includes:

[0059] at least one processor; and,

[0060] a memory communicatively connected to the at least one processor; wherein,

[0061] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the contract classification methods described above.

[0062] Some embodiments of the present application provide a contract classification non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions can execute any of the above-described contract classification methods.

[0063] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0064] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0065] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0070] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0071] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0072] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0073] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the technical principle of the present application should all fall within the protection scope of the present application.

Claims

1. A contract classification method, characterized in that: The method comprises: Extracting features from the contract text to obtain feature keywords of the contract text; Converting the characteristic keywords to obtain a semantic representation of the contract text; When there are multiple contract texts, the semantic similarity between each contract text is calculated using the twin neural network technology based on the characteristic keywords and semantic representation of each contract text; Multiple contract texts are classified according to the semantic similarity between each contract text.

2. The method according to claim 1, characterized in that The feature extraction of the contract text to obtain the feature keywords of the contract text specifically includes: Obtaining a sample contract text and sample characteristic keywords of the sample contract text; The sample contract text is used as a sample input, the sample feature keywords are used as sample labels, and a text extraction recurrent neural network architecture is trained to obtain a text extraction recurrent neural network model; The contract text is input into the text extraction recurrent neural network model to obtain characteristic keywords of the contract text.

3. The method according to claim 1, characterized in that The converting of the characteristic keywords to obtain the semantic representation of the contract text specifically includes: According to the unlabeled sample text dataset, the initial bidirectional encoder representation model is self-supervised trained to obtain a pre-trained bidirectional encoder representation model; Generate a labeled sample data set according to sample feature keywords of the sample contract text and sample semantic representations of the sample feature keywords; Fine-tune the pre-trained bidirectional encoder representation model according to the labeled sample data set to obtain a bidirectional encoder representation model; The characteristic keywords are input into the bidirectional encoder representation model to obtain the semantic representation of the contract text.

4. The method according to claim 1, characterized in that: The method of calculating the semantic similarity between each contract text by using the twin neural network technology based on the characteristic keywords and semantic representation of each contract text specifically includes: Obtaining sample feature keywords and sample semantic representations of sample contract texts, as well as sample semantic similarities between sample contract texts; The sample contract text, the sample feature keywords and the sample semantic representation of the sample contract text are used as sample inputs, the sample semantic similarity between the sample contract texts is used as sample labels, and the similarity twin neural network model architecture is trained to obtain a similarity twin neural network model; Each contract text, the characteristic keywords and the semantic representation of each contract text are input into the similarity twin neural network model to obtain the semantic similarity between each contract text.

5. The method according to claim 1, characterized in that The classifying of the multiple contract texts according to the semantic similarity between each contract text specifically includes: Obtaining sample semantic similarity between a sample contract text and a sample contract text, and a sample category of the sample contract text; Taking the sample semantic similarity between the sample contract text and the sample contract text as sample input, taking the sample category of the sample contract text as sample label, training the classification convolutional neural network model architecture, and obtaining a classification convolutional neural network model; Each contract text and the semantic similarity between each contract text are input into the classification convolutional neural network model to obtain the category of the contract text.

6. The method according to claim 1, characterized in that After classifying the multiple contract texts according to the semantic similarity between each contract text, the method further includes: Parse the user's query request to obtain the query keywords; In the contract database, calculating the matching degree between the query keyword and the characteristic keyword of each contract text; Sort the contracts in order of highest to lowest matching degree; In order from high to low, with a preset number of contracts as a condition, in the contract sorting, the contract sorting to be displayed is selected, and the contract sorting to be displayed is fed back to the user.

7. The method according to claim 6, characterized in that The calculating the matching degree between the query keyword and the characteristic keyword of each contract text specifically includes: Matching the query keyword in a pre-constructed standard dictionary to obtain a standard keyword for the query keyword; In the standard dictionary, the characteristic keywords of each contract text are matched to obtain the standard characteristic keywords of each contract text; Comparing the standard keywords with the standard feature keywords to determine common keywords; Determine the matching degree of each contract based on the number of common keywords.

8. The method according to claim 1, characterized in that The method further comprises: When there is a contract text, output the contract text to a temporary contract text storage area; When other contract texts are received, the contract text is classified together with the other contract texts.

9. A contract classification device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a contract classification method as described in any one of claims 1 to 8.

10. A contract classification non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are capable of executing a contract classification method as described in any one of claims 1-8 above.

Images