System architecture diagram processing method and device, computer equipment and storage medium

Abstract

The invention relates to a system architecture diagram processing method and device, computer equipment and a storage medium. The method relates to the field of artificial intelligence, and comprises the following steps: obtaining a target description text for describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in the target domain; identifying a dependency relationship between target domain concepts in the target description text; generating a domain concept relationship graph based on the dependency relationship between the target domain concepts; performing domain concept redundancy detection on the domain concept relation graph to obtain a redundancy detection result of the domain concept relation graph; the redundancy detection result is used for performing redundancy elimination processing on the target domain concept expressed in the target system architecture diagram. By adopting the method, the rationality of the system architecture diagram can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer device and storage medium for processing a system architecture diagram. Background Art

[0002] With the development of computer technology and artificial intelligence, the demand for building complex systems is increasing. To build a system, you need to design a system architecture diagram first. The system architecture diagram includes the various components of the system. The components in the system architecture diagram can be determined according to the domain to which the system belongs. For example, the components can be represented by domain concepts. Domain concepts refer to concepts with specific meanings in a specific domain. For example, in the field of face recognition, face images, face recognition algorithms, and face features are all domain concepts.

[0003] In traditional technology, a manual method is usually used to design a system architecture diagram, that is, each component in the system architecture diagram is determined by a manual method.

[0004] However, when the system complexity is high, the system architecture diagram obtained by manually designing the system architecture diagram usually contains redundant information. For example, there are redundant domain concepts in the system architecture diagram, which makes the design of the system architecture diagram inaccurate. Summary of the invention

[0005] Based on this, it is necessary to provide a system architecture diagram processing method, apparatus, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of the system architecture diagram in response to the above technical problems.

[0006] On the one hand, the present application provides a method for processing a system architecture diagram. The method includes: obtaining a target description text that describes a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in the target domain; identifying the dependency relationship between target domain concepts in the target description text; generating a domain concept relationship diagram based on the dependency relationship between the target domain concepts; performing domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result of the domain concept relationship diagram; the redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram.

[0007] On the other hand, the present application also provides a system architecture diagram processing device. The device includes: a description text acquisition module, which is used to acquire a target description text that describes a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in the target domain; a dependency identification module, which is used to identify the dependency between target domain concepts in the target description text; a relationship diagram generation module, which is used to generate a domain concept relationship diagram based on the dependency between the target domain concepts; a redundancy detection module, which is used to perform domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result of the domain concept relationship diagram; the redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram.

[0008] In some embodiments, the dependency relationship is identified through a second text generation model; the device also includes a model acquisition module; the model acquisition module is used to obtain a first text generation model with domain concept recognition capability; the training text is input into the first text generation model to obtain the output result of the first text generation model; the training text is a text belonging to the target domain; based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, the first text generation model is adjusted to obtain the second text generation model; the standard domain concept relationship includes the dependency relationship between each training domain concept, and the training domain concept refers to the domain concept included in the training text.

[0009] In some embodiments, the training text is target domain knowledge of the target domain including multiple domain concepts; the model acquisition module is also used to input the target domain knowledge into the first text generation model to obtain the output result of the first text generation model.

[0010] In some embodiments, the model acquisition module is further used to construct a training text based on multiple domain concepts of the target domain; input the training text into the first text generation model to obtain the output result of the first text generation model.

[0011] In some embodiments, the model acquisition module is also used to obtain a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; obtain sample domain knowledge; the sample domain knowledge belongs to the target domain; input the sample domain knowledge into the pre-trained text generation model to obtain the output result of the pre-trained text generation model; based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge, adjust the model parameters of the pre-trained text generation model to obtain a first text generation model.

[0012] In some embodiments, the model acquisition module is also used to obtain multiple sample texts; the sample texts among the multiple sample texts belong to the target domain, at least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is domain knowledge of the target domain; for each of the sample texts, the sample text is input into the text generation model to be trained to obtain the output result of the text generation model to be trained; based on the difference between the sample text and the output result of the text generation model to be trained, the model parameters of the text generation model to be trained are adjusted to obtain the pre-trained text generation model.

[0013] In some embodiments, the relationship graph generation module is further used to use each of the target domain concepts as a node and the dependency relationships between the target domain concepts as an edge; and generate a domain concept relationship graph based on each of the nodes and the edges.

[0014] In some embodiments, the domain concept relationship graph includes a target loop, and each of the target domain concepts constituting the target loop includes redundant target domain concepts, and the redundancy detection result includes a loop detection result: the redundancy detection module is also used to perform loop detection on the domain concept relationship graph to obtain a loop detection result; the loop detection result includes the dependency relationship between each of the target domain concepts constituting the target loop.

[0015] In some embodiments, the redundancy detection result also includes a repeated concept detection result: the redundancy detection module is further used to detect target domain concepts with repeated dependencies in the domain concept relationship graph to obtain a repeated concept detection result.

[0016] In some embodiments, the target field is the palmprint recognition field, and the target system is a palmprint recognition system; the obtaining of the target description text describing the target system architecture diagram includes: receiving the target description text sent by a terminal corresponding to the palmprint recognition system; the method also includes: sending the redundant detection result to the terminal corresponding to the palmprint recognition system.

[0017] In some embodiments, the device is also used to: obtain a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; sample domain concepts and domain knowledge related to the sample domain concepts are input into the pre-trained text generation model to obtain output text of the pre-trained text generation model; based on the difference between the output text and the description text of the sample domain concept, the parameters of the pre-trained text generation model are adjusted to obtain a trained text generation model; the trained text generation model is used to generate description text of domain concepts.

[0018] On the other hand, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in the above-mentioned system architecture diagram processing method are implemented.

[0019] On the other hand, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the above-mentioned system architecture diagram processing method are implemented.

[0020] On the other hand, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned system architecture diagram processing method are implemented.

[0021] The above-mentioned system architecture diagram processing method, device, computer equipment, storage medium and computer program product obtain a target description text describing a target system architecture diagram, the target system architecture diagram is an architecture diagram of a target system in a target domain, identifies the dependency relationship between target domain concepts in the target description text, generates a domain concept relationship diagram based on the dependency relationship between target domain concepts, performs domain concept redundancy detection on the domain concept relationship diagram, obtains a redundancy detection result of the domain concept relationship diagram, and the redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram. Thus, by generating a domain concept relationship diagram and performing domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result, it is convenient to use the redundancy detection result to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram, so as to improve the rationality of the target system architecture diagram. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is an application environment diagram of the system architecture diagram processing method in some embodiments;

[0023] Figure 2 A schematic diagram of a process flow of a method for processing a system architecture diagram in some embodiments;

[0024] Figure 3 This is an architectural diagram of a palm-swiping system in some embodiments;

[0025] Figure 4 A schematic diagram of a domain concept relationship diagram in some embodiments;

[0026] Figure 5 This is an architecture diagram of a palm-swiping system after redundancy removal in some embodiments;

[0027] Figure 6 A schematic diagram of a pre-trained text generation model for training in some embodiments;

[0028] Figure 7 A schematic diagram of a domain concept relationship diagram in some embodiments;

[0029] Figure 8 A schematic diagram of a process flow of a method for processing a system architecture diagram in some embodiments;

[0030] Fig. 9 It is a structural block diagram of a system architecture diagram processing device in some embodiments;

[0031] Fig.10 is an internal structure diagram of a computer device in some embodiments;

[0032] Fig.11 1 is a diagram of the internal structure of a computer device in some embodiments. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0034] The system architecture diagram processing method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other servers.

[0035] Specifically, the terminal 102 sends a target description text to the server 104. The target description text is a text describing the target system architecture diagram. The server 104 identifies the dependency relationship between the target domain concepts in the target description text. The server 104 generates a domain concept relationship diagram based on the dependency relationship between the target domain concepts. The server 104 performs domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result of the domain concept relationship diagram. The redundancy detection result is used to perform redundancy removal processing on the domain concepts expressed in the target system architecture diagram. The server 104 can return the redundancy detection result to the terminal 102. The terminal 102 can be a terminal of a developer of the target system or a designer of the target system architecture diagram.

[0036] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, and this application is not limited here.

[0037] The system architecture diagram processing method provided in the embodiment of the present application may involve the field of artificial intelligence. For example, the server may input the target description text into a neural network model for identifying the dependency relationship between domain concepts, and identify the dependency relationship. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive discipline that involves a wide range of fields, including both hardware-level technology and software-level technology. Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technology mainly includes several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0038] The solution provided in the embodiments of the present application involves artificial intelligence, machine learning and other technologies, which are specifically described by the following embodiments:

[0039] In some embodiments, Figure 2 As shown, a system architecture diagram processing method is provided, which can be executed by a terminal or a server, or by a terminal and a server together. Figure 1 The server 104 in the example is used as an example to illustrate, and the following steps are included:

[0040] Step 202, obtaining a target description text describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in a target domain.

[0041] Among them, the target field can be any field, for example, the target field can be face recognition, fingerprint recognition, palmprint recognition, etc. The target system belongs to the target field. For example, when the target field is palmprint recognition, the target system can be a palm swiping system or a palmprint recognition system. The target system architecture diagram refers to the architecture diagram of the target system. The target system architecture diagram can be manually designed or automatically generated by computer equipment, and there is no restriction here. The target system architecture diagram includes multiple components, and there can be subordinate relationships between the components. In the target system architecture diagram, each component can be represented by a domain concept in the target field. A domain concept refers to a concept with a specific meaning in a specific field. For example, in the field of palmprint recognition, palm features, palm images, and left and right hands are all domain concepts. For example Figure 3 The figure shows the architecture of the palm-swiping system. Figure 3 It can be seen that the components of the palm swiping system include palmprint recognition, equipment management, card swiping, system management, data backup, report statistics and data recovery. Among them, the components of palmprint recognition include palmprint comparison, palmprint feature extraction, palmprint recognition algorithm, palmprint library and palmprint classification. The components of equipment management include equipment maintenance, equipment configuration, equipment status detection, equipment monitoring and equipment diagnosis. The components of card swiping include card reader, card swiping operation, ID card recognition and card processor. The components of system management include system configuration, log management, performance monitoring, user authority management and exception handling.

[0042] In order to detect whether the design of the target system architecture diagram is reasonable, the target system architecture diagram can be described to obtain the target description text. The target description text is a text that describes the target system architecture diagram, including the domain concepts representing each component in the target system architecture diagram. For example, the target description text is "The palm swiping system includes palmprint recognition, equipment management, card swiping, system management, data backup, report statistics and data recovery. Among them, palmprint recognition includes palmprint comparison, palmprint feature extraction, palmprint recognition algorithm, palmprint library and palmprint classification. Equipment management includes equipment maintenance, equipment configuration, equipment status detection, equipment monitoring and equipment diagnosis. Card swiping includes card reader, card swiping operation, ID card recognition and card processor. System management includes system configuration, log management, performance monitoring, user authority management, and exception handling. "The domain concepts representing the components in the target system architecture diagram can be called target domain concepts.

[0043] Specifically, the terminal sends a redundant identification request carrying the target description text to the server. The server responds to the redundant identification request and obtains the target description text from the redundant identification request.

[0044] Step 204 , identifying dependencies between target domain concepts in the target description text.

[0045] The target domain concept is a domain concept extracted from the target description text, that is, the domain concept representing the component in the target system architecture diagram. The dependencies between domain concepts include, but are not limited to, circular dependencies, subordinate relationships, and repeated relationships. A cycle refers to multiple concepts that overlap and cover each other. A circular dependency refers to a situation where two components are mutually dependent, causing the two components to wait for each other and cannot continue to execute, such as a deadlock.

[0046] Specifically, a second text generation model may be deployed in the server. The second text generation model may be generated based on the first text generation model, and the first text generation model has the ability to identify domain concepts from domain knowledge. The second text generation model is used to identify domain concepts in the target domain and the dependencies between domain concepts. That is, the second text generation model has the function of identifying domain concepts in the target domain and the dependencies between domain concepts. The server inputs the target description text into the second text generation model, and the target domain concepts in the target description text and the dependencies between the target domain concepts are obtained through the second text generation model.

[0047] In some embodiments, the second text generation model is obtained based on a pre-trained text generation model. The pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain. That is, the pre-trained text generation model has the ability to generate domain knowledge and domain concepts of the target domain.

[0048] In some embodiments, the pre-trained text generation model can be a neural network model, for example, it can be a GPT (Generative Pre-Trained Transformer). The GPT model is an Internet-based, data-trained, text-generating deep learning model. The GPT model is a neural network model based on the Transformer architecture, which can be applied to natural language processing tasks, such as text generation, text classification, or language modeling.

[0049] Step 206: Generate a domain concept relationship graph based on the dependency relationships between the target domain concepts.

[0050] The domain concept relationship graph is composed of nodes and edges. The nodes represent the target domain concepts, and the edges represent the dependency relationships between the target domain concepts linked by the edges. The domain concept relationship graph can also be called a dependency graph.

[0051] Specifically, the server may use each target domain concept as a node, use the dependency relationship between the target domain concepts as an edge, and generate a domain concept relationship graph based on each node and each edge.

[0052] In some embodiments, the server can determine the dependencies not described in the target description text, that is, the dependencies not presented in the target system architecture diagram, from the identified dependencies between the target domain concepts, and generate a domain concept relationship graph. For example, it is identified that there is a dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction", and the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction" is not presented in the target system architecture diagram. Then, based on the target system architecture diagram, the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction" is used as an edge connecting "palmprint recognition algorithm" and "palmprint feature extraction". It should be noted that the domain concept relationship graph generated by the second text generation model is in text form, not image form. Figure 4 As shown in Figure 1, a domain concept relationship diagram generated by the second text generation model is displayed using a visual method. Figure 4 It can be seen that the domain concept relationship diagram not only presents the relationships described in the target description text, but also presents the relationships not described in the target description text. For example, palmprint comparison, palmprint feature extraction and palmprint classification are all related to the palmprint recognition algorithm, the card processor is related to the card reader, the device diagnosis is related to the device dimension, and the device monitoring is related to the device status detection.

[0053] Step 208, performing domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result of the domain concept relationship diagram; the redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram.

[0054] Among them, redundancy detection is used to detect redundant target domain concepts from the domain concept relationship graph. In the case where there are no redundant target domain concepts in the domain concept relationship graph, the redundancy detection result indicates that there are no redundant target domain concepts in the domain concept relationship graph. In the case where there are redundant target domain concepts in the domain concept relationship graph, the redundancy detection result includes the redundant target domain concepts. The target system architecture diagram uses domain concepts to represent the components of the target system. The target domain concepts expressed in the target system architecture diagram are domain concepts that represent the components of the target system. The target description text includes the target domain concepts expressed in the target system architecture diagram.

[0055] Specifically, the target description text may be sent from the terminal of the target system to the server. After the server obtains the redundancy detection result, it may send the redundancy detection result to the terminal corresponding to the target system. The terminal of the target system may be a terminal of a developer of the target system or a designer of the target system architecture diagram.

[0056] In some embodiments, the terminal may also send the target system architecture diagram to the server. After the server obtains the redundancy detection result, it may remove the redundant target domain concepts in the target system architecture diagram according to the redundancy detection result to obtain an updated target system architecture diagram. The server may return the updated target system architecture diagram to the terminal. Figure 5 As shown in the figure, the architecture diagram of the palm-swiping system after redundancy removal is shown. Figure 5 compared to Figure 3 , removing the repeated or redundant domain concepts of palmprint feature extraction, palmprint comparison, palmprint classification, card processor, device diagnosis and device detection. Obviously, Figure 5 The architecture diagram in Figure 3 The architecture diagram in is more concise and reasonable.

[0057] In the above-mentioned system architecture diagram processing method, a target description text describing the target system architecture diagram is obtained, the target system architecture diagram is an architecture diagram of the target system in the target domain, the dependency relationship between the target domain concepts in the target description text is identified, a domain concept relationship diagram is generated based on the dependency relationship between the target domain concepts, a domain concept redundancy detection is performed on the domain concept relationship diagram, and a redundancy detection result of the domain concept relationship diagram is obtained. The redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram. Thus, by generating a domain concept relationship diagram and performing domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result, it is convenient to use the redundancy detection result to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram, so as to improve the rationality of the target system architecture diagram. In addition, the system architecture diagram processing method provided by the present application realizes a method for automatically performing domain concept redundancy detection, thereby improving the efficiency of redundancy detection.

[0058] In some embodiments, the dependency relationship is identified through a second text generation model; the step of obtaining the second text generation model includes: obtaining a first text generation model with domain concept recognition capability; inputting training text into the first text generation model to obtain the output result of the first text generation model; the training text is a text belonging to the target domain; based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, adjusting the first text generation model to obtain the second text generation model; the standard domain concept relationship includes the dependency relationship between each training domain concept, and the training domain concept refers to the domain concept existing in the training text.

[0059] The training text may be domain knowledge of the target domain including multiple domain concepts. For example, the training text is “extract palm features from palm images and perform identity recognition based on palm features”, that is, the training text may be a sentence or paragraph in the target domain. Alternatively, the training text may be a text composed of multiple domain concepts in the target domain, for example, the training text is “palm image, palm features, left and right hands”.

[0060] Specifically, the server inputs the training text into the first text generation model to obtain the output result of the first text generation model. The server generates the first model loss based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text. The difference between the output result of the first text generation model and the standard domain concept relationship of the training text is positively correlated with the first model loss. The server can calculate the gradient of the first model loss relative to the model parameters of the first text generation model, and adjust the model parameters along the direction of gradient descent, thereby realizing the training of the first text generation model. There can be multiple training texts, and the server can continuously train the first text generation model according to different training texts until the first text generation model converges to obtain a second text generation model. The second text generation model is the first text generation model at convergence.

[0061] In some embodiments, since the second text generation model has the function of extracting domain concepts. Therefore, the second text generation model can also be used to verify whether the domain concepts extracted manually are accurate or whether there are omissions. Specifically, the server obtains the specified text and the domain concepts extracted manually from the specified text. The specified text belongs to the target domain. The server inputs the specified text into the second text generation model to extract the domain concepts in the specified text. The server can verify the domain concepts extracted manually through the domain concepts extracted by the second text generation model.

[0062] In some embodiments, the dependency relationship between the training domain concepts included in the standard domain concept relationship may be determined by the server using a relationship extraction algorithm. The relationship extraction algorithm may be implemented by a neural network model for entity extraction. For example, the neural network model for entity extraction may be a CNN (Convolutional Neural Networks) or RNN (Recurrent Neural Network) for entity extraction.

[0063] In this embodiment, since the standard domain concept relationship includes the dependency relationship between the training domain concepts, based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, the first text generation model is adjusted to obtain the second text generation model, so that the second text generation model can learn the ability to identify the dependency relationship between domain concepts from domain knowledge.

[0064] In some embodiments, the training text is target domain knowledge including multiple domain concepts in the target domain; inputting the training text into the first text generation model to obtain the output result of the first text generation model includes: inputting the target domain knowledge into the first text generation model to obtain the output result of the first text generation model.

[0065] The target domain knowledge is the domain knowledge of the target domain, and the target domain knowledge includes multiple domain concepts of the target domain, where multiple refers to at least two. The training text is the domain knowledge of the target domain including multiple domain concepts. For example, the training text is "extract palm features from palm images and perform identity recognition based on palm features", that is, the training text can be a sentence or paragraph in the target domain.

[0066] In this embodiment, the target domain knowledge is input into the first text generation model to obtain the output result of the first text generation model, so that the second text generation model can learn the ability to identify the dependencies between domain concepts from the domain knowledge, that is, the second text generation model can learn the ability to identify the dependencies between domain concepts from sentences or paragraphs, so that the second text generation model can not only identify domain concepts from the domain knowledge, but also identify the dependencies between domain concepts.

[0067] In some embodiments, inputting the training text into the first text generation model to obtain the output result of the first text generation model includes: constructing the training text based on multiple domain concepts of the target domain; inputting the training text into the first text generation model to obtain the output result of the first text generation model.

[0068] The training text may be a text composed of multiple domain concepts in the target domain, for example, the training text may be “palm image, palm features, left and right hands”.

[0069] In this embodiment, a training text is constructed based on multiple domain concepts of the target domain, and the training text is input into the first text generation model to obtain the output result of the first text generation model, so that the second text generation model can learn the ability to identify the dependency relationship between domain concepts based on multiple domain concepts.

[0070] In some embodiments, obtaining a first text generation model with domain concept recognition capability includes: obtaining a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of a target domain; obtaining sample domain knowledge; the sample domain knowledge belongs to the target domain; inputting the sample domain knowledge into the pre-trained text generation model to obtain an output result of the pre-trained text generation model; based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge, adjusting the model parameters of the pre-trained text generation model to obtain the first text generation model.

[0071] The pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain. The pre-trained text generation model has the ability to understand the domain knowledge and domain concepts of the target domain. The pre-trained text generation model can be a neural network model, for example, a neural network model based on the Transformer architecture, for example, a GPT (Generative Pre-Trained Transformer) model.

[0072] The standard domain concept extraction result of the sample domain knowledge includes the domain concepts existing in the sample domain knowledge. For example, if the sample domain knowledge is "the palmprint recognition method is a method of identifying an identity by extracting palmprint features from a palmprint image", then the domain concepts existing in the sample domain knowledge are "palmprint recognition", "palmprint image" and "palmprint feature". The standard domain concept extraction result of the sample domain knowledge can be a text that conforms to natural language, for example, "the extracted domain concepts are palmprint recognition, palmprint image and palmprint feature". The domain concepts in the standard domain concept extraction result can be extracted from the sample domain knowledge by the server through the entity extraction algorithm. The entity extraction method includes but is not limited to CRF (Conditional Random Field) and BiLSTM-CRF. The full name of BiLSTM is Bi-directional Long Short-Term Memory network.

[0073] Specifically, the domain knowledge of the target domain is determined by the corpus of the target domain. The corpus in the corpus can come from text data sets or professional books, for example, from papers, manuals, standards, patents, domain dictionaries or technical documents. The corpus can be understood as a collection of texts. The corpus in the corpus can be text, audio or images. A text can be a character, a word, a sentence, multiple sentences, a paragraph or an article, etc. Multiple sentences refer to at least two sentences. It should be noted that the audio or image in the corpus is in text form, that is, text is used to represent audio and images. The domain knowledge of the target domain can be the corpus in the corpus of the target domain, for example, it can be a sentence, a paragraph or an article, etc.

[0074] In some embodiments, the server may preprocess the corpus, and the preprocessing includes but is not limited to noise removal, sentence segmentation, segmentation, stop word removal, etc., to obtain a processed corpus. The domain knowledge of the target domain may be the corpus in the processed corpus, for example, sentences, paragraphs or articles in the processed corpus.

[0075] In some embodiments, the server inputs the sample domain knowledge into the pre-trained text generation model to obtain the output result of the pre-trained text generation model. The server generates a second model loss based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge. The difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge is positively correlated with the second model loss.

[0076] In some embodiments, the server can calculate the gradient of the second model loss relative to the model parameters of the pre-trained text generation model, and adjust the model parameters along the direction of gradient descent, thereby realizing the training of the pre-trained text generation model. There can be multiple sample domain knowledge, and the server can continuously train the pre-trained text generation model according to different sample domain knowledge until the pre-trained text generation model converges to obtain the first text generation model. The first text generation model is the pre-trained text generation model at the time of convergence. During the training process, the hyperparameters of the model can be tuned, and the hyperparameters of the model include but are not limited to batch size (BatchSize) and learning rate (learning rate). During the training process, the server can verify and evaluate the model through the validation set and the test set, both of which are sets of domain knowledge. To evaluate the performance and accuracy of the model. When the performance and accuracy meet the preset requirements, the model converges. Figure 6As shown, the model extracts domain concepts from domain knowledge, evaluates the extracted domain concepts, and determines whether the extracted domain concepts are accurate. If they are inaccurate, the model parameters are updated to improve the rationality of the domain concepts extracted by the model. When the performance and accuracy of the model meet the preset requirements, the server can store the proposed domain concepts in correspondence with the corresponding domain knowledge.

[0077] In some embodiments, the server may generate a word list based on a corpus. The text in the corpus is segmented to obtain a segmentation set. The segmentations in the segmentation set are different. A segmentation table is constructed based on some or all of the segmentations in the segmentation set. For example, the server may determine the number of occurrences of each segmentation in the segmentation set in the corpus, and determine the frequency of occurrence of the segmentation based on the number of occurrences of the segmentation. The segmentations in the segmentation table may be arranged from large to small according to the frequency of occurrence. The greater the probability of occurrence of a segmentation, the higher the position of the segmentation in the segmentation table. A preset number of segmentations may be selected from the segmentation set in order of frequency of occurrence from large to small to form a segmentation table. The preset number is used to limit the size of the segmentation table so that the size of the segmentation table is appropriate. After obtaining the sample domain knowledge, the server may segment the sample domain knowledge through the segmentation table to obtain a segmentation sequence, and input the segmentation sequence into the pre-trained text generation model to speed up the processing efficiency of the text.

[0078] In this embodiment, based on the difference between the output results of the pre-trained text generation model and the standard domain concept extraction results of the sample domain knowledge, the model parameters of the pre-trained text generation model are adjusted to obtain a first text generation model, so that the first text generation model has the ability to extract domain concepts.

[0079] In some embodiments, obtaining a pre-trained text generation model includes: obtaining multiple sample texts; a sample text among the multiple sample texts belongs to a target domain, at least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is domain knowledge of the target domain; for each sample text, inputting the sample text into the text generation model to be trained to obtain an output result of the text generation model to be trained; based on the difference between the sample text and the output result of the text generation model to be trained, adjusting the model parameters of the text generation model to be trained to obtain a pre-trained text generation model.

[0080] Among them, multiple sample texts refer to at least two sample texts. At least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is domain knowledge of the target domain. For example, the target domain is the palmprint recognition field, and the multiple sample texts include sample text 1 and sample text 2, sample text 1 is the domain knowledge of the palmprint recognition field, and sample text 2 is the domain concept of the palmprint recognition field. Then sample text 1 is, for example, "the palmprint recognition method is a method of identifying an identity by extracting palmprint features from a palmprint image", and sample text 2 is, for example, "palmprint image". The text generation model to be trained can be a neural network model, for example, it can be a neural network model based on the Transformer architecture.

[0081] Specifically, the server inputs the sample text into the text generation model to be trained to obtain the output result of the text generation model to be trained. The output result of the text generation model to be trained is in text form, and the server can generate a third model loss based on the difference between the sample text and the output result of the text generation model to be trained. The difference between the sample text and the output result of the text generation model to be trained is positively correlated with the third model loss. The server can calculate the gradient of the third model loss relative to the model parameters of the text generation model to be trained, and adjust the model parameters in the direction of gradient descent, thereby realizing the training of the text generation model to be trained. The server continuously trains the text generation model to be trained until the text generation model to be trained converges, and obtains a pre-trained text generation model.

[0082] In this embodiment, based on the difference between the sample text and the output result of the text generation model to be trained, the model parameters of the text generation model to be trained are adjusted to obtain a pre-trained text generation model. Therefore, the purpose of training is to make the output of the model consistent with the input, that is, to make the model have the function of generating sample text. Since at least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is the domain knowledge of the target domain, the pre-trained text generation model has the ability to generate domain concepts and domain knowledge, so that the pre-trained text generation model has the ability to understand the domain knowledge of the target domain and the domain concepts of the target domain.

[0083] In some embodiments, generating a domain concept relationship graph based on the dependency relationships between target domain concepts includes: taking each target domain concept as a node and the dependency relationships between target domain concepts as an edge; and generating a domain concept relationship graph based on each node and each edge.

[0084] The domain concept relationship graph is composed of nodes and edges, and the nodes represent the target domain concepts. The dependency relationships between the target domain concepts include equivalence relationships, inclusion relationships, etc. Since the second text generation model is used to identify the domain concepts in the target domain and the dependency relationships between the domain concepts, the second text generation model can identify the relationships that are not presented in the target system architecture diagram. For example, the target system architecture diagram is Figure 3 , the target description text is "the palm swiping system includes palmprint recognition, equipment management, card swiping, system management, data backup, report statistics and data recovery. Among them, palmprint recognition includes palmprint comparison, palmprint feature extraction, palmprint recognition algorithm, palmprint library and palmprint classification. Equipment management includes equipment maintenance, equipment configuration, equipment status detection, equipment monitoring and equipment diagnosis. Card swiping includes card reader, card swiping operation, ID card recognition and card processor. System management includes system configuration, log management, performance monitoring, user authority management and exception handling. " It can be seen that neither the target description text nor the target system architecture diagram shows the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction". Since the second text generation model has the ability to identify the dependency relationship between domain concepts, the second text generation model can identify the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction", that is, "palmprint recognition algorithm" depends on "palmprint feature extraction".

[0085] Specifically, the server can generate an initial graph based on the dependency relationships presented in the target description text, and determine the dependency relationships not described in the target description text, that is, the dependency relationships not presented in the target system architecture diagram, from the identified dependency relationships between the target domain concepts, and supplement the not presented dependency relationships on the basis of the initial graph to generate a domain concept relationship graph. For example, it is identified that there is a dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction", and the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction" is not presented in the target system architecture diagram. Then, based on the target system architecture diagram, the dependency relationship between "palmprint recognition algorithm" and "palmprint feature extraction" is used as an edge connecting "palmprint recognition algorithm" and "palmprint feature extraction". Figure 4 As shown, the generated domain concept relationship diagram is displayed.

[0086] In some embodiments, the server uses each target domain concept as a node and the dependency relationship between the target domain concepts as an edge to generate a domain concept relationship graph with a tree structure. Figure 7 As shown in the figure, the generated domain concept relationship diagram is shown. The domain concept relationship diagram can be a tree structure, in which the root node is located at the first level, the child node of the root node is located at the second level, and so on. The target system architecture diagram is also designed in layers, for example, Figure 3In the example, "palm scan system" is the first layer, "palmprint recognition", "device management", "card swiping", "system management", "data backup", "report statistics", and "data recovery" are the second layer, and the remaining domain concepts are the third layer. When performing redundancy detection, the server can determine the target domain concepts with a level greater than the number of layers in the target system architecture diagram, such as 3 layers, from the domain concept relationship diagram, and determine the target domain concepts with a level greater than the number of layers as redundant target domain concepts. For example, Figure 7 In the figure, “palmprint classification”, “palmprint comparison”, “palmprint feature extraction”, “card processor”, “device diagnosis” and “device status detection” are at the fourth level, so “palmprint classification”, “palmprint comparison”, “palmprint feature extraction”, “card processor”, “device diagnosis” and “device status detection” are all redundant target domain concepts.

[0087] In this embodiment, each target domain concept is taken as a node, and the dependency relationship between the target domain concepts is taken as an edge. A domain concept relationship graph is generated based on each node and each edge, thereby quickly generating a domain concept relationship graph.

[0088] In some embodiments, the domain concept relationship graph includes a target loop, and the target domain concepts constituting the target loop include redundant target domain concepts. The redundancy detection result includes a loop detection result: performing domain concept redundancy detection on the domain concept relationship graph, and obtaining the redundant detection result of the domain concept relationship graph includes: performing loop detection on the domain concept relationship graph, and obtaining the loop detection result; the loop detection result includes the dependency relationship between the target domain concepts constituting the target loop.

[0089] Among them, loop detection refers to detecting whether loops are stored in the domain concept relationship graph. A loop refers to a route with the same starting point and end point. The loop detection results include: the target domain concepts that constitute the target loop, and the dependency relationships between the target domain concepts that constitute the target loop. Figure 4 In the example, "palmprint recognition" → "palmprint comparison" → "palmprint recognition algorithm" → "palmprint recognition" is a loop. The loop detection result includes "palmprint recognition", "palmprint comparison" and "palmprint recognition algorithm", and also includes the dependency relationship between "palmprint recognition", "palmprint comparison" and "palmprint recognition algorithm". For example, "palmprint recognition" has a dependency relationship with "palmprint recognition algorithm" and "palmprint comparison", and there is a dependency relationship between "palmprint recognition algorithm" and "palmprint comparison".

[0090] Specifically, the server may use a topological sorting algorithm to sort the domain concept relationship graph, thereby detecting whether there is a loop therein. The domain concept relationship graph may include at least one loop.

[0091] In this embodiment, loop detection is performed on the domain concept relationship graph to obtain loop detection results, so that redundant target domain concepts in the domain concept relationship graph can be accurately and quickly detected. Figure 4 In the process of “palmprint recognition” → “palmprint recognition algorithm” → “palmprint comparison” → “palmprint recognition”, it can be determined that there is redundancy. Since the “palmprint recognition algorithm” depends on the “palmprint comparison”, it can be determined that the “palmprint comparison” is redundant.

[0092] In some embodiments, the redundancy detection result also includes a repeated concept detection result: performing domain concept redundancy detection on the domain concept relationship graph, and obtaining the redundant detection result of the domain concept relationship graph also includes: detecting the target domain concepts with repeated dependencies in the domain concept relationship graph, and obtaining the repeated concept detection result.

[0093] The dependency relationship between the two target domain concepts is duplication, which means that the two target domain concepts are essentially duplicate domain concepts, that is, the meanings expressed by the two target domain concepts are consistent. The duplication concept detection result includes the duplicate target domain concepts.

[0094] In this embodiment, target domain concepts with repeated dependencies in the domain concept relationship diagram are detected to obtain repeated concept detection results, so that target domain concepts with consistent meanings can be detected, so as to facilitate deduplication of domain concepts in the target system architecture diagram, thereby improving the rationality of the target system architecture diagram.

[0095] In some embodiments, the target field is the palmprint recognition field, and the target system is a palmprint recognition system; obtaining a target description text describing the target system architecture diagram includes: receiving the target description text sent by a terminal corresponding to the palmprint recognition system; the method also includes: sending the redundant detection result to the terminal corresponding to the palmprint recognition system.

[0096] Specifically, the terminal corresponding to the palmprint recognition system may be a terminal of a developer of the palmprint recognition system or a person who designs an architecture diagram of the palmprint recognition system.

[0097] In some embodiments, the terminal sends a redundant identification request to the server, and the redundant identification request carries a target description text. The server inputs the target description text into the second text generation model, and obtains the target domain concept and the dependency relationship between the target domain concepts in the target description text through the second text generation model. The server generates a domain concept relationship graph based on the dependency relationship between the target domain concept and the target domain concept. The server performs a redundant detection on the domain concept relationship graph to obtain a redundant detection result of the domain concept relationship graph. The server returns the redundant detection result to the terminal corresponding to the palmprint recognition system.

[0098] In this embodiment, the redundant detection result is sent to the terminal corresponding to the palmprint recognition system, so that the personnel of the terminal corresponding to the palmprint recognition system can optimize the architecture diagram of the palmprint recognition system.

[0099] In some embodiments, the method also includes: obtaining a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; inputting sample domain concepts and domain knowledge related to the sample domain concepts into the pre-trained text generation model to obtain output text of the pre-trained text generation model; based on the difference between the output text and the descriptive text of the sample domain concept, adjusting the parameters of the pre-trained text generation model to obtain a trained text generation model; the trained text generation model is used to generate descriptive text of the domain concept.

[0100] The sample domain concept may be any domain concept in the target domain. The domain knowledge related to the sample domain concept may be a sentence or paragraph related to the sample domain concept. The description text of the sample domain concept is used to describe or explain the meaning of the sample domain concept.

[0101] Specifically, the server inputs the sample domain concept and the domain knowledge related to the sample domain concept into the pre-trained text generation model to obtain the output text of the pre-trained text generation model. The server generates the fourth model loss based on the difference between the output text and the description text. The difference between the output text and the description text is positively correlated with the fourth model loss.

[0102] In some embodiments, the server may calculate the gradient of the fourth model loss relative to the model parameters of the pre-trained text generation model, and adjust the model parameters along the direction of gradient descent, thereby implementing the training of the pre-trained text generation model. There may be multiple sample domain concepts, and the server may continuously train the pre-trained text generation model according to different sample domain concepts until the pre-trained text generation model converges to obtain a trained text generation model. This enables the trained text generation model to have the ability to generate descriptive text of domain concepts.

[0103] In some embodiments, the server receives the domain concept sent by the terminal, inputs the domain concept into a trained text generation model, predicts the descriptive text corresponding to the domain concept, and returns the descriptive text corresponding to the domain concept to the terminal, so that the user of the terminal can understand the meaning of the domain concept more quickly.

[0104] In some embodiments, the server inputs the sample domain concept and the domain knowledge related to the sample domain concept into the pre-trained text generation model, and the pre-trained text generation model includes a text encoder, and the sample domain concept and the domain knowledge related to the sample domain concept are encoded by the text encoder to obtain hidden features, and the hidden features include the features of the sample domain concept and the features of the domain knowledge related to the sample domain concept. The pre-trained text generation model also includes a feature fusion network, and the hidden features are fused with the features of the domain knowledge related to the sample domain concept by the feature fusion network to obtain fused features. The pre-trained text generation model also includes a text decoder, and the fused features are decoded by the text decoder to obtain output text. Among them, the feature fusion network can be a neural network based on an attention mechanism.

[0105] In some embodiments, the server may input the domain knowledge and the domain concepts in the domain knowledge into a trained text generation model, encode the domain knowledge and the domain concepts in the domain knowledge through a text encoder, obtain the features of the domain knowledge and the features of the domain concepts, and fuse the features of the domain knowledge and the features of the domain concepts through a feature fusion network to obtain target fusion features. The target fusion features can be understood as the semantic representation and description of the domain concepts. The target fusion features can be used to classify domain concepts, and can also be used to determine whether a text is related to the domain concept.

[0106] In this embodiment, based on the difference between the output text and the descriptive text of the sample domain concepts, the parameters of the pre-trained text generation model are adjusted to obtain a trained text generation model, so that the trained text generation model has the ability to generate descriptive texts of domain concepts, so as to quickly generate descriptive texts corresponding to the domain concepts according to the trained text generation model, so as to facilitate a faster understanding of the meaning of the domain concepts.

[0107] In some embodiments, Figure 8 As shown, a system architecture diagram processing method is provided. The method can be executed by a terminal or a server, or can be executed by a terminal and a server together. The method is applied to a server as an example for explanation, and includes the following steps:

[0108] Step 802, obtaining a plurality of sample texts; a sample text among the plurality of sample texts belongs to a target domain.

[0109] Among them, at least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is domain knowledge of the target domain.

[0110] Step 804: for each sample text, input the sample text into the text generation model to be trained to obtain the output result of the text generation model to be trained.

[0111] Step 806: Based on the difference between the sample text and the output result of the text generation model to be trained, adjust the model parameters of the text generation model to be trained to obtain a pre-trained text generation model.

[0112] Step 808, obtaining sample domain knowledge; inputting the sample domain knowledge into the pre-trained text generation model to obtain the output result of the pre-trained text generation model.

[0113] Among them, the sample domain knowledge belongs to the target domain.

[0114] Step 810: Based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge, the model parameters of the pre-trained text generation model are adjusted to obtain a first text generation model.

[0115] Step 812: input the training text into the first text generation model to obtain the output result of the first text generation model.

[0116] Among them, the training text is the text belonging to the target field.

[0117] Step 814: Based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, the first text generation model is adjusted to obtain a second text generation model.

[0118] Among them, the standard domain concept relationship includes the training domain concept and the dependency relationship between the training domain concepts, and the training domain concept refers to the domain concept included in the training text.

[0119] Step 816, obtaining a target description text describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in a target domain.

[0120] Step 818: input the target description text into the second text generation model to identify the dependency relationship between the target domain concepts in the target description text.

[0121] Step 820: Generate a domain concept relationship graph based on the dependency relationships between target domain concepts.

[0122] Step 822, performing redundancy detection on the domain concept relationship graph to obtain a redundancy detection result of the domain concept relationship graph.

[0123] In this embodiment, a target description text describing a target system architecture diagram is obtained, the target system architecture diagram is an architecture diagram of a target system in a target domain, the target description text is input into a second text generation model, the dependency relationship between target domain concepts in the target description text is identified, the second text generation model is used to identify the domain concepts of the target domain and the dependency relationship between domain concepts, a domain concept relationship graph is generated based on the dependency relationship between the target domain concepts, a redundancy detection is performed on the domain concept relationship graph, and a redundancy detection result of the domain concept relationship graph is obtained, and the redundancy detection result is used to perform redundancy removal processing on the target system architecture diagram. Thus, the dependency relationship between the target domain concepts is quickly and accurately identified by the second text generation model, and a redundancy detection result is obtained by generating a domain concept relationship graph and performing redundancy detection on the domain concept relationship graph, so as to facilitate the redundancy removal processing of the target system architecture diagram using the redundancy detection result, so as to improve the rationality of the target system architecture diagram.

[0124] The system architecture diagram processing method provided in the present application can be used for redundancy removal of system architecture diagrams in any field. The fields include but are not limited to face recognition, fingerprint recognition, palmprint recognition and other fields. Palmprint recognition is used as an example for illustration. The server can obtain a target description text by describing the architecture diagram of the palm-swiping system. The architecture diagram of the palm-swiping system can be manually designed or generated by a computer. The server can input the target description text into the second text generation model, identify the target domain concepts in the target description text and the dependencies between the target domain concepts, and generate a domain concept relationship diagram based on the dependencies between the target domain concepts and the palmprint recognition domain concepts. Then, the server performs redundancy detection on the domain concept relationship diagram, obtains the redundancy detection results of the domain concept relationship diagram, and performs redundancy removal on the architecture diagram of the palm-swiping system through the redundancy detection results.

[0125] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0126] Based on the same inventive concept, the embodiment of the present application also provides a system architecture diagram processing device for implementing the system architecture diagram processing method involved above. The implementation solution for solving the problem provided by the device is similar to the implementation solution recorded in the above method, so the specific limitations in the one or more system architecture diagram processing device embodiments provided below can refer to the limitations of the system architecture diagram processing method above, and will not be repeated here.

[0127] In some embodiments, Fig. 9 As shown, a system architecture diagram processing device is provided, including: a description text acquisition module 902, a dependency identification module 904, a relationship diagram generation module 906 and a redundancy detection module 908, wherein:

[0128] The description text acquisition module 902 is used to acquire a target description text describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in a target domain.

[0129] The dependency identification module 904 is used to identify the dependency between target domain concepts in the target description text.

[0130] The relationship graph generation module 906 is used to generate a domain concept relationship graph based on the dependency relationships between target domain concepts.

[0131] The redundancy detection module 908 is used to perform domain concept redundancy detection on the domain concept relationship diagram to obtain the redundancy detection result of the domain concept relationship diagram; the redundancy detection result is used to perform redundancy removal processing on the target domain concepts expressed in the target system architecture diagram.

[0132] In some embodiments, the dependency relationship is identified through a second text generation model; the device also includes a model acquisition module; the model acquisition module is used to obtain a first text generation model with domain concept recognition capability; the training text is input into the first text generation model to obtain the output result of the first text generation model; the training text is a text belonging to the target domain; based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, the first text generation model is adjusted to obtain the second text generation model; the standard domain concept relationship includes the dependency relationship between each training domain concept, and the training domain concept refers to the domain concept included in the training text.

[0133] In some embodiments, the training text is target domain knowledge including multiple domain concepts in the target domain; the model acquisition module is also used to input the target domain knowledge into the first text generation model to obtain the output result of the first text generation model.

[0134] In some embodiments, the model acquisition module is further used to construct a training text based on multiple domain concepts of the target domain; input the training text into the first text generation model to obtain the output result of the first text generation model.

[0135] In some embodiments, the model acquisition module is also used to obtain a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; obtain sample domain knowledge; the sample domain knowledge belongs to the target domain; input the sample domain knowledge into the pre-trained text generation model to obtain the output result of the pre-trained text generation model; based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge, adjust the model parameters of the pre-trained text generation model to obtain the first text generation model.

[0136] In some embodiments, the model acquisition module is also used to obtain multiple sample texts; the sample texts among the multiple sample texts belong to the target domain, at least one of the multiple sample texts is a domain concept of the target domain, and at least one of the multiple sample texts is domain knowledge of the target domain; for each sample text, the sample text is input into the text generation model to be trained to obtain the output result of the text generation model to be trained; based on the difference between the sample text and the output result of the text generation model to be trained, the model parameters of the text generation model to be trained are adjusted to obtain a pre-trained text generation model.

[0137] In some embodiments, the relationship graph generation module 906 is further configured to use each target domain concept as a node, and use the dependency relationship between the target domain concepts as an edge; and generate a domain concept relationship graph based on each node and each edge.

[0138] In some embodiments, the domain concept relationship graph includes a target loop, and the target domain concepts constituting the target loop include redundant target domain concepts, and the redundant detection result includes a loop detection result: the redundant detection module 908 is also used to perform loop detection on the domain concept relationship graph to obtain a loop detection result; the loop detection result includes the dependency relationship between the target domain concepts constituting the target loop.

[0139] In some embodiments, the redundancy detection result also includes a repeated concept detection result: the redundancy detection module 908 is further used to detect target domain concepts with repeated dependencies in the domain concept relationship graph to obtain repeated concept detection results.

[0140] In some embodiments, the target field is the palmprint recognition field, and the target system is a palmprint recognition system; obtaining a target description text describing the target system architecture diagram includes: receiving the target description text sent by a terminal corresponding to the palmprint recognition system; the method also includes: sending the redundant detection result to the terminal corresponding to the palmprint recognition system.

[0141] In some embodiments, the device is also used to: obtain a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; input sample domain concepts and domain knowledge related to the sample domain concepts into the pre-trained text generation model to obtain output text of the pre-trained text generation model; based on the difference between the output text and the descriptive text of the sample domain concept, adjust the parameters of the pre-trained text generation model to obtain a trained text generation model; the trained text generation model is used to generate descriptive text of the domain concept.

[0142] Each module in the above-mentioned system architecture diagram processing device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0143] In some embodiments, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data involved in the system architecture diagram processing method. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a system architecture diagram processing method is implemented.

[0144] In some embodiments, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.11As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a system architecture diagram processing method is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.

[0145] Those skilled in the art will understand that Fig.10 and Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0146] In some embodiments, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned system architecture diagram processing method when executing the computer program.

[0147] In some embodiments, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned system architecture diagram processing method are implemented.

[0148] In some embodiments, a computer program product is provided, including a computer program, which implements the steps in the above-mentioned system architecture diagram processing method when executed by a processor.

[0149] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0150] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0151] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0152] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing a system architecture diagram, It is characterized in that The method comprises: Obtaining a target description text describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in the target field; Identifying dependencies between target domain concepts in the target description text; Generate a domain concept relationship graph based on the dependency relationship between the target domain concepts; A domain concept redundancy check is performed on the domain concept relationship diagram to obtain a redundancy check result of the domain concept relationship diagram; the redundancy check result is used to perform redundancy removal processing on the target domain concept expressed in the target system architecture diagram.

2. The method according to claim 1, It is characterized in that The dependency relationship is identified by a second text generation model; The step of obtaining the second text generation model comprises: Obtaining a first text generation model capable of recognizing domain concepts; Inputting a training text into the first text generation model to obtain an output result of the first text generation model; the training text is a text belonging to the target field; Based on the difference between the output result of the first text generation model and the standard domain concept relationship of the training text, the first text generation model is adjusted to obtain a second text generation model; the standard domain concept relationship includes the dependency relationship between each training domain concept, and the training domain concept refers to the domain concept included in the training text.

3. The method according to claim 2, It is characterized in that The training text is target domain knowledge including multiple domain concepts in the target domain; The step of inputting the training text into the first text generation model to obtain an output result of the first text generation model includes: The target domain knowledge is input into the first text generation model to obtain an output result of the first text generation model.

4. The method according to claim 2, It is characterized in that The step of inputting the training text into the first text generation model to obtain an output result of the first text generation model includes: Forming training text based on multiple domain concepts of the target domain; The training text is input into the first text generation model to obtain an output result of the first text generation model.

5. The method according to claim 2, It is characterized in that The step of obtaining a first text generation model capable of recognizing domain concepts includes: Acquire a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; Acquire sample domain knowledge; the sample domain knowledge belongs to the target domain; Inputting the sample domain knowledge into the pre-trained text generation model to obtain an output result of the pre-trained text generation model; Based on the difference between the output result of the pre-trained text generation model and the standard domain concept extraction result of the sample domain knowledge, the model parameters of the pre-trained text generation model are adjusted to obtain a first text generation model.

6. The method according to claim 5, It is characterized in that The obtaining of the pre-trained text generation model comprises: Acquire a plurality of sample texts; sample texts among the plurality of sample texts belong to the target domain, at least one of the plurality of sample texts is a domain concept of the target domain, and at least one of the plurality of sample texts is domain knowledge of the target domain; For each of the sample texts, the sample text is input into a text generation model to be trained to obtain an output result of the text generation model to be trained; Based on the difference between the sample text and the output result of the text generation model to be trained, the model parameters of the text generation model to be trained are adjusted to obtain the pre-trained text generation model.

7. The method according to claim 1, It is characterized in that The generating of the domain concept relationship graph based on the dependency relationship between the target domain concepts comprises: Taking each of the target domain concepts as a node, and taking the dependency relationship between the target domain concepts as an edge; A domain concept relationship graph is generated based on the nodes and the edges.

8. The method according to claim 1, It is characterized in that The domain concept relationship graph includes a target loop, each of the target domain concepts constituting the target loop includes a redundant target domain concept, and the redundancy detection result includes a loop detection result: The performing domain concept redundancy detection on the domain concept relationship graph to obtain a redundancy detection result of the domain concept relationship graph includes: A loop detection is performed on the domain concept relationship graph to obtain a loop detection result; the loop detection result includes the dependency relationship between each of the target domain concepts constituting the target loop.

9. The method according to claim 8, It is characterized in that The redundant detection results also include repeated concept detection results: The performing domain concept redundancy detection on the domain concept relationship graph to obtain a redundancy detection result of the domain concept relationship graph further includes: Detect target domain concepts with repeated dependency relationships in the domain concept relationship graph to obtain repeated concept detection results.

10. The method according to claim 1, It is characterized in that The target field is the palmprint recognition field, and the target system is the palmprint recognition system; The obtaining of the target description text describing the target system architecture diagram includes: Receiving the target description text sent by the terminal corresponding to the palmprint recognition system; The method further comprises: The redundant detection result is sent to a terminal corresponding to the palmprint recognition system.

11. The method according to any one of claims 1 to 10, It is characterized in that The method further comprises: Acquire a pre-trained text generation model; the pre-trained text generation model is used to generate domain knowledge or domain concepts of the target domain; Inputting a sample domain concept and domain knowledge related to the sample domain concept into the pre-trained text generation model to obtain an output text of the pre-trained text generation model; Based on the difference between the output text and the description text of the sample domain concept, the parameters of the pre-trained text generation model are adjusted to obtain a trained text generation model; the trained text generation model is used to generate the description text of the domain concept.

12. A system architecture diagram processing device, It is characterized in that The device comprises: A description text acquisition module is used to acquire a target description text describing a target system architecture diagram; the target system architecture diagram is an architecture diagram of a target system in the target field; A dependency identification module, used to identify the dependency between target domain concepts in the target description text; A relationship graph generation module, used to generate a domain concept relationship graph based on the dependency relationship between the target domain concepts; A redundancy detection module is used to perform domain concept redundancy detection on the domain concept relationship diagram to obtain a redundancy detection result of the domain concept relationship diagram; the redundancy detection result is used to perform redundancy removal processing on the target domain concept expressed in the target system architecture diagram.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.

14. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

15. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images