Archives intelligent retrieval method, device and electronic equipment based on big data

Abstract

The present application relates to the field of intelligent retrieval, and specifically discloses a method, device and electronic device for intelligent archive retrieval based on big data, which uses big data technology and artificial intelligence technology based on the field of deep learning to perform feature extraction and encoding on archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, which improves the comprehensiveness and accuracy of detection.

Description

Technical Field

[0001] The present application relates to the field of intelligent retrieval, and more specifically to an intelligent archive retrieval method, device and electronic device based on big data. Background Art

[0002] With the development of science and technology, artificial intelligence (AI) has experienced highs and lows, but it has always been on the main track of scientific and technological innovation. As the core driving force of industrial transformation, AI empowers traditional businesses, frees people from boring labor, reduces manpower input, and improves work efficiency. At present, archives departments are also actively exploring and looking for new ways and methods to promote the transformation of archives management models using new generation information technologies such as artificial intelligence and big data. In office scenarios, traditional archive retrieval can only retrieve structured information in the system due to defects in the search engine, and the recall rate and precision rate are not satisfactory.

[0003] Therefore, an optimized intelligent archive retrieval solution is expected. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiments of this application provide a method, device and electronic device for intelligent archive retrieval based on big data, which uses big data technology and artificial intelligence technology based on the field of deep learning to perform feature extraction and encoding on archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, which improves the comprehensiveness and accuracy of detection.

[0005] According to one aspect of the present application, a method for intelligent archive retrieval based on big data is provided, which includes:

[0006] Obtain archival search terms and archival search databases;

[0007] Analyzing the archive retrieval database to obtain an archive database feature vector;

[0008] Processing the archive search term to obtain an archive search term feature vector;

[0009] The archive search term feature vector and the archive database feature vector are comprehensively analyzed to generate an archive search result.

[0010] According to another aspect of the present application, there is also provided a file intelligent retrieval device based on big data, which includes:

[0011] The archive data acquisition module is used to obtain archive search terms and archive search databases;

[0012] An archive retrieval database analysis module is used to analyze the archive retrieval database to obtain an archive database feature vector;

[0013] An archive search term analysis module, used for processing the archive search term to obtain an archive search term feature vector;

[0014] The retrieval result generating module is used to comprehensively analyze the archive retrieval term feature vector and the archive database feature vector to generate the archive retrieval result.

[0015] According to another aspect of the present application, there is also provided an electronic device, comprising:

[0016] one or more processors;

[0017] A memory for storing one or more programs;

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the big data-based intelligent archive retrieval method.

[0019] Compared with the prior art, the intelligent archive retrieval method, device and electronic device based on big data provided by the present application use big data technology and artificial intelligence technology based on the field of deep learning to extract and encode features of archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, which improves the comprehensiveness and accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 The present invention is a flowchart of an intelligent archive retrieval method based on big data according to an embodiment of the present application.

[0022] Figure 2 The present invention is a flowchart for analyzing the archive retrieval database in the big data-based intelligent archive retrieval method according to an embodiment of the present application to obtain an archive database feature vector.

[0023] Figure 3 The present invention is a flowchart of performing multi-scale feature extraction on the multiple archive context semantic feature vectors to obtain the archive database feature vector in the big data-based archive intelligent retrieval method according to an embodiment of the present application.

[0024] Figure 4 It is a block diagram of an intelligent archive retrieval device based on big data according to an embodiment of the present application.

[0025] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0027] Figure 1 FIG. 1 is a flow chart of an intelligent archive retrieval method based on big data according to an embodiment of the present application. Figure 1 As shown, the intelligent archive retrieval method based on big data according to the embodiment of the present application includes: S110, obtaining archive retrieval terms and archive retrieval database; S120, analyzing the archive retrieval database to obtain an archive database feature vector; S130, processing the archive retrieval terms to obtain an archive retrieval term feature vector; and S140, comprehensively analyzing the archive retrieval term feature vector and the archive database feature vector to generate an archive retrieval result.

[0028] In step S110, archive search terms and archive search databases are obtained. As mentioned in the above background technology, in office scenarios, due to the defects of the search engine, traditional archive search can only search for structured information in the system, and the recall rate and precision rate are not satisfactory. Therefore, an optimized archive intelligent search solution is expected.

[0029] In response to the above technical problems, a big data-based intelligent archive retrieval method is proposed, which uses big data technology and artificial intelligence technology based on deep learning to extract and encode features of archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, improving the comprehensiveness and accuracy of detection.

[0030] Specifically, first, obtain archive search terms and archive search databases. Archive search terms are keywords or phrases entered by users to query archives, and are a way of expressing user needs. By obtaining archive search terms, the system can understand the user's search intent and accurately match relevant archive data. Archive search databases are data sets that store a large amount of archive information, including archive content that needs to be retrieved. The purpose of obtaining archive search databases is to extract the semantic features of archives from them so as to match them with the search terms entered by users, thereby generating relevant search results.

[0031] In step S120, the archive retrieval database is analyzed to obtain an archive database feature vector.

[0032] Figure 2 The flowchart of analyzing the archive retrieval database in the big data-based intelligent archive retrieval method according to the embodiment of the present application to obtain the archive database feature vector. Figure 2 As shown, analyzing the archive retrieval database to obtain an archive database feature vector includes: S121, passing the archive retrieval database through an archive database feature extraction module including an embedding layer to obtain a plurality of archive context semantic feature vectors; and S122, performing multi-scale feature extraction on the plurality of archive context semantic feature vectors to obtain an archive database feature vector.

[0033] Specifically, in order to extract and represent the semantic information of archival data, the archival retrieval database is subjected to an archival database feature extraction module including an embedding layer based on a converter context encoder model to obtain multiple archival context semantic feature vectors, thereby providing richer and more accurate features for subsequent retrieval and matching processes. The archival database feature extraction module is a context encoder model including an embedding layer based on a converter. The information in the archives is often complex and abstract. By using a converter-based context encoder model, the text data can be converted into a high-dimensional semantic vector representation, thereby better capturing the semantic information of the archival content. The converter-based model can learn the semantic features in the text data and convert it into a continuous vector space representation. These feature vectors contain the semantic information of the text, which helps to distinguish the similarities and differences between different archives. The context encoder model can consider the context information in the text data, thereby better understanding the context and meaning of the text. This helps to improve the understanding and representation capabilities of archival data. By obtaining multiple archival context semantic feature vectors, representations at different levels and angles can be obtained, making the semantic information of the archival data richer and more comprehensive.

[0034] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the archive database feature extraction module is a context encoder model based on a converter including an embedding layer.

[0035] Specifically, in the above-mentioned big data-based intelligent archive retrieval method, the archive retrieval database is passed through an archive database feature extraction module including an embedding layer to obtain a plurality of archive context semantic feature vectors, including: performing word segmentation processing on the archive retrieval database to obtain a plurality of archive words; passing the plurality of archive words through an embedding layer to convert each of the plurality of archive words into an archive word embedding vector to obtain a sequence of archive word embedding vectors, wherein the embedding layer uses a learnable embedding matrix to embed-code each archive word; and inputting the sequence of archive word embedding vectors into the archive database feature extraction module to obtain the plurality of archive context semantic feature vectors.

[0036] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the sequence of the archive word embedding vectors is input into the archive database feature extraction module to obtain the multiple archive semantic feature vectors, including: arranging the sequence of the archive word embedding vectors as an input vector; converting the input vector into a query vector and a key vector respectively through a learnable embedding matrix; calculating the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalizing the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and, multiplying the self-attention feature matrix with each archive word embedding vector in the sequence of the archive word embedding vectors to obtain the multiple archive context semantic feature vectors.

[0037] Figure 3 The present invention is a flowchart of performing multi-scale feature extraction on the plurality of archive context semantic feature vectors to obtain archive database feature vectors in the archive intelligent retrieval method based on big data according to an embodiment of the present application. Figure 3 As shown, Figure 3 As shown, multi-scale feature extraction is performed on the multiple archival context semantic feature vectors to obtain an archival database feature vector, including: S11, cascading the multiple archival context semantic feature vectors to obtain a first-scale archival semantic feature vector; S12, performing mixed feature extraction on the multiple archival context semantic feature vectors to obtain a second-scale archival semantic feature vector; and S13, fusing the first-scale archival semantic feature vector and the second-scale archival semantic feature vector to obtain an archival database feature vector.

[0038] More specifically, in order to integrate semantic information at different levels and angles, multiple archival context semantic feature vectors are cascaded to obtain the first-scale archival semantic feature vector, thereby obtaining a more comprehensive and integrated archival representation. Different archival context semantic feature vectors may capture semantic information at different levels and angles. By cascading these feature vectors, multiple semantic information can be comprehensively considered, making the obtained first-scale archival semantic feature vector more comprehensive and rich. During the cascading process, information is not lost, but integrated into a higher-dimensional vector. This helps to retain important information in the original feature vector and avoid information loss.

[0039] Specifically, in the above-mentioned big data-based intelligent archive retrieval method, hybrid feature extraction is performed on the multiple archive context semantic feature vectors to obtain a second-scale archive semantic feature vector, including: arranging the multiple archive context semantic feature vectors in two dimensions into an archive context semantic feature matrix and then passing through an archive context feature enhancement module including multiple hybrid convolutional layers to obtain a second-scale archive semantic feature vector.

[0040] More specifically, in order to further extract and integrate the semantic information of the archival data, after arranging multiple archival context semantic feature vectors in two dimensions into an archival context semantic feature matrix, an archival context feature enhancement module including multiple hybrid convolutional layers is used to obtain a second-scale archival semantic feature vector, thereby obtaining a higher-level and more abstract feature representation. The archival context feature enhancement module is a convolutional neural network model including multiple hybrid convolutional layers. Convolutional neural networks (CNNs) can effectively extract local features when processing two-dimensional data. By arranging multiple archival context semantic feature vectors into a matrix, CNN can capture local features in the matrix using convolution operations, thereby better understanding the structure of archival data. The CNN model with multiple hybrid convolutional layers can extract features at different levels through convolution kernels of different depths and scales. This helps to integrate local features into higher-level abstract features, thereby more comprehensively representing the semantic information of archival data. The convolution operation in CNN has the characteristic of parameter sharing, which can reduce the number of parameters of the model and improve the generalization ability and efficiency of the model. Through the convolution operation, CNN can retain the spatial structure information of the input data, which helps to capture the spatial relationship and context information in the archival data. At the same time, the activation functions and deep structures in CNN can help the model learn nonlinear relationships, thereby better representing the complex semantic information of archival data.

[0041] Specifically, in the above-mentioned big data-based intelligent archive retrieval method, a hybrid feature extraction is performed on the multiple archive context semantic feature vectors to obtain a second-scale archive semantic feature vector, including: using each hybrid convolutional layer of the archive context feature enhancement module to perform the following on the input data in the forward pass of the layer: multi-scale convolution encoding on the input data to obtain a multi-scale convolutional feature map; pooling the multi-scale convolutional feature map to obtain a pooled feature map; and activating the pooled feature map to obtain an activated feature map; wherein the output of the last hybrid convolutional layer of the archive context feature enhancement module is the second-scale archive semantic feature vector.

[0042] Furthermore, in order to comprehensively utilize semantic information of different scales and levels, the first-scale archival semantic feature vector and the second-scale archival semantic feature vector are fused to obtain the archival database feature vector, thereby obtaining a more comprehensive, rich and representative archival database feature representation. The first-scale archival semantic feature vector and the second-scale archival semantic feature vector capture semantic information of different levels and scales respectively. By fusing these two feature vectors, multi-scale information can be comprehensively considered, making the obtained archival database feature vector more comprehensive and rich. The first-scale archival semantic feature vector and the second-scale archival semantic feature vector extract different types of features at different levels. By fusing these features, the expressive power of the features can be improved and the semantic information of the archival database can be better described. At the same time, fusing feature vectors of different scales can help suppress noise and redundant information and improve the robustness and discrimination of the features.

[0043] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the first-scale archive semantic feature vector and the second-scale archive semantic feature vector are fused to obtain an archive database feature vector, including: fusing the first-scale archive semantic feature vector and the second-scale archive semantic feature vector with the following cascade formula to obtain an archive database feature vector; wherein the cascade formula is:

[0044]

[0045] in, represents the first scale archive semantic feature vector, represents the second scale archive semantic feature vector, represents the feature vector of the archive database, Represents a cascade function.

[0046] In step S130, the archive search term is processed to obtain an archive search term feature vector. Converting the archive search term into a feature vector can represent it as a continuous, semantically rich vector, thereby better capturing the meaning and context of the search term. By representing the archive search term as a feature vector, the search term can be matched with the document in the archive database in vector space, thereby improving the accuracy of the search result. Specifically, the specific method of processing the archive search term refers to the specific method of processing the archive database.

[0047] In step S140, the archive search term feature vector and the archive database feature vector are comprehensively analyzed to generate an archive search result.

[0048] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the archive retrieval term feature vector and the archive database feature vector are comprehensively analyzed to generate the archive retrieval result, including: fusing the archive retrieval term feature vector and the archive database feature vector to obtain an archive retrieval feature matrix; performing class coherence compensation based on feature implicit expression on the archive retrieval feature matrix to obtain a compensated archive retrieval feature matrix; and, passing the compensated archive retrieval feature matrix through a generator to generate the archive retrieval result.

[0049] More specifically, in order to comprehensively utilize the semantic information of the search terms and the archive database, the archive search term feature vector and the archive database feature vector are fused to obtain the archive search feature matrix, so as to more effectively perform archive retrieval. The fusion of the search term feature vector and the archive database feature vector can comprehensively utilize the semantic information of the query and the document, making the retrieval more comprehensive and accurate. The fusion of the search term feature vector and the archive database feature vector can better capture the semantic relevance between the query and the document, and improve the matching accuracy. At the same time, the fusion of the feature vectors of the two can perform personalized retrieval according to specific query requirements and context information, and improve the relevance of the retrieval results and user satisfaction.

[0050] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the archive retrieval term feature vector and the archive database feature vector are fused to obtain an archive retrieval feature matrix, including: fusing the archive retrieval term feature vector and the archive database feature vector as follows to obtain an archive retrieval feature matrix; wherein the fusion formula is:

[0051]

[0052] in represents vector multiplication, represents the archive retrieval feature matrix, represents the archive retrieval term feature vector, represents the feature vector of the archive database, represents the transpose of the archive database feature vector.

[0053] In particular, in the technical solution of the present application, it is considered that the accuracy of archive retrieval depends largely on the semantic relevance between the search terms and the archive content. Class coherence compensation can strengthen the semantic connection between the archive search terms and the feature vectors of the archive database, thereby improving the relevance of the retrieval. Through class coherence compensation, the overlap between different archive feature vectors can be reduced, making the feature expression of each archive more unique and more discriminative, which helps to improve the accuracy of the retrieval results. In practical applications, the search terms may have spelling errors, synonyms, or subtle semantic differences. Class coherence compensation helps the model remain robust to these changes and provide accurate retrieval results even in the face of imperfect search terms. The archive database may contain a large number of features, some of which may not be important for the retrieval task. Class coherence compensation can help identify the most relevant features, thereby performing feature selection and dimensionality reduction, simplifying the model and improving computational efficiency. Through class coherence compensation, it is possible to better understand which archive features have a greater impact on the retrieval results. This interpretability is very useful for understanding model behavior and improving retrieval strategies. The content of the archives and the way the search terms are used may change over time. Class coherence compensation can help the model adapt to these changes and capture the dynamic characteristics of archive content through implicit feature expression. Based on this, in the technical solution of this application, class coherence compensation based on implicit feature expression is performed on the archive retrieval feature matrix.

[0054] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, the archive retrieval feature matrix is ​​subjected to class coherence compensation based on feature implicit expression to obtain a compensated archive retrieval feature matrix, including: using the principal component analysis method to process the archive retrieval feature matrix to map the archive retrieval feature matrix to a low-dimensional feature space to obtain a generated feature implicit expression; calculating the class coherence matrix of the generated feature implicit expression using the following class coherence calculation formula, wherein the class coherence calculation formula is:

[0055]

[0056] in, represents the implicit expression of the generated features, represents the mean of the implicit expression of the generated features, represents the transpose of a vector, represents the scale of the implicit expression of the generated features, representing the class coherence matrix; and, based on the class coherence matrix, performing class-related compensation on the generated feature implicit expression to obtain a compensated archive retrieval feature matrix.

[0057] Here, for the archive retrieval feature matrix, firstly, by projecting the data into a low-dimensional feature space, the computational efficiency of the model is improved, the algorithm complexity is reduced, and the calculation speed is accelerated. Then, with the goal of enhancing robustness, the implicit expression of the features is used to capture the intrinsic structure of the data and enhance the anti-interference ability of the model. Next, the overlap between samples of different categories is reduced through class coherence compensation, thereby improving the accuracy of the generation algorithm. Furthermore, in the feature selection stage, class coherence compensation helps identify the most relevant features, thereby optimizing the generation effect. At the same time, the implicit expression of features also provides insights for enhancing the interpretability of the generation model, making it easier for users to understand the data structure.

[0058] Specifically, in the above-mentioned intelligent archive retrieval method based on big data, based on the class coherence matrix, class-related compensation is performed on the generated feature implicit expression to obtain the compensated archive retrieval feature matrix, including: creating a Spring Web project; writing a class-related compensation algorithm in the Spring Web project, wherein the class-related compensation algorithm is to multiply the generated feature implicit expression by the class coherence matrix; creating a Spring Controller class to process HTTP requests and responses; writing a method in the Spring Controller class to accept the generated feature implicit expression and the class coherence matrix as input; and configuring routing to map HTTP requests to the SpringController class.

[0059] Specifically, the following effects can be achieved by using the Spring Web mechanism to deploy and implement the class-related compensation of the generated feature implicit expression based on the class-coherent matrix to obtain the compensated archive retrieval feature matrix: 1. Easy to develop and maintain: The Spring framework provides a large number of functions and components, making the development of Web applications easier and more efficient. Spring Boot is particularly suitable for quickly building Web applications, reducing the workload of developers. 2. RESTful API support: The Spring framework provides good support for RESTful APIs, and it is easy to create RESTful-style APIs to process data requests and responses. This is very useful for processing feature data and returning processing results. 3. Dependency injection: The dependency injection function of the Spring framework can help manage component dependencies in applications, making the code more modular and easy to maintain. 4. Integration testing: The Spring framework provides good support for integration testing, and test cases can be easily written and run to verify the functionality of the application. 5. Security: Spring Security is an important module of the Spring framework, which can help you implement authentication, authorization and other security functions to protect your Web applications from malicious attacks. 6. Extensibility: The Spring framework is a very flexible and extensible framework that can easily integrate other libraries and tools to meet different needs. In general, using the Spring Web mechanism makes it easier for you to develop, deploy, and maintain Web applications, while providing rich features and tools to support you in implementing feature compensation functions and providing a robust infrastructure to build your applications.

[0060] More specifically, based on the class coherence matrix, the code portion for performing class coherence compensation on the generated feature implicit expression to obtain the compensated archive retrieval feature matrix is ​​as follows:

[0061]

[0062] Then, the compensated archive retrieval feature matrix is ​​used to generate archive retrieval results through a generator. Generating retrieval results through a generator can increase the diversity of results, enrich user choices, and avoid overly similar results. For fuzzy or ambiguous queries, the generator can generate more targeted retrieval results based on contextual information and semantic associations. The archive retrieval results may include the following: 1. Document list: The retrieval results are presented in the form of a document list, listing documents or archives that match the query conditions. 2. Document summary: For each document, a summary or brief description is usually provided so that users can quickly understand the document content. 3. Relevance score: For each retrieval result, the system may give a relevance score indicating the degree of match between the result and the query condition. 4. Ranking information: The retrieval results are usually ranked according to relevance or other sorting criteria, and users can quickly find the most relevant documents based on the ranking information. 5. Link or preview: For online documents or archives, the retrieval results may contain a direct link to the location of the document, or provide a preview function to allow users to view part of the content without opening the document. 6. Label or classification information: The retrieval results may include the label or classification information of the document to help users better understand the document content and filter it. 7. Filtering options: The system provides various filtering options, allowing users to filter and sort search results according to different conditions (such as time, author, type, etc.).

[0063] Specifically, in the above-mentioned big data-based intelligent archive retrieval method, the model of the generator can be a generative adversarial network model or a variational autoencoder model, which trains the compensated archive retrieval feature matrix to map the compensated archive retrieval feature matrix to the representation of hot issues of public concern.

[0064] In summary, the intelligent archive retrieval method based on big data according to the embodiment of the present application has been explained, which uses big data technology and artificial intelligence technology based on the field of deep learning to extract and encode features of archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, which improves the comprehensiveness and accuracy of detection.

[0065] Based on the same inventive concept, the embodiment of the present application also provides an intelligent archive retrieval device based on big data, which can be used to implement the method described in the above embodiment, as described in the following embodiment. Since the principle of solving the problem by the intelligent archive retrieval device based on big data is similar to the above method, the implementation of the intelligent archive retrieval device based on big data can refer to the implementation of the above method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0066] Figure 4 FIG. 1 is a block diagram of an intelligent archive retrieval device based on big data according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the archive intelligent retrieval device 100 based on big data includes: an archive data acquisition module 110, which is used to acquire archive search terms and archive search databases; an archive search database analysis module 120, which is used to analyze the archive search database to obtain archive database feature vectors; an archive search term analysis module 130, which is used to process the archive search terms to obtain archive search term feature vectors; and a retrieval result generation module 140, which is used to comprehensively analyze the archive search term feature vectors and the archive database feature vectors to generate archive search results.

[0067] Here, those skilled in the art can understand that the specific functions and operations of the various units and modules in the above-mentioned intelligent archive retrieval device based on big data 100 have been referred to above. Figures 1 to 4 The description of the archival intelligent retrieval method based on big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0068] In summary, the intelligent archive retrieval device based on big data according to the embodiment of the present application has been explained, which uses big data technology and artificial intelligence technology based on the field of deep learning to extract and encode features of archive retrieval terms and archive retrieval databases to generate archive retrieval results. In this way, the required archive data can be intelligently retrieved, which improves the comprehensiveness and accuracy of detection.

[0069] The devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is an electronic device, and the electronic device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0070] In a typical example, the electronic device specifically includes one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the big data-based intelligent archive retrieval method.

[0071] Figure 5 FIG. 1 is a block diagram of an electronic device according to an embodiment of the present application. Figure 5 As shown, based on the above data processing method, this specification also provides an electronic device, one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent archive retrieval method based on big data. The processor can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor, a logic gate, a switch, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller and an embedded microcontroller, etc. The memory includes but is not limited to a random access memory (Random Access Memory, RAM), a read-only memory (Read-Only Memory, ROM), a cache (Cache), a hard disk (Hard Disk Drive, HDD) or a memory card (Memory Card). The electronic device can be a server of the method described in the above embodiment, etc. Of course, the electronic device can also include a module structure such as an input device and a transmission module. In this embodiment, the functions and effects specifically implemented by the electronic device can be explained in comparison with other embodiments, and will not be repeated here.

[0072] In addition, the embodiments of this specification also provide a computer-readable storage medium on which computer instructions are stored, and when the instructions are executed, the steps of the method described in any one or more of the above embodiments are implemented. The storage medium may include a physical device for storing information, which is usually to digitize the information and then store it in a medium using electrical, magnetic or optical methods. The storage medium may include: devices that store information using electrical energy, such as various types of memories, such as RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memories, bubble memories, and U disks; devices that store information optically, such as CDs or DVDs. Of course, there are other types of readable storage media, such as quantum memories, graphene memories, and so on.

[0073] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0074] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0075] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0076] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0077] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for intelligent archive retrieval based on big data, characterized in that: include: Obtain archival search terms and archival search databases; Analyzing the archive retrieval database to obtain an archive database feature vector; Processing the archive search term to obtain an archive search term feature vector; Comprehensively analyzing the archive search term feature vector and the archive database feature vector to generate an archive search result; The comprehensive analysis of the archive search term feature vector and the archive database feature vector to generate the archive search result includes: Merging the archive retrieval term feature vector and the archive database feature vector to obtain an archive retrieval feature matrix; Performing class coherence compensation based on implicit feature expression on the archive retrieval feature matrix to obtain a compensated archive retrieval feature matrix; Passing the compensated archive retrieval feature matrix through a generator to generate an archive retrieval result; The process of performing class coherence compensation based on implicit feature expression on the archive retrieval feature matrix to obtain a compensated archive retrieval feature matrix includes: Using principal component analysis to process the archive retrieval feature matrix to map the archive retrieval feature matrix to a low-dimensional feature space to obtain an implicit expression of generated features; The class coherence matrix of the implicit expression of the generated feature is calculated using the following class coherence calculation formula, wherein the class coherence calculation formula is: in, represents the implicit expression of the generated features, represents the mean of the implicit expression of the generated features, represents the transpose of a vector, represents the scale of the implicit expression of the generated features, represents the class coherence matrix; Based on the class coherence matrix, class-correlation compensation is performed on the generated feature implicit expression to obtain a compensated archive retrieval feature matrix; Wherein, based on the class coherence matrix, class-correlation compensation is performed on the generated feature implicit expression to obtain a compensated archive retrieval feature matrix, including: Create a Spring Web project; Writing a class-related compensation algorithm in the Spring Web project, wherein the class-related compensation algorithm is to multiply the generated feature implicit expression by the class-related matrix; Create a Spring Controller class to handle HTTP requests and responses; Writing a method in the Spring Controller class to accept the generated feature implicit expression and the class coherence matrix as input; Configure routing to map HTTP requests to the said Spring Controller classes.

2. The intelligent archive retrieval method based on big data according to claim 1 is characterized in that: Analyze the archive retrieval database to obtain the archive database feature vector, including: The archive retrieval database is passed through an archive database feature extraction module including an embedding layer to obtain a plurality of archive context semantic feature vectors; Multi-scale feature extraction is performed on the multiple archive context semantic feature vectors to obtain an archive database feature vector.

3. The intelligent archive retrieval method based on big data according to claim 2 is characterized in that: The archive database feature extraction module is a transformer-based context encoder model including an embedding layer.

4. The intelligent archive retrieval method based on big data according to claim 3 is characterized in that: Performing multi-scale feature extraction on the multiple archive context semantic feature vectors to obtain an archive database feature vector, including: Cascading the plurality of archive context semantic feature vectors to obtain a first scale archive semantic feature vector; Performing mixed feature extraction on the multiple archive context semantic feature vectors to obtain a second scale archive semantic feature vector; The first-scale archive semantic feature vector and the second-scale archive semantic feature vector are fused to obtain an archive database feature vector.

5. The intelligent archive retrieval method based on big data according to claim 4 is characterized in that: Performing mixed feature extraction on the multiple archive context semantic feature vectors to obtain a second-scale archive semantic feature vector, including: The plurality of archival context semantic feature vectors are arranged in two dimensions into an archival context semantic feature matrix and then passed through an archival context feature enhancement module comprising a plurality of mixed convolutional layers to obtain a second-scale archival semantic feature vector.

6. An intelligent archive retrieval device based on big data, characterized in that: include: The archive data acquisition module is used to obtain archive search terms and archive search databases; An archive retrieval database analysis module is used to analyze the archive retrieval database to obtain an archive database feature vector; An archive search term analysis module, used for processing the archive search term to obtain an archive search term feature vector; A search result generation module, used for comprehensively analyzing the archive search term feature vector and the archive database feature vector to generate an archive search result; The comprehensive analysis of the archive search term feature vector and the archive database feature vector to generate the archive search result includes: Merging the archive retrieval term feature vector and the archive database feature vector to obtain an archive retrieval feature matrix; Performing class coherence compensation based on implicit feature expression on the archive retrieval feature matrix to obtain a compensated archive retrieval feature matrix; Passing the compensated archive retrieval feature matrix through a generator to generate an archive retrieval result; The process of performing class coherence compensation based on implicit feature expression on the archive retrieval feature matrix to obtain a compensated archive retrieval feature matrix includes: Using principal component analysis to process the archive retrieval feature matrix to map the archive retrieval feature matrix to a low-dimensional feature space to obtain an implicit expression of generated features; The class coherence matrix of the implicit expression of the generated feature is calculated using the following class coherence calculation formula, wherein the class coherence calculation formula is: in, represents the implicit expression of the generated features, represents the mean of the implicit expression of the generated features, represents the transpose of a vector, represents the scale of the implicit expression of the generated features, represents the class coherence matrix; Based on the class coherence matrix, class-correlation compensation is performed on the generated feature implicit expression to obtain a compensated archive retrieval feature matrix; Wherein, based on the class coherence matrix, class-correlation compensation is performed on the generated feature implicit expression to obtain a compensated archive retrieval feature matrix, including: Create a Spring Web project; Writing a class-related compensation algorithm in the Spring Web project, wherein the class-related compensation algorithm is to multiply the generated feature implicit expression by the class-related matrix; Create a Spring Controller class to handle HTTP requests and responses; Writing a method in the Spring Controller class to accept the generated feature implicit expression and the class coherence matrix as input; Configure routing to map HTTP requests to the said Spring Controller classes.

7. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the big data-based intelligent archive retrieval method as described in any one of claims 1-5.

Images