Self-adaptive knowledge construction method, system and equipment based on representation learning and storage medium

Abstract

The invention discloses a self-adaptive knowledge construction method, system and device based on representation learning and a storage medium, and the method comprises the steps: collecting data covering different fields and different types, carrying out the preprocessing of the collected data, and forming a resource library; performing representation learning on the data in the resource library through the trained long-short-term memory network model to obtain low-dimensional vector representation of the data; based on low-dimensional vector representation of the data, knowledge of the data in the resource library is extracted in a structured form through a natural language processing method, the extracted knowledge is screened and optimized, and the screened and optimized knowledge data is integrated and classified based on a clustering algorithm; and storing, managing and updating the integrated and classified knowledge data. The method does not depend on annotated data, efficient knowledge construction can be carried out on complex multi-source heterogeneity, and meanwhile the method has the self-adaptive capacity for dynamic change data.

Description

Technical Field

[0001] The present invention relates to knowledge construction, and in particular to an adaptive knowledge construction method, system, device and storage medium based on representation learning. Background Art

[0002] Knowledge construction is the process of systematically organizing and integrating knowledge. It is not just a simple collection of information, but also emphasizes the extraction, association, and expansion of knowledge from a large amount of data and information to form a valuable knowledge system. For example, in the field of academic research, scholars construct knowledge by using a large number of research reports, experimental data and other information to form a complete theoretical framework. An accurate, comprehensive and easily accessible knowledge system can provide strong support for decision-making. Therefore, efficient construction and management of knowledge are essential. Traditional knowledge construction relies on manual implementation and can no longer meet the needs of knowledge construction in the context of a sharp increase in data volume. With the development of machine learning, knowledge construction methods based on statistical machine learning have emerged. These methods use a large amount of labeled data to train models to extract knowledge, such as using algorithms such as support vector machines to perform text classification to build a knowledge classification system. This type of method relies heavily on a large amount of high-quality labeled data. The cost of obtaining labeled data is extremely high, and the generalization ability of the model is limited. When the data distribution changes, the performance of the model will drop significantly, and the data adaptability is poor. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to provide an adaptive knowledge construction method, system, device and storage medium based on representation learning to solve one or more problems existing in the prior art.

[0004] Technical solution: The present invention provides an adaptive knowledge construction method based on representation learning, comprising: Collect data from different fields and types, pre-process the collected data, and form a resource library; Through the trained long short-term memory network model, the data in the resource library is represented and learned to obtain a low-dimensional vector representation of the data; Based on the low-dimensional vector representation of data, the knowledge of the data in the resource library is extracted in a structured form through natural language processing methods, the extracted knowledge is screened and optimized, and the screened and optimized knowledge data is integrated and classified based on the clustering algorithm; Store, manage and update integrated and classified knowledge data.

[0005] Furthermore, the different types of data include text, image and video data; and the preprocessing of the collected data includes: Remove noise, outliers, and incomplete information from data: For text data, outliers include meaningless characters and punctuation errors; for image and video data, incomplete information includes blurred and damaged parts; Normalization processing; Classify, organize and merge the normalized data to form a resource library.

[0006] Furthermore, the data in the resource library is represented and learned by the trained long short-term memory network model to obtain a low-dimensional vector representation of the data, including: When the trained LSTM model processes input data, the hidden layer As a low-dimensional vector representation of the data, it is expressed as: ; in, Represents the time step A low-dimensional vector representation of the input data at this time step Features and latent semantic information; hidden layer , is the output gate, is element-wise multiplication, is the hyperbolic tangent function, To update the cell status; As time steps progress, the hidden layer states at different time steps together constitute a dynamic low-dimensional vector representation sequence of the input data.

[0007] Furthermore, the low-dimensional vector representation based on the data extracts the knowledge of the data in the resource library in a structured form through a natural language processing method, including: The preprocessed data is divided into independent words, and the preprocessed text data is set to , the word sequence after word segmentation is , Expressing words, is the number of words; expressed as a function ; For image and video data, after preprocessing, extract the information and convert it into text form, and then perform the same operation; Using deep learning named entity recognition model to identify word sequences Perform named entity recognition. The output after processing by the deep learning named entity recognition model is an entity set. , is the number of entities, After vectorization, its features are represented. ; expressed as a function ; The identified entities are classified according to pre-set standards, and then the relationships between entities are extracted from the data based on syntactic analysis to achieve structured knowledge extraction.

[0008] Furthermore, the screening and optimization of the extracted knowledge includes: Based on predefined natural language grammar rules, the extracted knowledge is automatically detected and analyzed for grammatical errors; for detected grammatical errors, the knowledge entries that do not conform to the grammatical rules are automatically removed or corrected.

[0009] Furthermore, the clustering algorithm is used to integrate and classify the screened and optimized knowledge data, including: Random Selection data points as the initial cluster centers, and the set of these cluster centers is expressed as ; For each data point in the dataset , calculate it to each cluster center Distance , the formula is: ; in, Represents data points No. eigenvalues, Represents the cluster center No. eigenvalues, Represents the dimension of the data point; The data points Assign to the nearest cluster center, and repeat the assignment and update until the cluster center no longer changes significantly or reaches the preset number of iterations.

[0010] Furthermore, the storing, managing and updating of the integrated and classified knowledge data includes: Storage is done in the form of cloud database; Management includes version control and authority management of knowledge; Update knowledge regularly: Automatically monitor changes in data sources, review knowledge regularly, and automatically discover new knowledge and trends based on data mining algorithms and integrate them into the knowledge system; Use association rule mining algorithms to discover the associations between different knowledge. When new knowledge and trends are discovered, they are integrated into the knowledge system and the newly discovered knowledge is fed back to users.

[0011] Based on the same inventive concept, an adaptive knowledge construction system based on representation learning of the present invention comprises: The data collection module is used to collect data covering different fields and types, pre-process the collected data, and form a resource library; The representation learning module is used to perform representation learning on the data in the resource library through the trained long short-term memory network model to obtain a low-dimensional vector representation of the data; The knowledge extraction module is used to extract the knowledge of the data in the resource library in a structured form based on the low-dimensional vector representation of the data through natural language processing methods, screen and optimize the extracted knowledge, and integrate and classify the screened and optimized knowledge data based on the clustering algorithm; The knowledge management module is used to store, manage and update the integrated and classified knowledge data.

[0012] Based on the same inventive concept, an adaptive knowledge construction device based on representation learning of the present invention includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the adaptive knowledge construction method based on representation learning as described above.

[0013] Based on the same inventive concept, a computer-readable storage medium of the present invention stores a computer program thereon, and when the computer program is executed by a processor, the steps of the above-mentioned adaptive knowledge construction method based on representation learning are implemented.

[0014] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) The present invention captures the potential semantic information in the data through the representation learning method of the long short-term memory network (LSTM), and converts complex multi-source heterogeneous data covering different fields and types into low-dimensional vector representation, thereby reducing the dimension of the data, improving computing efficiency, and better retaining the key features and semantic relationships of the data during the knowledge extraction process, thus solving the problem that the existing knowledge construction methods are heavily dependent on a large amount of high-quality labeled data.

[0015] (2) Based on the low-dimensional vector representation of the data, combined with natural language processing methods, the efficiency of knowledge construction is improved, the extracted knowledge is screened, optimized and structured for storage, and the model performance is improved. The dynamic management and automatic updating of knowledge are also realized, ensuring the adaptive ability when processing constantly changing data, thus solving the problem of poor data adaptability of existing knowledge construction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of an adaptive knowledge construction method based on representation learning disclosed in an embodiment of the present invention; Figure 2It is a structural diagram of an adaptive knowledge construction system based on representation learning disclosed in an embodiment of the present invention; Figure 3 It is a structural diagram of an adaptive knowledge construction device based on representation learning disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It will be understood by those skilled in the art that the purposes and advantages that can be achieved with the present invention are not limited to the specific description of the above beneficial effects, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description.

[0018] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed in the present invention can be implemented in hardware, software or a combination of the two. Whether it is implemented in hardware or software depends on the specific application, design and conditions of the technical solution. Professional and technical personnel can use different methods to implement the functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0019] Reference to "embodiments" in the present invention means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0020] Example 1: Please refer to Figure 1 , Figure 1 1 is a flow chart of an adaptive knowledge construction method based on representation learning disclosed in an embodiment of the present invention. The adaptive knowledge construction method based on representation learning may include the following operations: S1. Collect data from different fields and types, pre-process the collected data, and form a resource library. Specifically: Data from different fields include data from business, finance, medical and other fields, and different types of data include text, image and video data.

[0021] Data preprocessing, including: 1) Remove noise, outliers, and incomplete information from the data to improve the accuracy and reliability of the data.

[0022] For text data, outliers include meaningless characters and punctuation errors; for image and video data, incomplete information includes blurred and damaged parts.

[0023] 2) Normalization processing: Normalization processing makes different types of data have similar scales and ranges, avoiding calculation problems caused by different data dimensions, which helps to improve the effectiveness of subsequent data analysis and model training.

[0024] 3) Classify, organize and merge the normalized data to form a resource library.

[0025] Data is classified based on the time sequence and usage of the data. For example, data can be classified by day, week, month, etc. For example, data can be classified by predicting sales data, analyzing user behavior, etc.

[0026] The classification method is based on a decision tree. By gradually testing the features of the data, the branches of the tree are constructed according to the values ​​of different features, and finally the data is divided into different categories. Each internal node represents a test of a feature, each branch is an output of the test, and the leaf node represents the category.

[0027] S2. Through the trained long short-term memory network model, the data in the resource library is represented and learned to obtain a low-dimensional vector representation of the data, capturing the potential semantic information in the data. Specifically: A long short-term memory network model is constructed, and the long short-term memory network model is trained, verified and adjusted to obtain an applicable long short-term memory network model. When the trained long short-term memory network model processes input data, the state of the hidden layer is a low-dimensional vector representation of the data, which contains the characteristics of the data and potential semantic information.

[0028] When the trained LSTM model processes input data, the hidden layer As a low-dimensional vector representation of the data, it is expressed as: ; in, Represents the time step A low-dimensional vector representation of the input data at this time step Features and latent semantic information; hidden layer , is the output gate, is element-wise multiplication, is the hyperbolic tangent function, To update the cell status.

[0029] As time steps progress, the hidden layer states at different time steps together constitute a dynamic low-dimensional vector representation sequence of the input data.

[0030] The present invention obtains low-dimensional vector representation of data through the representation learning method of long short-term memory network, which can greatly reduce the complexity of data. The original data may have high dimensions and a large amount of redundant information, while the low-dimensional vector concisely summarizes the key features of the data, which is convenient for subsequent calculations and analysis and improves the calculation efficiency. Long short-term memory network can effectively process sequence data. For text and time series data types, it can capture the long-term dependencies and semantic information therein, which enables the model to understand the context and meaning of the data, rather than just the surface features.

[0031] S3. Based on the low-dimensional vector representation of data, the knowledge of the data in the resource library is extracted in a structured form through natural language processing methods, the extracted knowledge is screened and optimized, and the screened and optimized knowledge data is integrated and classified based on the clustering algorithm. Specifically: S301. Based on the low-dimensional vector representation of the data, the knowledge of the data in the resource library is extracted in a structured form through natural language processing methods.

[0032] The preprocessed data is divided into independent words, and the preprocessed text data is set to , the word sequence after word segmentation is , Expressing words, is the number of words; expressed as a function . For image and video data, after preprocessing, information is extracted and converted into text form, and then the same operation is performed. For the method of converting images to text, image classification to text based on deep learning is an existing technology and will not be described in detail here. For the method of converting videos to text, image-text conversion and integration based on key frames, generating text descriptions for each key frame, is also an existing technology and will not be described in detail here.

[0033] Using deep learning named entity recognition model to identify word sequences Perform named entity recognition. The output after processing by the deep learning named entity recognition model is an entity set. , is the number of entities, After vectorization, its features are represented. ; expressed as a function ; Vectorization processing is an existing technology and will not be described in detail here.

[0034] The identified entities are classified according to pre-set criteria, and then the relationships between entities are extracted from the data based on syntactic analysis to achieve structured knowledge extraction. The pre-set criteria consider the degree of semantic association between entities to divide the categories. In text processing, for lexical entities, co-occurring and expressing similar topics can be classified into one category.

[0035] S302, screening and optimizing the extracted knowledge.

[0036] The present invention uses a rule-based method to screen and optimize the extracted knowledge. Specifically, according to the predefined natural language grammar rules, the extracted knowledge is automatically detected and analyzed for grammatical errors. For the detected grammatical errors, these knowledge items that do not conform to the grammatical rules are automatically eliminated or corrected, thereby ensuring that the knowledge items in the final knowledge base are grammatically correct and structurally standardized.

[0037] S303: Integrate and classify the screened and optimized knowledge data based on clustering algorithms.

[0038] Random Selection data points as the initial cluster centers, and the set of these cluster centers is expressed as .

[0039] For each data point in the dataset , calculate it to each cluster center Distance , the formula is: ; in, Represents data points No. eigenvalues, Represents the cluster center No. eigenvalues, Represents the dimensions of the data points.

[0040] The data points Assign to the nearest cluster center, and repeat the assignment and update until the cluster center no longer changes significantly or reaches the preset number of iterations.

[0041] The criteria for no longer having significant changes are: By setting a distance threshold, each time the cluster center is updated, the Euclidean distance indicator between the new cluster center and the cluster center of the previous round is calculated to measure the degree of change (Euclidean distance is an existing technology). If the distance between all cluster centers in the current round and the previous round is less than the set threshold, it is considered that the cluster center no longer changes significantly and the iteration is stopped.

[0042] The present invention divides the data into independent words, which helps to analyze the text content more carefully, accurately identify the meaning and function of each word, and avoid misunderstandings caused by improper word combination. At the same time, this also provides a basis for subsequent named entity recognition and syntactic analysis, and improves the accuracy of knowledge extraction. Based on the clustering algorithm, knowledge data is integrated, similar knowledge can be classified, which is convenient for knowledge organization and retrieval. The integrated knowledge is more orderly, and users can find the required knowledge more quickly, which improves the accessibility of knowledge.

[0043] S4. Store, manage and update the integrated and classified knowledge data to ensure the security and accessibility of knowledge. Specifically: Storage is done in the form of cloud database; Management includes version control and authority management of knowledge; Update knowledge regularly: By automatically monitoring changes in data sources, review knowledge regularly, and based on data mining algorithms, automatically discover new knowledge and trends and integrate them into the knowledge system.

[0044] Use association rule mining algorithms to discover the associations between different knowledge. When new knowledge and trends are discovered, they are integrated into the knowledge system and the newly discovered knowledge is fed back to users.

[0045] The present invention uses cloud database to store knowledge, ensuring the security and reliability of data. Cloud services usually provide powerful backup and recovery mechanisms to prevent data loss due to hardware failure, natural disasters or human errors. Cloud storage is highly scalable and can easily expand storage capacity according to the growth of knowledge without worrying about storage resource limitations. Automatic monitoring of data source changes and regular review of knowledge ensure the timeliness of knowledge, which can promptly reflect changes in the real world and keep knowledge up to date.

[0046] In summary, the present invention widely collects various types of data resources to provide rich materials for knowledge construction. Representation learning based on long short-term memory networks can capture the potential semantic information in the data and convert the data into a low-dimensional vector representation. This representation method not only reduces the dimension of the data and improves the computational efficiency, but also can better retain the key features and semantic relationships of the data. The integrated and classified knowledge data is stored, managed and updated to ensure the security of knowledge. By adopting advanced database management technology and security measures, the loss, damage and unauthorized access of knowledge can be prevented. When the data source changes or user needs change, it can automatically adapt to new situations.

[0047] Example 2: Please refer to Figure 2 , Figure 2: is a schematic diagram of a structure of an adaptive knowledge construction system based on representation learning disclosed in an embodiment of the present invention. The system can realize adaptive knowledge construction, specifically including: The data collection module is used to collect data covering different fields and types, pre-process the collected data, and form a resource library; The representation learning module is used to perform representation learning on the data in the resource library through the trained long short-term memory network model to obtain a low-dimensional vector representation of the data; The knowledge extraction module is used to extract the knowledge of the data in the resource library in a structured form based on the low-dimensional vector representation of the data through natural language processing methods, screen and optimize the extracted knowledge, and integrate and classify the screened and optimized knowledge data based on the clustering algorithm; The knowledge management module is used to store, manage and update the integrated and classified knowledge data.

[0048] Example 3: Please refer to Figure 3 , Figure 3 The present invention discloses a schematic diagram of a structure of an adaptive knowledge construction device based on representation learning disclosed in an embodiment of the present invention. The device may include a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the method described in the above embodiment and can achieve the technical effect consistent with the above method.

[0049] The memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the memory may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). A program / utility having a set (at least one) of program modules may be stored in, for example, the memory, such program modules including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment. The program modules typically perform the functions and / or methods in the embodiments described herein.

[0050] The processor executes various functional applications and data processing by running the programs stored in the memory, such as implementing the method provided in the embodiment of the present invention.

[0051] Embodiment 4: The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in the above embodiment are implemented and the technical effect consistent with the above method can be achieved.

[0052] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0053] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0054] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0055] Computer program code for performing the operation of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0056] Of course, the computer executable instructions of a storage medium including computer executable instructions provided by an embodiment of the present invention are not limited to the above method operations, and can also execute related operations in the method provided by any embodiment of the present invention.

[0057] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An adaptive knowledge construction method based on representation learning, characterized in that: include: Collect data from different fields and types, pre-process the collected data, and form a resource library; Through the trained long short-term memory network model, the data in the resource library is represented and learned to obtain a low-dimensional vector representation of the data; Based on the low-dimensional vector representation of data, the knowledge of the data in the resource library is extracted in a structured form through natural language processing methods, the extracted knowledge is screened and optimized, and the screened and optimized knowledge data is integrated and classified based on the clustering algorithm; Store, manage and update integrated and classified knowledge data.

2. The adaptive knowledge construction method based on representation learning according to claim 1, characterized in that: The different types of data include text, image and video data; the preprocessing of the collected data includes: Remove noise, outliers, and incomplete information from data: For text data, outliers include meaningless characters and punctuation errors; for image and video data, incomplete information includes blurred and damaged parts; Normalization processing; Classify, organize and merge the normalized data to form a resource library.

3. The adaptive knowledge construction method based on representation learning according to claim 1, characterized in that: The method of performing representation learning on the data in the resource library through the trained long short-term memory network model to obtain a low-dimensional vector representation of the data includes: When the trained LSTM model processes input data, the hidden layer As a low-dimensional vector representation of the data, it is expressed as: ； in, Represents the time step A low-dimensional vector representation of the input data at this time step Features and latent semantic information; hidden layer , is the output gate, is element-wise multiplication, is the hyperbolic tangent function, To update the cell status; As time steps progress, the hidden layer states at different time steps together constitute a dynamic low-dimensional vector representation sequence of the input data.

4. The adaptive knowledge construction method based on representation learning according to claim 1, characterized in that: The low-dimensional vector representation based on data extracts the knowledge of data in the resource library in a structured form through natural language processing methods, including: The preprocessed data is divided into independent words, and the preprocessed text data is set to , the word sequence after word segmentation is , Expressing words, is the number of words; expressed as a function ; For image and video data, after preprocessing, extract the information and convert it into text form, and then perform the same operation; Using deep learning named entity recognition model to identify word sequences Perform named entity recognition. The output after processing by the deep learning named entity recognition model is an entity set. , is the number of entities, After vectorization, its features are represented. ; expressed as a function ; The identified entities are classified according to pre-set standards, and then the relationships between entities are extracted from the data based on syntactic analysis to achieve structured knowledge extraction.

5. The adaptive knowledge construction method based on representation learning according to claim 4 is characterized in that: The screening and optimization of the extracted knowledge includes: Based on predefined natural language grammar rules, the extracted knowledge is automatically detected and analyzed for grammatical errors; for detected grammatical errors, the knowledge entries that do not conform to the grammatical rules are automatically removed or corrected.

6. The adaptive knowledge construction method based on representation learning according to claim 5, characterized in that: The clustering algorithm is used to integrate and classify the screened and optimized knowledge data, including: Random Selection data points as the initial cluster centers, and the set of these cluster centers is expressed as ; For each data point in the dataset , calculate it to each cluster center Distance , the formula is: ； in, Represents data points No. eigenvalues, Represents the cluster center No. eigenvalues, Represents the dimension of the data point; The data points Assign to the nearest cluster center, and repeat the assignment and update until the cluster center no longer changes significantly or reaches the preset number of iterations.

7. The adaptive knowledge construction method based on representation learning according to claim 1, characterized in that: The storing, managing and updating of the integrated and classified knowledge data includes: Storage is done in the form of cloud database; Management includes version control and authority management of knowledge; Update knowledge regularly: Automatically monitor changes in data sources, review knowledge regularly, and automatically discover new knowledge and trends based on data mining algorithms and integrate them into the knowledge system; Use association rule mining algorithms to discover the associations between different knowledge. When new knowledge and trends are discovered, they are integrated into the knowledge system and the newly discovered knowledge is fed back to users.

8. An adaptive knowledge construction system based on representation learning, characterized in that: include: The data collection module is used to collect data covering different fields and types, pre-process the collected data, and form a resource library; The representation learning module is used to perform representation learning on the data in the resource library through the trained long short-term memory network model to obtain a low-dimensional vector representation of the data; The knowledge extraction module is used to extract the knowledge of the data in the resource library in a structured form based on the low-dimensional vector representation of the data through natural language processing methods, screen and optimize the extracted knowledge, and integrate and classify the screened and optimized knowledge data based on the clustering algorithm; The knowledge management module is used to store, manage and update the integrated and classified knowledge data.

9. An adaptive knowledge construction device based on representation learning, characterized in that: The device comprises a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the adaptive knowledge construction method based on representation learning as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the adaptive knowledge construction method based on representation learning as described in any one of claims 1 to 7.

Images