Knowledge organization system construction method, device, equipment, medium and program

Abstract

The invention relates to the field of computers, and provides a knowledge organization system construction method and device, equipment, a medium and a program.The method comprises the steps that text paragraphs containing target domain terms are retrieved from a target domain data set obtained in advance, and corresponding retrieval results are input into a term generation model to obtain term generation results; for each term in the term generation result, inputting the term into a basic information prediction model and a semantic recognition model to obtain basic information and a semantic recognition result of the corresponding term; according to the terms and the basic information of the terms, utilizing a classification model to obtain corresponding classification results; and according to the term basic information, attributes and relationships corresponding to the terms and the term classification result, generating entries corresponding to the terms, and based on all generated entries, constructing a knowledge organization system. According to the method, multi-dimensional knowledge types are comprehensively considered, so that the expression capability of a knowledge organization system on complex knowledge is improved, and various types of knowledge can be better presented.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a knowledge organization system construction method, device, equipment, medium and program. Background Art

[0002] With the advent of the digital age, massive amounts of data and text information are growing explosively. Faced with complex, widely sourced, and diversely structured massive data, the Knowledge Organization System (KOS) is introduced to integrate and sort out these massive amounts of information, and to classify and index them according to certain logic and rules, so that the originally chaotic data becomes orderly, making it easier for users to quickly locate and find the required information. The Knowledge Organization System is a general term for semantic tools that express and organize human knowledge structures, including taxonomies, subject tables, thesauri, etc., as well as new knowledge organization models that have emerged with the development of semantic web technology, such as ontologies, linked data, visualized subject maps, knowledge maps, and knowledge graphs.

[0003] At present, the method used to build a knowledge organization system is mainly to collect data, extract knowledge from the collected data and build associations, and when the knowledge organization system needs to be updated, the updated data is reviewed and screened to determine whether to update the knowledge organization system.

[0004] However, the above methods focus on constructing the attributes of entities and the relationships between entities, ignoring the basic information, definition, classification and broader conceptual framework of entities, thus failing to fully understand the essential characteristics of entities and their position in the knowledge system, making the knowledge representation incomplete and affecting the comprehensiveness and accuracy of the system. In addition, when collecting a large amount of data in a certain field in the above methods, it is necessary to manually extract, classify, and associate these data, and as the information continues to expand, a large amount of manpower is required to update the knowledge organization system from time to time. Summary of the invention

[0005] The present invention provides a method, device, equipment, medium and program for constructing a knowledge organization system, which is used to solve the defect of the prior art that only focuses on the attributes of entities and the relationships between entities, resulting in incomplete knowledge representation. It comprehensively considers multi-dimensional knowledge types and improves the knowledge organization system's ability to represent complex knowledge, so that it can better process and present various types of knowledge.

[0006] The present invention provides a method for constructing a knowledge organization system, comprising: obtaining a target domain data set, the target domain data set comprising a plurality of text paragraphs obtained based on target domain files; retrieving text paragraphs containing target domain terms from the target domain data set, and inputting corresponding retrieval results into a term generation model to obtain term generation results output by the term generation model, the term generation results comprising at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training texts; for each term in the term generation results, inputting the term into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training texts; For each term in the term generation result, the term is input into the semantic recognition model to obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning the previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; for each term in the term generation result, the corresponding classification result is obtained by using the classification model according to the term and the basic information of the term; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; according to the basic information, attributes and relationships corresponding to each term and the classification result, the entries of the corresponding terms are generated, and based on all the generated entries, a knowledge organization system is constructed.

[0007] According to a knowledge organization system construction method provided by the present invention, text paragraphs containing target domain terms are retrieved from a target domain data set, and the corresponding search results are input into a term generation model to obtain term generation results output by the term generation model, including: using keywords previously set based on the target domain terms, keyword search or similarity matching is performed on the target domain data set to obtain text paragraphs containing the target domain terms as search results; based on the search results, in combination with preset prompt words, the term generation model is called to generate term generation results corresponding to the search results based on the preset prompt words; wherein the preset prompt words are set based on the context environment required for the terms of the target domain.

[0008] According to a knowledge organization system construction method provided by the present invention, after obtaining the term generation result output by the term generation model, the method includes: screening the terms in the term generation result to eliminate duplicate terms; and / or, based on preset verification rules, verifying the terms in the term generation result to eliminate terms that do not comply with the preset verification rules; wherein the preset verification rules include rules for limiting the terms to comply with the target field.

[0009] According to a knowledge organization system construction method provided by the present invention, before inputting terms into a semantic recognition model, the method includes: obtaining term training data and attribute labels and relationship labels corresponding to the term training data; converting the term training data and the attribute labels and relationship labels corresponding to the term training data according to a preset format to obtain instruction fine-tuning data after format conversion; wherein the preset format is used to limit the arrangement format of attributes and relationships corresponding to the entry; and fine-tuning the semantic recognition model according to the instruction fine-tuning data using a preset fine-tuning method.

[0010] According to a knowledge organization system construction method provided by the present invention, the classification model includes a primary classification number classifier, a secondary classification number classifier, a first-level classification number classifier and a second-level classification number classifier. The primary classification number classifier is obtained based on the training of term classification training samples and the primary classification number corresponding to the term classification training samples. The secondary classification number classifier is obtained based on the training of term classification training samples and the secondary classification number corresponding to the term classification training samples. The first-level classification number classifier is obtained by training the primary classification number classifier using the corresponding subordinate third-level classification number and its corresponding term classification training samples. The second-level classification number classifier is obtained by training the secondary classification number classifier using the corresponding subordinate third-level classification number and its corresponding term classification training samples. The primary classification number, the secondary classification number and the third classification number are obtained by dividing the term classification training samples based on a preset hierarchical division rule from high to low. For each term in the term generation result, according to the term and the basic information of the term, the classification model is used to obtain the corresponding classification result, including: for each term in the term generation result, each term is compared with The basic information corresponding to the terms is spliced ​​to obtain the text to be classified; the text to be classified is input into the first-level class number classifier and the second-level class number classifier respectively, and the first-level prediction classification result of the text to be classified output by the first-level class number classifier and the second-level prediction classification result of the text to be classified output by the second-level class number classifier are obtained, the first-level prediction classification result includes the first-level prediction classification number of the text to be classified, and the second-level prediction classification result includes the second-level prediction classification number of the text to be classified; determine whether the second-level prediction classification number of the same text to be classified belongs to the corresponding first-level prediction classification number, if the second-level prediction classification number of the same text to be classified belongs to the corresponding first-level prediction classification number, use the second third-level class number classifier to classify the text to be classified, and obtain the third-level prediction classification result of the text to be classified; otherwise, use the first third-level class number classifier to classify the text to be classified, and obtain the third-level prediction classification result of the text to be classified; according to the first-level prediction classification result of the text to be classified, the second-level prediction classification result of the text to be classified and the third-level prediction classification result of the text to be classified, the classification result of the corresponding term is obtained.

[0011] According to a knowledge organization system construction method provided by the present invention, a three-level class number classifier includes a fusion classifier and a base classifier. The three-level class number classifier is used to classify the text to be classified to obtain a three-level predicted classification number, including: inputting the text to be classified into each base classifier respectively to obtain the probability value output by each base classifier; outputting the probability value output by each base classifier to the fusion classifier for classification decision, and obtaining the three-level predicted classification result output by the fusion classifier.

[0012] The present invention also provides a knowledge organization system construction device, including: a data acquisition module, which acquires a target domain data set, the target domain data set includes multiple text paragraphs obtained based on target domain files; a term generation module, which retrieves text paragraphs containing target domain terms from the target domain data set, and inputs the corresponding retrieval results into a term generation model to obtain a term generation result output by the term generation model, the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; an information prediction module, which inputs the term into a basic information prediction model for each term in the term generation result, and obtains basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; and a semantic recognition module. A module, for each term in the term generation result, inputs the term into the semantic recognition model to obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning the previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; a term classification module, for each term in the term generation result, uses the classification model to obtain the corresponding classification result according to the term and the basic information of the term; wherein the classification model is obtained by fine-tuning the previously trained model based on the term training data and the classification labels corresponding to the term training data; a system construction module, generates entries for the corresponding terms according to the basic information, attributes and relationships corresponding to each term and the classification results, and constructs a knowledge organization system based on all the generated entries.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-described knowledge organization system construction methods is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for constructing a knowledge organization system as described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned knowledge organization system construction methods.

[0016] The knowledge organization system construction method, device, equipment, medium and program provided by the present invention generate corresponding terms for the target domain data set acquired in advance by using the term generation model, and further use multiple models such as the basic information prediction model, the semantic recognition model and the classification model to process the terms from different angles. Through the clear division of labor and mutual cooperation of each model, the basic information, attributes and relationships of the terms and the classification results are comprehensively and deeply excavated. On the basis of improving the quality and efficiency of term processing, multi-dimensional knowledge types are comprehensively considered so that different types of knowledge can be integrated into one entry, which helps to break down the barriers between knowledge and avoid the limitation of only understanding single-dimensional information. It helps to deeply understand and accurately apply the terms. When new knowledge or information appears, since the entry is constructed based on the information of each dimension, it only needs to update the information of the corresponding dimension, thereby enabling the knowledge organization system to more truly reflect the knowledge structure and relationship in the objective world, and improve the knowledge organization system's ability to represent complex knowledge, so that it can better process and present various types of knowledge. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is one of the flow charts of the knowledge organization system construction method provided by the present invention; Figure 2 This is the second flow chart of the knowledge organization system construction method provided by the present invention; Figure 3 It is a structural schematic diagram of the knowledge organization system construction device provided by the present invention; Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Figure 1 is a flow chart of the knowledge organization system construction method provided by the present invention, such as Figure 1 As shown, the method includes: S11, obtaining a target domain dataset, where the target domain dataset includes a plurality of text paragraphs obtained based on target domain files; S12, retrieving text paragraphs containing target domain terms from the target domain dataset, and inputting the retrieval results into a term generation model to obtain a term generation result output by the term generation model, wherein the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; S13, for each term in the term generation result, input the term into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; S14, for each term in the term generation result, input the term into a semantic recognition model to obtain a semantic recognition result of the corresponding term output by the semantic recognition model, the semantic recognition result including attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning a previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; S15, for each term in the term generation result, using a classification model according to the term and basic information of the term, obtaining a corresponding classification result; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; S16, generating entries of corresponding terms according to basic information, attributes and relationships of the terms corresponding to each term and the term classification results, and constructing a knowledge organization system based on all generated entries.

[0021] It should be noted that the step numbers "S1N" in this specification do not represent the order of the knowledge organization system construction method. Figure 2 The knowledge organization system construction method of the present invention is described.

[0022] Step S11, obtaining a target domain dataset, where the target domain dataset includes a plurality of text paragraphs obtained based on target domain files.

[0023] In this embodiment, obtaining a target domain data set includes: collecting files in the target domain; performing text extraction on the collected files to obtain text content corresponding to each file; performing denoising and segmentation processing on each text content to obtain corresponding text paragraphs; and removing duplicates from all obtained text paragraphs to obtain a target domain data set.

[0024] It should be noted that the files may include documents, reports and professional books in the target field, such as those obtained through crawler technology, application programming interface (API) or other data collection tools. The corresponding files can be collected according to the actual design requirements, and the denoising method can be set according to the noise that may be introduced by the actual text extraction, such as deleting redundant Hypertext Markup Language (HTML) tags, Uniform Resource Locator (URL) and other denoising methods, which are not further limited here. In addition, the collected files are unstructured data.

[0025] Further, each text content is segmented to obtain corresponding text paragraphs, including: reading the text content and identifying paragraphs in the text according to punctuation marks and format information, or segmenting according to content theme or semantics. It should be added that segmentation according to content theme or semantics can use some existing natural language processing (NLP) tools, which will not be further elaborated here.

[0026] In an optional embodiment, after obtaining the target domain data set, it also includes: monitoring the corresponding file source, so as to obtain the corresponding update file after the file is updated, and after processing it into a data set in the above manner, obtaining the corresponding updated entry according to the following steps S12-S15, so as to update the knowledge organization system with the updated entry, which will not be repeated here.

[0027] It should be noted that using updated terms to update a previously created knowledge organization system includes: retrieving terms within the knowledge organization system based on the terms corresponding to the updated terms, and obtaining system terms in the knowledge organization system that match the terms of the updated terms; comparing the updated terms with the system terms, and if they are consistent, not updating the knowledge organization system; otherwise, replacing the system terms in the knowledge organization system with the updated terms to update the knowledge organization system.

[0028] In addition, the target domain dataset may also be a dataset within a previously disclosed target domain and containing terms in the domain, which is not further limited here.

[0029] Step S12, retrieve text paragraphs containing target domain terms from the target domain dataset, and input the corresponding search results into the term generation model to obtain the term generation results output by the term generation model, and the term generation results include at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text. It should be noted that the term generation results can be in the form of a list, a document, or other representations, which are specifically determined according to the output of the selected and trained term generation model, and are not further limited here.

[0030] In this embodiment, text paragraphs containing target domain terms are retrieved from the target domain data set, and the corresponding search results are input into the term generation model to obtain term generation results output by the term generation model, including: using keywords previously set based on the target domain terms, keyword search or similarity matching is performed on the target domain data set to obtain text paragraphs containing target domain terms as search results; based on the search results, in combination with preset prompt words, the term generation model is called to generate term generation results corresponding to the search results based on the preset prompt words; wherein the preset prompt words are set based on the context environment required for the terms of the target domain.

[0031] It is worth noting that if a keyword search is performed on the target domain data set using keywords previously set based on the target domain, then before the keyword search is performed on the target domain data set using keywords previously set based on the target domain, a retrieval system is constructed on the target domain data set using a preset retrieval technology to facilitate the subsequent keyword-based retrieval of relevant terminology texts in the target domain data set. It should be added that the preset retrieval technology can adopt existing methods such as inverted indexes, which can be selected based on actual retrieval requirements and are not further limited here. In addition, when similarity matching is performed on the target domain data set using keywords previously set based on the target domain, a similarity matching algorithm can adopt a cosine similarity algorithm, which can be selected based on actual design requirements and are not further limited here.

[0032] In an optional embodiment, the term generation model can be selected based on actual design requirements or prior experience, for example, it can be a general language model of thousands of questions. The preset prompt words can be designed based on actual design requirements or prior experience, such as using prompt words to construct a COSTAR framework, etc., which is not further limited here.

[0033] It should be added that when using the COSTAR framework to write preset prompt words, it is necessary to write them according to the context required by the terms in the target field. The context includes the field background and the related concepts involved in the field, so as to write prompt words that can guide the model to generate relevant terms according to the context. For example, the preset prompt words can be: [Background description] Please generate a list of terms related to the XX field based on the above background, and explain it in a calm and objective tone of a scholar. The audience is knowledge engineers, and please use the lightweight data exchange format JSON format.

[0034] In an optional embodiment, after obtaining the term generation results output by the term generation model, it includes: screening the terms in the term generation results to eliminate duplicate terms; and / or, based on preset verification rules, verifying the terms in the term generation results to eliminate terms that do not comply with the preset verification rules; wherein the preset verification rules include rules for limiting the terms to comply with the target field.

[0035] Specifically, based on preset verification rules, the terms in the term generation result are verified to eliminate terms that do not meet the preset verification rules, including: based on the preset verification rules, each term in the term generation result is matched with the terms in the preset term library, and based on the matching consistency, it is determined that the corresponding term in the term generation result has passed the verification, otherwise, it fails the verification and the corresponding term in the term generation result is eliminated; wherein, the preset term library is constructed based on expert experience or prior experience using the terms of the target field.

[0036] Furthermore, after determining that the corresponding term in the term generation result has not passed the verification, it also includes: pushing the term that has not passed the verification to the corresponding user terminal for review to determine whether it conforms to the corresponding target field, and receiving the review result returned by the user terminal; based on the review result that the term is consistent with the target field, retaining the corresponding term, and using the term to update the corresponding preset term library; based on the review result that the term is not consistent with the target field, removing the corresponding term from the term generation result.

[0037] Step S13, for each term in the term generation result, the term is input into the basic information prediction model to obtain the basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text.

[0038] It should be added that when using the basic information prediction model to predict the basic information of the corresponding term, it is also necessary to define the content of the basic information, such as the Chinese word form, pinyin and English translation of the term, etc., select an entry containing the above information from the professional books in the target field as an example for model learning, configure the corresponding prompt sample, and input the obtained prompt sample into the basic information prediction model so that the model can generate the basic information of the corresponding term based on the pattern learned from the prompt sample. For example, when the target field is the new energy vehicle field, the basic information related to the entry "biofuel" in the Chinese Science and Technology Word System (New Energy Vehicle Volume) published in 2012 can be selected as an example for model learning.

[0039] In an optional embodiment, a single-shot learning 1-shot technology may be used to generate basic information of the corresponding term.

[0040] Step S14, for each term in the term generation result, the term is input into the semantic recognition model to obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning a previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data.

[0041] In an optional embodiment, before inputting the term into the semantic recognition model, the process includes: downloading the original model file of the semantic recognition model and deploying it to a local server. It should be added that the semantic recognition model can be selected based on actual recognition requirements or prior experience, such as selecting the open source large language model ChatGLM, which is not further limited here.

[0042] In an optional embodiment, before inputting the terms into the semantic recognition model, it also includes: obtaining terminology training data and attribute labels and relationship labels corresponding to the terminology training data; converting the terminology training data and the attribute labels and relationship labels corresponding to the terminology training data according to a preset format to obtain instruction fine-tuning data; wherein the preset format is used to limit the arrangement format of attributes and relationships corresponding to the entry; and fine-tuning the semantic recognition model according to the instruction fine-tuning data using a preset fine-tuning method.

[0043] It should be added that the preset fine-tuning method can be selected according to actual design requirements, such as selecting the prompt fine-tuning (P-Tuning) method, and fine-tuning the original model file of the semantic recognition model previously deployed to the local server through P-Tuning to obtain the semantic recognition model. In addition, according to the instruction fine-tuning data, the semantic recognition model is fine-tuned using the preset fine-tuning method, including: using the preset fine-tuning method to replace the traditional discrete text prompt words with learnable continuous embeddings to optimize only the prompt-related parameters (such as the embedding layer and the encoder), freeze the main parameters of the pre-trained model, and then use the instruction fine-tuning data to fine-tune the model to significantly reduce the amount of training; and, during the fine-tuning process, introduce an encoder composed of a long short-term memory network (Long Short-Term Memory, referred to as LSTM) or a multilayer perceptron (Multilayer Perceptron, referred to as MLP) to extract features from virtual tags, thereby enhancing the relevance of features and accelerating training.

[0044] It should be noted that by using trainable LSTM or MLP to dynamically generate virtual tag embeddings, different embeddings can be generated according to different input data to better understand and adapt to complex, sequentially dependent tasks, providing higher flexibility and adaptability, and through the LSTM loop structure, it can better process and understand long-term dependencies and complex contextual information, which is suitable for tasks that require fine control and understanding of complex contexts. In addition, in P-Tuning, the parameters of the LSTM model can be shared between multiple tasks, thereby improving the generalization ability of the fine-tuned semantic recognition model and reducing the training requirements for each individual task.

[0045] Furthermore, the acquired terminology training data and its corresponding attributes and relationships can be obtained from corresponding professional books according to the actual target field involved. For example, when the target field is the new energy vehicle field, the Chinese Science and Technology Word System (New Energy Vehicle Volume) published in 2012 can be selected to include both attributes and relationships, totaling 510 entries, to obtain the terminology training data and the attribute labels and relationship labels corresponding to the terminology training data. No further limitation is made here.

[0046] In addition, attributes are used to represent the characteristics of terms used to describe, distinguish or classify, and relationships are used to represent the connection, dependency or interaction between the corresponding terms and other terms. Relationships include hierarchical relationships and composition relationships. Hierarchical relationships are used to represent the position of terms in the classification, such as the class or group to which the term belongs. Composition relationships are used to represent which larger entities or collections the term belongs to as a component. The preset format can be configured according to the actual arrangement requirements of the attributes and relationships involved. For example, it can be: { "instruction": "You are an expert in the field of new energy vehicles and have a deep understanding of the terms related to the field of new energy vehicles. Based on the given term, please give the attributes and relationships of the term. Please keep the complete format.\nTerm: Methane\nRelated attributes and relationships:", "input": "", "output": "

Entry attributes

Features

Key indicators

Disadvantages

Features

Features

Physical properties

Scale

Boiling point

Density

Flash point

Melting point

Flash point

Form

[0047] It should be added that the operating environment of the P-Tuning fine-tuning task can be configured based on considerations or prior experience such as computing requirements, resource availability, model complexity, and the size of the training dataset. For example, the operating environment of a single GPU V100 32GB card is selected, and no further limitation is made here.

[0048] Step S15, for each term in the term generation result, a corresponding classification result is obtained using a classification model according to the term and its basic information; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data.

[0049] In this embodiment, the classification model includes a first-level class number classifier, a second-level class number classifier, a first-level class number classifier and a second-level class number classifier. The first-level class number classifier is obtained based on the training of term classification training samples and the first-level classification number corresponding to the term classification training samples. The second-level class number classifier is obtained based on the training of term classification training samples and the second-level classification number corresponding to the term classification training samples. The first-level class number classifier is obtained by training the first-level class number classifier using the corresponding subordinate third-level classification number and its corresponding term classification training samples. The second-level class number classifier is obtained by training the second-level class number classifier using the corresponding subordinate third-level classification number and its corresponding term classification training samples. The first-level classification number, the second-level classification number and the third-level classification number are obtained by dividing the term classification training samples based on a preset hierarchical division rule from high to low. The term classification training samples include terms and basic information corresponding to the terms.

[0050] It should be noted that the number of third-level classification number classifiers should be configured as much as possible according to the types of second-level classification numbers involved, so as to ensure that regardless of whether the second-level prediction classification number of the same term belongs to the corresponding first-level prediction classification number, there is a corresponding third-level classification number classifier to classify the term.

[0051] In addition, the preset hierarchical division rules can be configured according to the Chinese Library Classification (CLC) or other classification requirements. The preset hierarchical division rules include at least three division levels and classification rules within each level. Taking three levels as an example, they include the first level, the second level and the third level from high to low. The first level divides the terms in the target field according to the subject field to obtain the first classification number. The second level can further divide the terms belonging to the same first classification number according to the professional field or the specific technical field on the basis of the first level to obtain the second classification number. The third level can further divide the terms belonging to the same second classification number according to the specific research topics, technical points or applications on the basis of the secondary classification number to obtain the third classification number.

[0052] Accordingly, for each term in the term generation result, the corresponding classification result is obtained according to the term and the basic information of the term and using the classification model, including: for each term in the term generation result, each term and the basic information corresponding to the term are spliced ​​to obtain the text to be classified; the text to be classified is input into the first-level class number classifier and the second-level class number classifier respectively, to obtain the first-level prediction classification result of the text to be classified output by the first-level class number classifier and the second-level prediction classification result of the text to be classified output by the second-level class number classifier, the first-level prediction classification result includes the first-level prediction classification number of the text to be classified, and the second-level prediction classification result includes the Secondary prediction classification number; determine whether the secondary prediction classification number of the same text to be classified belongs to the corresponding primary prediction classification number. If the secondary prediction classification number of the same text to be classified belongs to the corresponding primary prediction classification number, use the second third-level classification number classifier to classify the text to be classified to obtain the third-level prediction classification result of the corresponding text to be classified; otherwise, use the first third-level classification number classifier to classify the text to be classified to obtain the third-level prediction classification result of the corresponding text to be classified; obtain the classification result of the corresponding term according to the first-level prediction classification result of the corresponding text to be classified, the secondary prediction classification result of the corresponding text to be classified and the third-level prediction classification result of the corresponding text to be classified.

[0053] It should be supplemented that, for each text to be classified, when the same text to be classified is input into the first-level class number classifier and the second-level class number classifier respectively, a single text to be classified can be input into the first-level class number classifier and the second-level class number classifier respectively for classification; or, the corresponding text to be classified obtained according to at least two terms in the term generation result is input into the first-level class number classifier, and the text to be classified input into the first-level class number classifier is input into the second-level class number classifier at the same time. Accordingly, the first-level prediction classification result includes the first-level classification number of each corresponding input text to be classified, and the second-level prediction classification result includes the second-level classification number of each corresponding input text to be classified.

[0054] Specifically, the three-level class number classifier includes a fusion classifier and a base classifier. The three-level class number classifier is used to classify the text to be classified to obtain a three-level predicted classification number, including: inputting the text to be classified into each base classifier respectively to obtain the probability value output by each base classifier; outputting the probability value output by each base classifier to the fusion classifier for classification decision, and obtaining the three-level predicted classification result output by the fusion classifier.

[0055] It should be added that the base classifier can be set according to actual design requirements. For example, the three-level class number classifier can integrate 10 base binary classifiers and 1 fusion classifier for each low-level category using ensemble learning stacking. No further limitation is made here.

[0056] In addition, if there is only one text to be classified input into the first-level class number classifier and the second-level class number classifier, the first-level class number classifier performs vector conversion on the corresponding input text to be classified and classifies the converted vector features. The second-level class number classifier is the same as the first-level class number classifier and is not further limited here.

[0057] If the texts to be classified input into the primary class number classifier and the secondary class number classifier include at least two, the primary class number classifier and the secondary class number classifier can be implemented using the fast text classification algorithm FastText. Specifically, the primary class number classifier includes a word embedding layer, a summing pooling layer and a classification layer. The texts to be classified are input into the primary class number classifier to obtain the primary predicted classification number output by the primary class number classifier, including: inputting at least two texts to be classified into the word embedding layer for vector conversion to obtain corresponding vector features; inputting the corresponding vector features into the summing pooling layer for summing to obtain text vector features; inputting the text vector features into the classification layer for classification mapping to obtain the corresponding primary predicted classification number. Similarly, the method of using the secondary class number classifier to classify the texts to be classified can refer to the method of using the first class number classifier for classification in the previous text, and will not be repeated here.

[0058] It should be noted that in steps S13-S15, reference Figure 2 For each term in the term generation result, the term is input into the basic information prediction model, the term is input into the semantic recognition model, and the corresponding classification result is obtained by using the classification model according to the term and the basic information of the term. It can be as follows: by traversing each term in the term generation result, and inputting the traversed terms into the basic information prediction model and the terms into the semantic recognition model respectively, and after obtaining the basic information output by the basic information prediction model, the basic information is spliced ​​with the corresponding traversed terms and input into the classification model for classification.

[0059] Correspondingly, after generating the entry of the corresponding term, it also includes: judging whether the traversal of the term generation result is completed, if the traversal is completed, then ending the traversal; otherwise, continuing to traverse the term generation result to generate the entry in the above manner. It should be noted that the knowledge organization system can be constructed after the corresponding entry is first obtained, and after the entry is obtained subsequently, the knowledge organization system can be updated using the subsequently generated entry, or the knowledge organization system can be created for all entries after all entries are generated. The specific generation of entries and the creation of the knowledge organization system can refer to the following description, which will not be repeated here.

[0060] Step S16, generating entries of corresponding terms according to basic information, attributes and relationships corresponding to each term and classification results, and constructing a knowledge organization system based on all generated entries.

[0061] In an optional embodiment, before generating entries for corresponding terms according to the basic information, attributes and relationships corresponding to each term and the classification results, the following steps are performed: for each term in the term generation results, automatically accessing web pages and extracting definitions and annotations of the terms by writing scripts or using crawler tools; and / or, according to each term in the term generation results, querying a pre-built electronic dictionary or database, which may include professional dictionaries such as WordNet, CNKI (China National Knowledge Infrastructure), etc., which are based on cognitive linguistics; and / or, sending a query request to a pre-selected search website through an API or a web page interface, and extracting definitions and annotations of the terms from the returned results. The search website may select an existing search website, which is not further limited here.

[0062] In an optional embodiment, the formalized concept description includes terms, definitions and annotations, attributes and relationships. Before generating entries for the corresponding terms based on the basic information, attributes and relationships, and classification results corresponding to each term, it also includes: generating a formalized concept description for the corresponding term according to a preset structured form based on the basic information, attributes and relationships, classification results, definitions and annotations corresponding to the term.

[0063] Furthermore, the basic information, attributes and relationships, classification results, definitions and annotations corresponding to the terms are generated in a preset structured form to generate a formalized concept description of the corresponding terms, including: creating a template according to the terms, definitions and annotations, attributes and relationships in a preset structured form; filling the basic information, attributes and relationships, classification results, definitions and annotations corresponding to the terms into the template to obtain the corresponding formalized concept description.

[0064] It should be added that the formal concept description is used to characterize the form of representing the corresponding term in a preset structured form. Furthermore, the formal concept description also includes examples, which represent one or more examples used in an actual context. Before generating the entry of the corresponding term based on the basic information, attributes and relationships corresponding to each term and the classification results, it also includes: automatically accessing web pages and extracting text containing the target term by writing scripts or using crawler tools; using regular expressions or other text processing techniques to identify sentences containing the target term; screening sentences containing the target term; and / or, querying a pre-built electronic dictionary or database based on each term in the term generation result; and / or, sending a query request to a pre-selected search website through an API or web page interface, and extracting examples of the term from the returned results.

[0065] Accordingly, a formal concept description of the corresponding term is generated, including: generating a formal concept description of the corresponding term according to a preset structured form based on the basic information, attributes and relationships, classification results, definitions and annotations, and examples corresponding to the term. The specific implementation method can refer to the method of generating a formal concept description by creating a template above. When creating a template, it is also necessary to create it in combination with an example, which will not be repeated here.

[0066] In an optional embodiment, after the corresponding entries are generated, the process includes: verifying the accuracy of the entries and removing the entries that fail the verification.

[0067] Specifically, the accuracy of the entries is verified, including: using natural language processing (NLP) tools to check whether the grammar and spelling of the entries are correct, and verifying that they are correct; and / or, based on a previously constructed reference data set, using a text similarity algorithm to evaluate the consistency of the entry with the corresponding entry in the reference data set, and verifying that they are consistent; wherein the reference data set is a previously acquired professional dictionary, professional knowledge base, or an entry data set constructed based on prior experience or expert experience; and / or, using NLP technology to analyze examples to evaluate the correctness of the examples, or to check whether the source of the examples is reliable.

[0068] It should be added that the natural language processing tool may be a text proofreading tool such as LanguageTool or a grammar checking tool such as Grammarly API, and the text similarity algorithm may be a cosine similarity, edit distance, etc. The specific method may be selected based on actual verification requirements and is not further limited here.

[0069] In this embodiment, a knowledge organization system is constructed based on all generated entries, including: extracting terms from the entries and determining synonyms, near-synonyms and related words of the terms; determining the hierarchical relationship between the terms based on the classification results; organizing the extracted terms, synonyms, near-synonyms, related words of the terms and the hierarchical relationship between the terms into a structured thesaurus; basic information, attributes and relationships, classification results, definitions and annotations, and formalized concept descriptions are formatted using preset elements and preset coding rules to create metadata for each entry; and obtaining a knowledge organization system based on the thesaurus of the entry and the metadata of the entry.

[0070] In summary, the embodiment of the present invention generates corresponding terms for the target domain data set acquired in advance by using a term generation model, and further uses multiple models such as a basic information prediction model, a semantic recognition model and a classification model to process the terms from different angles. Through the clear division of labor and mutual cooperation of each model, the basic information, attributes and relationships of the terms and the classification results are comprehensively and deeply mined. On the basis of improving the quality and efficiency of term processing, multi-dimensional knowledge types are comprehensively considered so that different types of knowledge can be integrated into one entry, which helps to break down the barriers between knowledge and avoid the limitations of only understanding single-dimensional information. It helps to deeply understand and accurately apply the terms. When new knowledge or information appears, since the entry is constructed based on information of each dimension, it only needs to update the information of the corresponding dimension, thereby enabling the knowledge organization system to more truly reflect the knowledge structure and relationship in the objective world, and improve the knowledge organization system's ability to represent complex knowledge, so that it can better process and present various types of knowledge.

[0071] The knowledge organization system construction device provided by the present invention is described below. The knowledge organization system construction device described below and the knowledge organization system construction method described above can be referenced to each other.

[0072] Figure 3 A schematic diagram of the structure of a knowledge organization system construction device is shown, the device comprising: A data acquisition module 31 acquires a target domain data set, where the target domain data set includes a plurality of text paragraphs obtained based on target domain files; The term generation module 32 retrieves text paragraphs containing target domain terms from the target domain data set, and inputs the corresponding search results into the term generation model to obtain a term generation result output by the term generation model, wherein the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; The information prediction module 33 inputs each term in the term generation result into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; The semantic recognition module 34 inputs each term in the term generation result into a semantic recognition model to obtain a semantic recognition result of the corresponding term output by the semantic recognition model, wherein the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning a previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; The term classification module 35 obtains corresponding classification results for each term in the term generation result by using a classification model according to the term and the basic information of the term; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; The system construction module 36 generates entries of corresponding terms according to the basic information, attributes and relationships corresponding to each term and the classification results, and constructs a knowledge organization system based on all the generated entries.

[0073] It should be noted that in the embodiment of the present invention, the relevant functional modules can be implemented by a hardware processor, and the specific principles are the same as the principles of the method described in the above embodiment, and the more detailed explanations are not repeated here.

[0074] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communications interface 420 and the memory 430 communicate with each other via the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the knowledge organization system construction method, which includes: obtaining a target domain data set, the target domain data set including a plurality of text paragraphs obtained based on target domain files; retrieving text paragraphs containing target domain terms from the target domain data set, and inputting the corresponding retrieval results into a term generation model to obtain a term generation result output by the term generation model, the term generation result including at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; for each term in the term generation result, inputting the term into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is based on the unsupervised term training text The method is to perform unsupervised training on each term in the term generation result, input the term into the semantic recognition model, and obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning the previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; for each term in the term generation result, the corresponding classification result is obtained by using the classification model according to the term and the basic information of the term; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; according to the basic information, attributes and relationships corresponding to each term and the classification result, the entries of the corresponding terms are generated, and based on all the generated entries, a knowledge organization system is constructed.

[0075] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0076] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the knowledge organization system construction method provided by the above methods, which method includes: obtaining a target domain data set, the target domain data set includes multiple text paragraphs obtained based on target domain files; retrieving text paragraphs containing target domain terms from the target domain data set, and inputting the corresponding retrieval results into a term generation model to obtain a term generation result output by the term generation model, and the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; for each term in the term generation result, input the term into a basic information prediction model to obtain a term output by the basic information prediction model. The basic information of the corresponding term; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; for each term in the term generation result, the term is input into the semantic recognition model to obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning the previously trained model based on the term training data and the attribute labels and relationship labels corresponding to the term training data; for each term in the term generation result, the corresponding classification result is obtained by using the classification model according to the term and the basic information of the term; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; according to the basic information, attributes and relationships corresponding to each term and the classification result, the entry of the corresponding term is generated, and based on all the generated entries, a knowledge organization system is constructed.

[0077] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the knowledge organization system construction method provided by the above-mentioned methods, the method comprising: obtaining a target domain data set, the target domain data set comprising a plurality of text paragraphs obtained based on target domain files; retrieving text paragraphs containing target domain terms from the target domain data set, and inputting the corresponding retrieval results into a term generation model to obtain a term generation result output by the term generation model, the term generation result comprising at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; for each term in the term generation result, inputting the term into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is used to predict the basic information of the term. This information prediction model is obtained by unsupervised training based on unsupervised terminology training texts; for each term in the terminology generation result, the term is input into the semantic recognition model to obtain the semantic recognition result of the corresponding term output by the semantic recognition model, and the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning the previously trained model based on the terminology training data and the attribute labels and relationship labels corresponding to the terminology training data; for each term in the terminology generation result, the classification model is used according to the term and the basic information of the term to obtain the corresponding classification result; wherein the classification model is trained based on the terminology training data and the classification labels corresponding to the terminology training data; according to the basic information, attributes and relationships corresponding to each term and the classification result, the entries of the corresponding terms are generated, and based on all the generated entries, a knowledge organization system is constructed.

[0078] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0079] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a knowledge organization system, characterized in that: include: Acquire a target domain dataset, wherein the target domain dataset includes a plurality of text paragraphs obtained based on target domain files; Retrieving text paragraphs containing target domain terms from the target domain dataset, and inputting corresponding retrieval results into a term generation model to obtain a term generation result output by the term generation model, wherein the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; For each term in the term generation result, the term is input into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; For each term in the term generation result, the term is input into a semantic recognition model to obtain a semantic recognition result of the corresponding term output by the semantic recognition model, wherein the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning a previously trained model based on term training data and attribute labels and relationship labels corresponding to the term training data; For each term in the term generation result, a corresponding classification result is obtained by using a classification model according to the term and basic information of the term; wherein the classification model is trained based on the term training data and the classification labels corresponding to the term training data; According to the basic information, attributes and relationships corresponding to each of the terms and the classification results, entries of the corresponding terms are generated, and a knowledge organization system is constructed based on all the generated entries.

2. The method for constructing a knowledge organization system according to claim 1, characterized in that: Retrieve text paragraphs containing target domain terms from the target domain dataset, and input corresponding retrieval results into the term generation model to obtain term generation results output by the term generation model, including: Using the keywords previously set based on the target domain terms, keyword retrieval or similarity matching is performed on the target domain data set to obtain text paragraphs containing the target domain terms as retrieval results; According to the search results, combined with preset prompt words, the term generation model is called to generate term generation results corresponding to the search results according to the preset prompt words; wherein the preset prompt words are set based on the context environment required for the terms in the target field.

3. The method for constructing a knowledge organization system according to claim 1, characterized in that: After obtaining the term generation result output by the term generation model, the method includes: Screening the terms in the term generation results to remove duplicate terms; and / or, Based on preset verification rules, the terms in the term generation result are verified to eliminate the terms that do not comply with the preset verification rules; wherein the preset verification rules include rules for limiting the terms to comply with the target field.

4. The method for constructing a knowledge organization system according to claim 1, characterized in that: Before inputting the terms into the semantic recognition model, include: Obtaining terminology training data and attribute labels and relationship labels corresponding to the terminology training data; Converting the terminology training data and the attribute labels and relationship labels corresponding to the terminology training data according to a preset format to obtain instruction fine-tuning data after format conversion; wherein the preset format is used to limit the arrangement format of the attributes and relationships corresponding to the terms; The semantic recognition model is fine-tuned according to the instruction fine-tuning data and using a preset fine-tuning method.

5. The method for constructing a knowledge organization system according to claim 1, characterized in that: The classification model includes a primary classification number classifier, a secondary classification number classifier, a first tertiary classification number classifier and a second tertiary classification number classifier. The primary classification number classifier is obtained based on the training of term classification training samples and the primary classification number corresponding to the term classification training samples. The secondary classification number classifier is obtained based on the training of term classification training samples and the secondary classification number corresponding to the term classification training samples. The first tertiary classification number classifier is obtained by training the primary classification number classifier with the corresponding subordinate tertiary classification number and its corresponding term classification training samples. The second tertiary classification number classifier is obtained by training the secondary classification number classifier with the corresponding subordinate tertiary classification number and its corresponding term classification training samples. The primary classification number, the secondary classification number and the tertiary classification number are obtained by dividing the term classification training samples based on a preset hierarchical division rule from high to low. For each term in the term generation result, a classification model is used according to the term and the basic information of the term to obtain a corresponding classification result, including: For each term in the term generation result, concatenate each term with the basic information corresponding to the term to obtain a text to be classified; Input the text to be classified into the primary class number classifier and the secondary class number classifier respectively, obtain the primary prediction classification result of the text to be classified output by the primary class number classifier and the secondary prediction classification result of the text to be classified output by the secondary class number classifier, wherein the primary prediction classification result includes the primary prediction classification number corresponding to the text to be classified, and the secondary prediction classification result includes the secondary prediction classification number corresponding to the text to be classified; Determine whether the secondary prediction classification number of the same text to be classified belongs to the corresponding primary prediction classification number, if the secondary prediction classification number of the same text to be classified belongs to the corresponding primary prediction classification number, classify the text to be classified using the second third-level classification number classifier to obtain the third-level prediction classification result of the corresponding text to be classified; otherwise, classify the text to be classified using the first third-level classification number classifier to obtain the third-level prediction classification result of the corresponding text to be classified; The classification result of the corresponding term is obtained according to the first-level prediction classification result of the corresponding text to be classified, the second-level prediction classification result of the corresponding text to be classified and the third-level prediction classification result of the corresponding text to be classified.

6. The method for constructing a knowledge organization system according to claim 5, characterized in that: The three-level classification number classifier includes a fusion classifier and a base classifier. The three-level classification number classifier is used to classify the text to be classified to obtain a three-level predicted classification number, including: Input the text to be classified into each base classifier respectively to obtain the probability value output by each base classifier; The probability values ​​output by each base classifier are output to the fusion classifier for classification decision, and the three-level prediction classification result output by the fusion classifier is obtained.

7. A knowledge organization system construction device, characterized in that: include: A data acquisition module, which acquires a target domain data set, wherein the target domain data set includes a plurality of text paragraphs obtained based on target domain files; A term generation module, which retrieves text paragraphs containing target domain terms from the target domain data set, and inputs the corresponding search results into a term generation model to obtain a term generation result output by the term generation model, wherein the term generation result includes at least one term; wherein the term generation model is obtained by unsupervised training based on unsupervised term training text; An information prediction module, for each term in the term generation result, inputs the term into a basic information prediction model to obtain basic information of the corresponding term output by the basic information prediction model; wherein the basic information prediction model is obtained by unsupervised training based on unsupervised term training text; A semantic recognition module, for each term in the term generation result, inputs the term into a semantic recognition model to obtain a semantic recognition result of the corresponding term output by the semantic recognition model, wherein the semantic recognition result includes attributes and relationships; wherein the semantic recognition model is obtained by fine-tuning a previously trained model based on term training data and attribute labels and relationship labels corresponding to the term training data; A term classification module, for each term in the term generation result, obtains a corresponding classification result by using a classification model according to the term and basic information of the term; wherein the classification model is obtained by fine-tuning a previously trained model based on the term training data and the classification labels corresponding to the term training data; The system construction module generates entries of the corresponding terms according to the basic information, attributes and relationships corresponding to each of the terms and the classification results, and constructs a knowledge organization system based on all the generated entries.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the knowledge organization system construction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for constructing a knowledge organization system as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for constructing a knowledge organization system as claimed in any one of claims 1 to 6 is implemented.

Images