Term multi-level standardization method and device, equipment and storage medium

Abstract

The invention provides a term multi-level standardization method, device and equipment and a storage medium, pre-processing and entity identification are carried out on an original text, a term text to be standardized is automatically extracted, accurate matching is carried out on the basis of a local vocabulary library, and a multi-level matching model is introduced, so that the accuracy of the term text to be standardized is improved. And comprehensively capturing the surface morphological features and the deep semantic features of the terms to generate a semantic similarity score table, thereby improving the accuracy and robustness of term semantic matching. Furthermore, by calling a large language model to conduct reasoning analysis on the semantic relation between the candidate terms, and combining context information and similarity scores, semantic equivalent terms can be accurately recognized, a unique standard term node can be determined, and intelligent synonymous normalization of the terms is achieved. By establishing a term mapping relation table, to-be-standardized terms and standard term nodes are mapped, and semantic matching confidence is generated based on semantic similarity scores, so that a standardized term set has interpretability and hierarchical consistency.

Description

Technical Field

[0001] This invention relates to the field of terminology standardization technology, and in particular to a method, apparatus, device, and storage medium for multi-level terminology standardization. Background Technology

[0002] In the process of building healthcare information systems, significant differences exist in the terminology used by different hospitals, departments, and information systems. For example, the same disease may be described in different forms in an electronic medical record system, such as "diabetes," "type II diabetes," or "type 2 diabetes patient"; the same examination item may appear with different abbreviations or spellings in a laboratory information system. Due to the numerous synonyms, near-synonyms, alternative names, and spelling differences in medical terminology, data is difficult to identify and share uniformly across different systems, affecting the interconnectivity of medical information and the construction of knowledge graphs.

[0003] Existing terminology standardization methods typically rely on manual rules or dictionary-based fuzzy matching, which have limited effectiveness when dealing with large-scale data and complex semantics. On the one hand, dictionary-based precise matching struggles to handle semantic ambiguity and contextual dependencies, resulting in low matching accuracy. On the other hand, semantic matching based on statistics or shallow models fails to capture multi-layered semantic features, easily leading to mismatches or missed matches. Furthermore, existing methods often only achieve single-level alignment at the terminology level, failing to perform in-depth analysis of the contextual semantics and hierarchical relationships of terms, and lacking the ability for multi-level semantic fusion and synonym normalization.

[0004] In summary, the shortcomings of the existing technology urgently need to be addressed. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for multi-level standardization of terminology, which addresses the deficiencies in the prior art and improves the accuracy, automation, and adaptability of terminology standardization.

[0006] This invention provides a method for multi-level standardization of terminology, comprising: Obtain the raw text data and extract the terminology text from the raw text data to obtain the terms to be standardized; The terms to be standardized are precisely matched based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined. When the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. A multi-level matching model is invoked to calculate the semantic similarity between the candidate standard terms and the original text data, thereby obtaining a semantic similarity score table between the term to be standardized and each candidate standard term. Based on the semantic similarity score table, the large language model is invoked to analyze the semantic relationships between candidate standard terms and determine the standard term nodes. The terms to be standardized are mapped to the standard term nodes to obtain a standardized term set.

[0007] According to a terminology multi-level standardization method provided by the present invention, the step of extracting terminology text from the original text data specifically includes: Preprocessing operations are performed on the original text data, including sentence segmentation, noise reduction, and part-of-speech tagging; The terminology text is extracted using an entity recognition model to obtain the terms to be standardized.

[0008] According to the terminology multi-level standardization method provided by the present invention, the step of performing precise matching of the term to be standardized based on a local vocabulary, determining the corresponding standard term node when the matching result is unique, and generating candidate standard terms when the matching result is not unique or fails to match, specifically includes: Retrieve term entries that are the same as or similar to the term to be standardized from the local vocabulary database; When a term entry that is a complete match exists in the search results, the standard term node corresponding to that term entry is determined. When the search results contain multiple partially matching term entries, they are sorted according to edit distance and word frequency similarity to generate candidate standard terms; When no matching entry is found, the extended vocabulary generation module is invoked to generate new candidate standard terms based on a semantic similarity dictionary or rule template.

[0009] According to a terminology multi-level standardization method provided by the present invention, the multi-level matching model includes: A character-level matching layer is used to calculate the edit distance, spelling similarity, and word form variation relationship between the term to be standardized and the candidate standard terms; A word vector-level matching layer is used to calculate the semantic vector similarity between the term to be standardized and the candidate standard term using a pre-trained word vector model; A context-level matching layer is used to combine contextual information from the original text data and capture the semantic differences of the terms to be standardized in the context through a bidirectional encoder structure.

[0010] According to the terminology multi-level standardization method provided by the present invention, the step of calling a multi-level matching model to calculate the semantic similarity between the candidate standard terms and the original text data, and obtaining a semantic similarity score table between the term to be standardized and each candidate standard term, specifically includes: The candidate standard terms and the terms to be standardized are respectively input into the modules of each layer of the multi-level matching model to obtain the level similarity score; The similarity scores at each level are weighted and fused to obtain the overall semantic similarity score between the term to be standardized and each candidate standard term; Based on the overall semantic similarity score, a semantic similarity score table is generated, which is used to characterize the semantic closeness between each candidate standard term and the term to be standardized.

[0011] According to the terminology multi-level standardization method provided by the present invention, the step of analyzing the semantic relationships between candidate standard terms and determining standard term nodes by calling a large language model based on the semantic similarity score table specifically includes: The semantic similarity score table and the candidate standard terms are input into a large language model for semantic relationship analysis to obtain semantically equivalent terms; Based on the semantic similarity score table and context matching degree, a unique standard term node corresponding to the semantically equivalent term is determined.

[0012] According to a terminology multi-level standardization method provided by the present invention, the step of mapping the term to be standardized to the standard term node to obtain a standardized term set specifically includes: Establish a terminology mapping table, which includes the terminology to be standardized, the target standard terminology node, and the semantic matching confidence. The mapping results are written into a standardized terminology set, which includes standard terminology codes, hierarchical paths, and semantic consistency indicators. The semantic matching confidence is determined based on the scores of each candidate standard term in the semantic similarity score table, and is used to characterize the semantic closeness between the term to be standardized and each candidate standard term; when there are multiple candidate nodes, the node with the highest confidence is selected as the mapping result based on the confidence ranking.

[0013] The present invention also provides a terminology multi-level standardization device, comprising: The data acquisition module is used to acquire raw text data and extract terminology text from the raw text data to obtain the terms to be standardized. The precise matching module is used to perform precise matching on the terms to be standardized based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined; when the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. The similarity calculation module is used to call a multi-level matching model to perform semantic similarity calculation on the candidate standard terms and the original text data, and obtain a semantic similarity score table between the term to be standardized and each candidate standard term. The semantic analysis module is used to analyze the semantic relationships between candidate standard terms based on the semantic similarity score table and call a large language model to determine the standard term nodes. The terminology mapping module is used to map the terms to be standardized to the standard term nodes to obtain a standardized terminology set.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the terminology multilevel standardization method as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the terminology multilevel standardization method as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the terminology multilevel standardization method as described above.

[0017] The terminology multi-level standardization method, apparatus, device, and storage medium provided by this invention automatically extract terminology text to be standardized by preprocessing and entity recognition of the original text, effectively improving the accuracy and automation of terminology extraction and reducing manual intervention.

[0018] Secondly, this invention performs precise matching based on a local vocabulary database. When the matching result is unique, it directly determines the standard term node to achieve efficient standardization. When the matching result is not unique or the match is unsuccessful, it generates a set of candidate standard terms to ensure matching coverage in complex semantic scenarios.

[0019] Furthermore, this invention introduces a multi-level matching model, which calculates the semantic similarity between terms at three levels: character level, word vector level, and context level. It comprehensively captures the surface morphological features and deep semantic features of terms to generate a semantic similarity score table, thereby improving the accuracy and robustness of semantic matching of terms.

[0020] Furthermore, this invention uses a large language model to reason about the semantic relationships between candidate terms. By combining contextual information and similarity scores, it can accurately identify semantically equivalent terms and determine unique standard term nodes, thereby achieving intelligent synonym normalization of terms.

[0021] Finally, this invention establishes a term mapping table to map the terms to be standardized to standard term nodes, and generates semantic matching confidence based on semantic similarity scores, so that the standardized term set has interpretability and hierarchical consistency, providing a high-quality standardized semantic foundation for medical knowledge graph construction, cross-system data exchange and intelligent question answering.

[0022] In summary, this invention achieves intelligent standardization of medical terms throughout the entire process, from text extraction and semantic matching to synonym normalization, through the collaborative processing of multi-level semantic modeling and large language models. This significantly improves the accuracy, automation, and adaptability of terminology standardization, and has good practical value and promotional significance. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the multi-level standardization method for terms provided by the present invention; Figure 2 This is a schematic diagram of the structure of the multi-level terminology standardization device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] To address the problems in existing technologies, this invention proposes a multi-level terminology standardization method to improve the accuracy, automation, and adaptability of terminology standardization. The multi-level terminology standardization method is described below, as follows: Figure 1 As shown, including but not limited to the following steps: Step 110: Obtain the original text data and extract the terminology text from the original text data to obtain the terminology to be standardized.

[0027] In step 110, the system first obtains raw text data from the target corpus source. The raw text data may include document content from medical, financial, legal, scientific research, or other professional fields. The system uses Natural Language Processing (NLP) technology to perform word segmentation, part-of-speech tagging, and named entity recognition on the text, extracting terminology text with professional significance.

[0028] In this embodiment, the extraction module can use a BiLSTM-CRF-based named entity recognition model to perform sequence annotation on the original text in order to identify a set of potential term entities, which will be used as input for the term to be standardized into the subsequent processing flow.

[0029] Step 120: Perform precise matching on the terms to be standardized based on the local vocabulary. When the matching result is unique, determine the corresponding standard term node; when the matching result is not unique or the match is unsuccessful, generate candidate standard terms.

[0030] In step 120, the system maintains a preset local vocabulary, which stores standard term nodes in the target domain and their synonym mapping relationships. The system first performs an exact match operation on the term to be standardized. When the match result is unique, it is considered that the term already exists in the standard system, and its corresponding standard term node is directly determined.

[0031] If the matching result has multiple candidate terms or no completely identical match can be found in the vocabulary, the system will take the term as the "fuzzy matching target" and generate several candidate standard terms based on string edit distance, word form restoration and synonym expansion rules to enter the subsequent semantic calculation steps.

[0032] Step 130: Call the multi-level matching model to calculate the semantic similarity between the candidate standard terms and the original text data, and obtain the semantic similarity score table between the term to be standardized and each candidate standard term.

[0033] In step 130, the system calls the pre-trained multi-level semantic matching model to quantitatively calculate the semantic relationship between the term to be standardized and each candidate standard term.

[0034] In one optional implementation, the multi-level matching model includes a shallow matching layer based on word vectors, a deep matching layer based on contextual semantics, and a structural matching layer based on syntactic dependencies. The shallow matching layer calculates preliminary matching results using word vector similarity metrics; the deep matching layer uses a Transformer structure to perform semantic modeling of the term context; and the structural matching layer further utilizes dependency syntax trees to calculate structural consistency scores.

[0035] The system combines and weights the results of the three-layer calculation to obtain the semantic similarity score of each candidate term, and outputs it in tabular form as a "semantic similarity score table". This score table is used to intuitively reflect the semantic similarity between the term to be standardized and the candidate terms.

[0036] Step 140: Based on the semantic similarity score table, call the large language model to analyze the semantic relationship between candidate standard terms and determine the standard term nodes.

[0037] In step 140, the system takes the semantic similarity score table as input and calls a large language model (such as GPT, LLM, or BERT derivative models) to perform semantic hierarchical relationship analysis. The large language model uses contextual semantic reasoning and concept aggregation mechanisms to determine the synonym, hypernym, hyponym, or related relationships between candidate terms, thereby comprehensively determining the standard term node that best matches the original semantic expression.

[0038] In this embodiment, the inference output of the large language model includes confidence scores and explanatory text. The system selects the final standard term nodes based on the confidence threshold, thereby achieving the unification and standardization of term semantics.

[0039] Step 150: Map the terms to be standardized to the standard term nodes to obtain a standardized term set.

[0040] In step 150, the system establishes a mapping relationship between the semantically validated standard term nodes and the corresponding terms to be standardized, forming a standardized term set.

[0041] This terminology set can be used to build knowledge graphs, unify business data labeling systems, or perform cross-system term mapping, providing a consistent knowledge foundation for subsequent information extraction, semantic search, and intelligent question answering.

[0042] In one alternative implementation, the system can also store the standardized terminology set in a unified terminology knowledge base, supporting subsequent batch updates and incremental learning, thereby further improving standardization efficiency and knowledge coverage.

[0043] As a further optional embodiment, the step of extracting terminology text from the original text data specifically includes: Preprocessing operations are performed on the original text data, including sentence segmentation, noise reduction, and part-of-speech tagging; The terminology text is extracted using an entity recognition model to obtain the terms to be standardized.

[0044] In this embodiment, the sentence segmentation operation is used to divide the original text into multiple sentence units based on punctuation marks, semantic pauses, or format features for subsequent processing; the noise reduction operation is used to remove invalid characters, HTML tags, and non-semantic symbol information, thereby improving text quality; and the part-of-speech tagging is used to identify the part-of-speech features of each word in the text, providing contextual feature input for the entity recognition model.

[0045] Furthermore, the entity recognition model can be a sequence labeling model based on bidirectional long short-term memory network and conditional random field (BiLSTM-CRF). By performing feature encoding and label prediction on the preprocessed text sequence, it can identify terminology text with professional meaning and form a set of terms to be standardized.

[0046] This method can accurately extract semantically loaded term entities while maintaining the semantic integrity of the original text, providing high-quality input for subsequent term standardization matching.

[0047] As a further optional embodiment, the step of performing precise matching of the terms to be standardized based on a local vocabulary, determining the corresponding standard term node when the matching result is unique, and generating candidate standard terms when the matching result is not unique or fails to match, specifically includes: Retrieve term entries that are the same as or similar to the term to be standardized from the local vocabulary database; When a term entry that is a complete match exists in the search results, the standard term node corresponding to that term entry is determined. When the search results contain multiple partially matching term entries, they are sorted according to edit distance and word frequency similarity to generate candidate standard terms; When no matching entry is found, the extended vocabulary generation module is invoked to generate new candidate standard terms based on a semantic similarity dictionary or rule template.

[0048] In this embodiment, the local vocabulary database can be a pre-built professional domain terminology database, which includes terminology encoding, standard definitions, and multilingual mapping relationships.

[0049] The edit distance is used to measure the degree of character difference between the term to be standardized and the term in the vocabulary, and the word frequency similarity is used to measure the co-occurrence pattern of the two in the domain corpus, thereby improving the accuracy of matching and ranking.

[0050] Furthermore, the extended vocabulary generation module can generate extended candidate terms that are semantically related to the target term based on WordNet semantic hierarchy, thesaurus, or custom rule templates, in order to cover the matching omission problem caused by the diversification of term expressions.

[0051] By following the steps above, a dynamic balance can be achieved between precise term matching and semantic expansion, ensuring maximum coverage and high-confidence matching of the terms to be standardized.

[0052] As a further optional embodiment, the multi-level matching model includes: A character-level matching layer is used to calculate the edit distance, spelling similarity, and word form variation relationship between the term to be standardized and the candidate standard terms; A word vector-level matching layer is used to calculate the semantic vector similarity between the term to be standardized and the candidate standard term using a pre-trained word vector model; A context-level matching layer is used to combine contextual information from the original text data and capture the semantic differences of the terms to be standardized in the context through a bidirectional encoder structure.

[0053] In this embodiment, the character-level matching layer quantifies the character differences between the term to be standardized and the candidate standard terms by calculating the edit distance between them. At the same time, it combines spell similarity to evaluate the impact of common spelling errors or typing deviations on matching, and further considers lexical inflection relationships, such as stemming, affixation and other linguistic features, to obtain a preliminary similarity score.

[0054] The word vector level matching layer uses a pre-trained medical domain word vector model to map the term to be standardized and the candidate standard term into high-dimensional semantic vectors. By calculating the cosine similarity or other distance metrics between the vectors, it quantifies the closeness between the two in the semantic space, so as to capture information on synonyms, near-synonyms and semantic relevance.

[0055] The context-level matching layer combines the contextual information of the original text data and adopts a bidirectional encoder structure (such as Bi-LSTM, Transformer or BERT-like models) to capture semantic dependencies and contextual differences while encoding the term to be standardized and its surrounding context, thereby further improving the accurate matching ability of terms in different contexts and generating semantic similarity scores at the context level.

[0056] Through the above three-level matching mechanism, the multi-level matching model can integrate character, semantic and contextual information to conduct a refined evaluation of the similarity between candidate terms and terms to be standardized, providing reliable input for the semantic reasoning of subsequent large language models.

[0057] As a further optional embodiment, the step of calling a multi-level matching model to calculate the semantic similarity between the candidate standard terms and the original text data, and obtaining a semantic similarity score table between the term to be standardized and each candidate standard term, specifically includes: The candidate standard terms and the terms to be standardized are respectively input into the modules of each layer of the multi-level matching model to obtain the level similarity score; The similarity scores at each level are weighted and fused to obtain the overall semantic similarity score between the term to be standardized and each candidate standard term; Based on the overall semantic similarity score, a semantic similarity score table is generated, which is used to characterize the semantic closeness between each candidate standard term and the term to be standardized.

[0058] In this embodiment, the terms to be standardized and their corresponding candidate standard terms are input into the modules of each layer of the multi-level matching model for processing. Specifically, the character-level matching layer calculates the scores for edit distance, spelling similarity, and word form change relationship; the word vector-level matching layer calculates the semantic vector similarity score; and the context-level matching layer captures contextual semantic differences based on the contextual information of the original text data and generates corresponding scores.

[0059] Subsequently, the similarity scores from each layer are weighted and fused to obtain the overall semantic similarity score between the term to be standardized and each candidate standard term. The weighting coefficients can be pre-set according to the matching importance of different levels or obtained through training optimization to take into account the combined influence of character, semantic, and contextual information.

[0060] Finally, a semantic similarity score table is generated based on the overall semantic similarity score of each candidate standard term. This score table is used to characterize the semantic closeness between each candidate standard term and the term to be standardized, providing input for subsequent semantic relationship analysis and standard term node determination in the large language model, thereby achieving accurate term standardization mapping.

[0061] As a further optional embodiment, the step of analyzing the semantic relationships between candidate standard terms and determining standard term nodes based on the semantic similarity score table by calling a large language model specifically includes: The semantic similarity score table and the candidate standard terms are input into a large language model for semantic relationship analysis to obtain semantically equivalent terms; Based on the semantic similarity score table and context matching degree, a unique standard term node corresponding to the semantically equivalent term is determined.

[0062] In this embodiment, the semantic similarity score table and the candidate standard terms are fed into the large language model as input. The large language model uses its pre-trained language understanding capabilities to analyze the semantic relationships between the candidate standard terms, identify semantically equivalent or highly similar term combinations, and form a set of semantically equivalent terms.

[0063] Subsequently, by combining the scores of each candidate term in the semantic similarity score table with the contextual information (contextual matching degree) of the original text data, a comprehensive evaluation of the semantically equivalent term set is performed to determine a unique standard term node. This standard term node ensures both semantic consistency with the term to be standardized and accuracy of the contextual information, thus providing a reliable basis for the final standardized term mapping.

[0064] As a further optional embodiment, the step of mapping the terms to be standardized to the standard term nodes to obtain a standardized term set specifically includes: Establish a terminology mapping table, which includes the terminology to be standardized, the target standard terminology node, and the semantic matching confidence. The mapping results are written into a standardized terminology set, which includes standard terminology codes, hierarchical paths, and semantic consistency indicators. The semantic matching confidence is determined based on the scores of each candidate standard term in the semantic similarity score table, and is used to characterize the semantic closeness between the term to be standardized and each candidate standard term; when there are multiple candidate nodes, the node with the highest confidence is selected as the mapping result based on the confidence ranking.

[0065] In this embodiment, the system first establishes a term mapping table based on the correspondence between standard term nodes and terms to be standardized obtained in the preceding steps. This mapping table records each term to be standardized, its corresponding target standard term node, and the semantic matching confidence score calculated based on the semantic similarity score table. The confidence score is used to characterize the semantic closeness between the term to be standardized and each candidate standard term.

[0066] When multiple candidate nodes exist for mapping, the system sorts the candidate nodes according to semantic matching confidence and selects the node with the highest confidence as the final mapping result. Subsequently, the final mapping result is written into a standardized terminology set, which includes standard term codes, hierarchical paths, and semantic consistency indicators, thereby achieving unified standardization of the original terms and ensuring the consistency and traceability of terms in subsequent applications.

[0067] The terminology multi-level standardization device provided by this invention is described below, such as... Figure 2 As shown, the terminology multilevel standardization apparatus described below and the terminology multilevel standardization method described above can be referred to in correspondence.

[0068] A terminology multi-level standardization device, comprising: The data acquisition module 210 is used to acquire raw text data and extract terminology text from the raw text data to obtain terms to be standardized. The precise matching module 220 is used to perform precise matching on the terms to be standardized based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined; when the matching result is not unique or the matching is unsuccessful, candidate standard terms are generated. The similarity calculation module 230 is used to call a multi-level matching model to perform semantic similarity calculation on the candidate standard terms and the original text data, and obtain a semantic similarity score table between the term to be standardized and each candidate standard term. The semantic analysis module 240 is used to analyze the semantic relationships between candidate standard terms based on the semantic similarity score table and call a large language model to determine the standard term nodes. The terminology mapping module 250 is used to map the terminology to be standardized to the standard terminology node to obtain a standardized terminology set.

[0069] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can invoke logical instructions in the memory 330 to execute a terminology multi-level standardization method, which includes: Obtain the raw text data and extract the terminology text from the raw text data to obtain the terms to be standardized; The terms to be standardized are precisely matched based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined. When the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. A multi-level matching model is invoked to calculate the semantic similarity between the candidate standard terms and the original text data, thereby obtaining a semantic similarity score table between the term to be standardized and each candidate standard term. Based on the semantic similarity score table, the large language model is invoked to analyze the semantic relationships between candidate standard terms and determine the standard term nodes. The terms to be standardized are mapped to the standard term nodes to obtain a standardized term set.

[0070] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] On the other hand, the present invention also provides a computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the terminology multi-level standardization method provided by the above methods, the method comprising: Obtain the raw text data and extract the terminology text from the raw text data to obtain the terms to be standardized; The terms to be standardized are precisely matched based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined. When the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. A multi-level matching model is invoked to calculate the semantic similarity between the candidate standard terms and the original text data, thereby obtaining a semantic similarity score table between the term to be standardized and each candidate standard term. Based on the semantic similarity score table, the large language model is invoked to analyze the semantic relationships between candidate standard terms and determine the standard term nodes. The terms to be standardized are mapped to the standard term nodes to obtain a standardized term set.

[0072] In another aspect, 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 perform the terminology multi-level standardization method provided by the methods described above, the method comprising: Obtain the raw text data and extract the terminology text from the raw text data to obtain the terms to be standardized; The terms to be standardized are precisely matched based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined. When the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. A multi-level matching model is invoked to calculate the semantic similarity between the candidate standard terms and the original text data, thereby obtaining a semantic similarity score table between the term to be standardized and each candidate standard term. Based on the semantic similarity score table, the large language model is invoked to analyze the semantic relationships between candidate standard terms and determine the standard term nodes. The terms to be standardized are mapped to the standard term nodes to obtain a standardized term set.

[0073] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0074] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for multi-level standardization of terminology, characterized in that, include: Obtain the raw text data and extract the terminology text from the raw text data to obtain the terms to be standardized; The terms to be standardized are precisely matched based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined. When the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. A multi-level matching model is invoked to calculate the semantic similarity between the candidate standard terms and the original text data, thereby obtaining a semantic similarity score table between the term to be standardized and each candidate standard term. Based on the semantic similarity score table, the large language model is invoked to analyze the semantic relationships between candidate standard terms and determine the standard term nodes. The terms to be standardized are mapped to the standard term nodes to obtain a standardized term set.

2. The terminology multi-level standardization method according to claim 1, characterized in that, The step of extracting terminology text from the original text data specifically includes: Preprocessing operations are performed on the original text data, including sentence segmentation, noise reduction, and part-of-speech tagging; The terminology text is extracted using an entity recognition model to obtain the terms to be standardized.

3. The terminology multi-level standardization method according to claim 1, characterized in that, The terminology to be standardized is precisely matched based on the local vocabulary database. When the matching result is unique, the corresponding standard term node is determined. When the matching results are not unique or no match is found, the step of generating candidate standard terms specifically includes: Retrieve term entries that are the same as or similar to the term to be standardized from the local vocabulary database; When a term entry that is a complete match exists in the search results, the standard term node corresponding to that term entry is determined. When the search results contain multiple partially matching term entries, they are sorted according to edit distance and word frequency similarity to generate candidate standard terms; When no matching entry is found, the extended vocabulary generation module is invoked to generate new candidate standard terms based on a semantic similarity dictionary or rule template.

4. The terminology multi-level standardization method according to claim 1, characterized in that, The multi-level matching model includes: A character-level matching layer is used to calculate the edit distance, spelling similarity, and word form variation relationship between the term to be standardized and the candidate standard terms; A word vector-level matching layer is used to calculate the semantic vector similarity between the term to be standardized and the candidate standard term using a pre-trained word vector model; A context-level matching layer is used to combine contextual information from the original text data and capture the semantic differences of the terms to be standardized in the context through a bidirectional encoder structure.

5. The terminology multi-level standardization method according to claim 4, characterized in that, The step of calling a multi-level matching model to calculate the semantic similarity between the candidate standard terms and the original text data, and obtaining a semantic similarity score table between the term to be standardized and each candidate standard term, specifically includes: The candidate standard terms and the terms to be standardized are respectively input into the modules of each layer of the multi-level matching model to obtain the level similarity score; The similarity scores at each level are weighted and fused to obtain the overall semantic similarity score between the term to be standardized and each candidate standard term; Based on the overall semantic similarity score, a semantic similarity score table is generated, which is used to characterize the semantic closeness between each candidate standard term and the term to be standardized.

6. The terminology multi-level standardization method according to claim 1, characterized in that, The step of analyzing the semantic relationships between candidate standard terms and determining standard term nodes based on the semantic similarity score table and calling a large language model specifically includes: The semantic similarity score table and the candidate standard terms are input into a large language model for semantic relationship analysis to obtain semantically equivalent terms; Based on the semantic similarity score table and context matching degree, a unique standard term node corresponding to the semantically equivalent term is determined.

7. The terminology multi-level standardization method according to claim 1, characterized in that, The step of mapping the terms to be standardized to the standard term nodes to obtain a standardized term set specifically includes: Establish a terminology mapping table, which includes the terminology to be standardized, the target standard terminology node, and the semantic matching confidence. The mapping results are written into a standardized terminology set, which includes standard terminology codes, hierarchical paths, and semantic consistency indicators. The semantic matching confidence is determined based on the scores of each candidate standard term in the semantic similarity score table, and is used to characterize the semantic closeness between the term to be standardized and each candidate standard term; when there are multiple candidate nodes, the node with the highest confidence is selected as the mapping result based on the confidence ranking.

8. A terminology multi-level standardization device, characterized in that, include: The data acquisition module is used to acquire raw text data and extract terminology text from the raw text data to obtain the terms to be standardized. The precise matching module is used to perform precise matching on the terms to be standardized based on the local vocabulary. When the matching result is unique, the corresponding standard term node is determined; when the matching result is not unique or the match is unsuccessful, candidate standard terms are generated. The similarity calculation module is used to call a multi-level matching model to perform semantic similarity calculation on the candidate standard terms and the original text data, and obtain a semantic similarity score table between the term to be standardized and each candidate standard term. The semantic analysis module is used to analyze the semantic relationships between candidate standard terms based on the semantic similarity score table and call a large language model to determine the standard term nodes. The terminology mapping module is used to map the terms to be standardized to the standard term nodes to obtain a standardized terminology set.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the terminology multilevel standardization method as described in any one of claims 1 to 7.

10. 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, it implements the terminology multilevel standardization method as described in any one of claims 1 to 7.

Images