Common language semantic recognition method based on energy industry material field

Abstract

The invention discloses a common language semantic recognition method based on the field of energy industry materials, and belongs to the technical field of speech recognition, and the method comprises the steps: S1, obtaining energy industry semantic data, and improving an entity recognition model based on a bidirectional long short-term memory network and a soft word list to obtain a target recognition model; s2, extracting semantic entities and semantic vectors according to the semantic data and the target recognition model, and performing entity linking based on the semantic entities and the semantic vectors to construct a three-dimensional entity set; s3, performing knowledge reasoning based on the three-dimensional entity set to obtain a knowledge ontology, and performing quality evaluation on the knowledge ontology to obtain a confidence coefficient; s4, constructing a semantic library based on the confidence coefficient and the knowledge ontology, and obtaining real-time statements of workers; s5, analyzing the real-time statement according to the semantic library and the target recognition model to obtain a material name corresponding to a common language in the real-time statement; according to the invention, the problem that the target material of the worker cannot be quickly identified because the correlation between the common language and the standard name is not considered in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to a colloquialism semantic recognition method based on the energy industry material field. Background Art

[0002] With the development of power system technology, the types of distribution network materials are increasing. Due to professional continuity and other reasons, these distribution network materials contain customary slangs that are not covered by the State Grid material master data, such as π boxes. These slangs sometimes need to be repeatedly confirmed in different links of the work, reducing the efficiency of submitting corresponding requirements; in addition, since these slangs need to be repeatedly confirmed, business personnel need to perform a large number of solidified and repetitive simple operations, wasting a lot of time.

[0003] Chinese patent, publication number: CN117493491A, publication date: February 2, 2024, discloses a natural language processing method and system based on machine learning, including: based on raw text data, using text cleaning and standardization algorithms, including denoising, word segmentation and stop word removal, to generate preprocessed text data; based on the preprocessed text data, using word embedding technology, based on Word2Vec or GloVe, to convert text into numerical features to generate a text feature vector; based on the text feature vector, using a Transformer model, through BERT, to perform context understanding and generate a context-aware feature representation; based on the context-aware feature representation, using a deep neural network to perform text classification, identify user intent, and generate a user intent recognition result; based on the user intent recognition result, using collaborative filtering and content-based recommendation algorithms to provide personalized recommendations and generate personalized recommendation content; based on the personalized recommendation content, using reinforcement learning and user behavior analysis to perform query response optimization and generate optimized search results; However, this invention does not consider the association between colloquialisms and standard names, and thus cannot quickly identify the intended materials of the staff. Summary of the Invention

[0004] The purpose of the present invention is to address the problem that the existing technology does not consider the association between colloquialisms and standard names, resulting in the inability to quickly identify the materials intended by staff; a colloquialism semantic recognition method based on the field of materials in the energy industry is proposed, and an entity recognition model is improved by a bidirectional long short-term memory network and a soft word list to obtain a target recognition model, and the semantic entities and semantic vectors extracted from the semantic data of the energy industry by the target recognition model are used to construct a three-dimensional entity set, and then a semantic library is constructed according to the knowledge ontology established based on the three-dimensional entity set and the confidence of the ontology in the knowledge ontology, and finally, the real-time sentences of the staff are analyzed according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentences, which effectively improves the recognition efficiency of the materials intended by the staff, and does not require repeated confirmation of the same colloquial name, which significantly improves the efficiency of reporting the requirements corresponding to the colloquial name.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is a method for semantic recognition of colloquialisms in the field of energy industry materials, comprising the following steps: S1. Obtain the semantic data of the energy industry, and improve the entity recognition model based on the bidirectional long short-term memory network and the soft word list to obtain a target recognition model; S2. Extracting semantic entities and semantic vectors based on the semantic data and the target recognition model, and performing entity linking based on the semantic entities and semantic vectors to construct a three-dimensional entity set; S3. Perform knowledge reasoning based on the three-dimensional entity set to obtain a knowledge ontology, and perform quality assessment on the knowledge ontology to obtain a confidence level. S4. Build a semantic library based on the confidence level and the knowledge ontology, and obtain real-time statements from staff members. S5. Analyze the real-time sentence according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentence.

[0006] In this solution, the entity recognition model is improved based on the bidirectional long short-term memory network and the soft word list to obtain the target recognition model, which can enhance the target recognition model's ability to capture sequence information, and extract semantic entities and semantic vectors from the semantic data of the energy industry to more deeply reflect the essential characteristics of the data. The fragmented semantic entities and semantic vectors are linked together by the entity linking technology to obtain a three-dimensional entity set, and knowledge reasoning is performed on the three-dimensional entity set to further explore the potential relationships and information hidden in the data to obtain a more complete knowledge ontology. The knowledge ontology is then quality evaluated, and the reliable ontology in the knowledge ontology is found with the help of the confidence of the ontology. Finally, the real-time sentences of the staff are analyzed according to the semantic library and target recognition model constructed based on the reliable ontology to obtain the names of materials corresponding to the slang in the real-time sentences, which effectively improves the recognition efficiency of the staff's target materials, and there is no need to repeatedly confirm the same slang, which significantly improves the efficiency of reporting the corresponding requirements of the slang.

[0007] Preferably, in S1, the specific process of improving the entity recognition model based on the bidirectional long short-term memory network and the soft word list to obtain the target recognition model is: S111, extracting an embedding module and a recognition module based on the structure of the entity recognition model; S112, improving the recognition module based on a bidirectional long short-term memory network to obtain a target recognition module, and adding a calibration entity and a calibration object to the embedding module based on a soft word list to obtain a logical embedding module; S113. Integrate the target recognition module and the logic embedding module based on the structure of the entity recognition model to obtain a target recognition model.

[0008] In this solution, the data faced by the target recognition model during information extraction is diverse in form and cross-language, which makes information extraction of the target recognition model difficult. Therefore, a bidirectional long short-term memory network is introduced into the entity recognition model to improve the bidirectional recognition capability of the entity recognition model and enhance the contextual application capability of the entity recognition model. A soft word list is introduced into the entity recognition model to formulate annotation specifications and enhance the entity recognition model's ability to process complex text data.

[0009] Preferably, in S2, the specific process of extracting semantic entities and semantic vectors based on semantic data and target recognition model is as follows: S211, inputting the semantic data into a target recognition model to perform entity calibration to obtain entities and objects, and classifying the entities, objects and linear classifiers to obtain entity relationships and entity attributes; S212: Arrange the entities, objects, entity relationships and entity attributes to obtain semantic entities and semantic vectors.

[0010] In this solution, entities and objects are classified through linear classifiers to make the data features extracted by the target recognition model more refined and accurate, and the entities, objects, entity relationships and entity attributes are organized to obtain structured semantic entities and semantic vectors, making the corresponding data features concise and clear, providing support for subsequent steps.

[0011] Preferably, in S2, the specific process of performing entity linking based on the semantic entities and semantic vectors to construct a three-dimensional entity set is: S221. Segment the semantic entities based on a natural language processing algorithm and organize them to obtain an entity set; Simultaneously, similarity analysis is performed on the entities in the entity set, and semantic vectors are reconstructed based on the results of the similarity analysis; S222, candidate entities are extracted from the energy industry semantic data based on the entity set, and the entities in the entity set and the candidate entities are paired based on the semantic vectors; S223: Construct a three-dimensional entity set based on the pairing result, the entity, and the candidate entity.

[0012] In this solution, semantic entities are segmented to obtain a more complete entity set. However, there is entity overlap in the entity set at this time. Therefore, similarity analysis is performed on the entities to eliminate entity overlap, and semantic vectors are reconstructed based on the results of the similarity analysis to eliminate the semantic vector overlap caused by entity overlap. Then, candidate entities that match the entities in the entity set in the energy industry semantic data are extracted to improve the professionalism and accuracy of the entity set, and the robustness and accuracy of the corresponding entity links are improved through the matching relationship.

[0013] Preferably, in S3, the specific process of performing knowledge reasoning based on the three-dimensional entity set to obtain the knowledge ontology is: S311, extracting entity parallel relationships and entity hierarchical relationships based on the three-dimensional entity set, and performing parallel similarity analysis based on the entity parallel relationships to obtain entity parallel features; S312: Based on the entity parallel characteristics and entity hierarchical relationships and the corresponding energy industry material supply rules, establish data definitions and structure definitions for the materials corresponding to the entities in the three-dimensional entity collection; S313: Construct a knowledge ontology based on the data definition and structure definition.

[0014] In this solution, the three-dimensional characteristics of the entities in the three-dimensional entity set are manifested through the parallel relationship of entities and the hierarchical relationship of entities, and the overlapping phenomenon in the three-dimensional characteristics of entities is eliminated through parallel similarity analysis, thereby enhancing the logical connection between different entities. Based on this three-dimensional characteristic and combined with the material supply rules of the energy industry, the data definition and structure definition of the materials corresponding to the entities in the three-dimensional entity set are established, thereby promoting the practicality of the knowledge ontology finally constructed and effectively improving the sharing performance of the knowledge ontology.

[0015] Preferably, in S3, the specific process of performing quality assessment on the knowledge ontology to obtain the confidence level is as follows: S321. Perform consistency analysis based on semantic entities and semantic vectors corresponding to the ontology in the knowledge ontology to obtain an ontology consistency degree, and perform overlap analysis based on the knowledge ontology and energy industry semantic data to obtain an ontology overlap degree; S322: Calculate the confidence of the ontology in the knowledge ontology according to the ontology consistency and ontology overlap.

[0016] In this solution, the confidence of the ontology in the knowledge ontology is calculated by using the consistency of the ontology itself and the similarity between the ontology and semantic data, that is, the overlap, so that the credibility of the ontology can be quantified.

[0017] Preferably, in S4, the specific process of constructing the semantic library based on the confidence level and knowledge ontology is: S411, screening ontologies in the knowledge ontology based on the confidence to obtain ontologies with qualified confidence, and updating the knowledge ontology according to the ontologies with qualified confidence; S412: Establish a vocabulary based on the semantic entities corresponding to the ontology in the updated knowledge ontology, and establish a semantic library based on the vocabulary and the semantic vectors corresponding to the semantic entities in the vocabulary.

[0018] In this scheme, confidence is used to screen ontologies in the knowledge ontology, which can eliminate ontologies with low credibility, thereby ensuring the quality of the vocabulary and effectively improving the reliability of the semantic library finally obtained by fusion.

[0019] Preferably, in S5, the specific process of analyzing the real-time sentence according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentence is: S511, inputting the real-time sentence into the target recognition model for analysis to obtain real-time semantic entities and real-time semantic vectors, and establishing a knowledge graph based on the semantic library; S512: Perform similarity analysis between the real-time semantic entity and the semantic entities in the semantic library, and fill the real-time semantic entity and the real-time semantic vector into the knowledge graph based on the similarity analysis result.

[0020] In this solution, the semantic entities of the real-time sentences are analyzed for similarity with the semantic entities in the semantic library to accurately determine the type of target materials in the real-time sentences, and a corresponding knowledge graph is established to clearly display the connections between the entities.

[0021] On the other hand, a technical solution also provided in an embodiment of the present invention is an electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the electronic device executes any one of the above methods.

[0022] On the other hand, a technical solution also provided in an embodiment of the present invention is a storage medium, which stores computer-executable instructions. When the computer-executable instructions are loaded and executed by a processor, the computer executes any one of the above methods.

[0023] The present invention has at least the following substantial effects: This application introduces a bidirectional long short-term memory network into the entity recognition model to improve the bidirectional recognition ability of the entity recognition model, enhance the context application ability of the entity recognition model, introduces a soft word list into the entity recognition model to formulate annotation specifications, and enhances the entity recognition model's processing ability for complex text data, thereby obtaining a target recognition model with a strong ability to capture sequence information, and extracts semantic entities and semantic vectors from the semantic data of the energy industry to more deeply reflect the essential characteristics of the data. Through entity linking technology, the fragmented semantic entities and semantic vectors are linked to obtain a three-dimensional entity set, and knowledge reasoning is performed on the three-dimensional entity set to further explore the potential relationships and information hidden in the data to obtain a more complete knowledge ontology, and the knowledge ontology is quality evaluated. With the help of the confidence of the ontology, a reliable ontology in the knowledge ontology is found. Finally, based on the semantic library constructed with the reliable ontology and the target recognition model, the real-time sentences of the staff are analyzed to obtain the names of materials corresponding to the slang in the real-time sentences, effectively improving the recognition efficiency of the staff's target materials, and there is no need to repeatedly confirm the same slang, which significantly improves the efficiency of reporting the corresponding requirements of the slang. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects, and advantages of the present invention will become more apparent upon reading the detailed description of the non-limiting embodiments made with reference to the following drawings. The drawings are for the purpose of illustrating preferred embodiments only and are not to be construed as limiting the present invention. Like reference characters are used throughout the drawings to designate like parts.

[0025] Figure 1 This is a flow chart of a colloquialism semantic recognition method based on the energy industry material field according to the present invention. DETAILED DESCRIPTION

[0026] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be performed in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but can also have additional steps not included in the figures; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0028] Example 1: like Figure 1 As shown, this embodiment provides a method for identifying colloquialisms in the field of energy industry materials, including the following steps: S1. Obtain the semantic data of the energy industry, and improve the entity recognition model based on the bidirectional long short-term memory network and the soft word list to obtain a target recognition model; S111, extracting an embedding module and a recognition module based on the structure of the entity recognition model; S112, improving the recognition module based on a bidirectional long short-term memory network to obtain a target recognition module, and adding a calibration entity and a calibration object to the embedding module based on a soft word list to obtain a logical embedding module; S113. Integrate the target recognition module and the logic embedding module based on the structure of the entity recognition model to obtain a target recognition model.

[0029] In this embodiment, considering that automatically acquiring basic information on domain entities and entity relationships is a prerequisite for constructing a large-scale domain knowledge graph, and that entities and their key elements in the domain need to be extracted from multiple sources, and that information extraction faces problems such as diverse information presentation forms, cross-language descriptions, possible mixing of multiple languages, unknown names of people and institutional entities, lack of labeling specifications, and lack of labeled samples, a SpanBERT+CRF entity recognition and relationship discovery joint extraction model can be used as the entity recognition model. However, the entity recognition model SpanBERT+CRF will have problems such as labeling confusion when faced with a large number of unknown entities. A bidirectional long short-term memory network BiLSTM sequence can be added to the context-related representation generated by the embedding module SpanBERT, and combined with a soft word list on the basis of the recognition module CRF to generate a target recognition model with powerful complex text processing capabilities. At the same time, an iterative method combining manual and machine methods can be used to further solve the problem of lack of labeled samples.

[0030] S2. Extract semantic entities and semantic vectors based on semantic data and target recognition models, and perform entity linking based on the semantic entities and semantic vectors to construct a three-dimensional entity set.

[0031] In one embodiment, the specific process of extracting semantic entities and semantic vectors based on semantic data and object recognition model is as follows: S211, inputting the semantic data into a target recognition model to perform entity calibration to obtain entities and objects, and classifying the entities, objects and linear classifiers to obtain entity relationships and entity attributes; S212: Arrange the entities, objects, entity relationships and entity attributes to obtain semantic entities and semantic vectors.

[0032] In this embodiment, the semantic data should include relevant management systems, work instructions, operating details and other data corresponding to the energy industry field, and then the target recognition model can mark the entities and objects corresponding to various materials based on these semantic data. Specifically, special tags are embedded in the corresponding positions of the calibration subject and the calibration object in the BERT-class pre-training module of the target recognition model, or special tags are added before and after the subject and object, and a linear classifier is used to classify the subject and object to extract entity relationships and entity attributes, that is, whether there is a relationship between the entities corresponding to different subjects and objects, and the attribute information corresponding to these entities.

[0033] In one embodiment, the specific process of constructing a three-dimensional entity set by entity linking based on the semantic entities and semantic vectors is as follows: S221. Segment the semantic entities based on a natural language processing algorithm and organize them to obtain an entity set; Simultaneously, similarity analysis is performed on the entities in the entity set, and semantic vectors are reconstructed based on the results of the similarity analysis; S222, candidate entities are extracted from the energy industry semantic data based on the entity set, and the entities in the entity set and the candidate entities are paired based on the semantic vectors; S223: Construct a three-dimensional entity set based on the pairing result, the entity, and the candidate entity.

[0034] In this embodiment, semantic entities are segmented based on the spacy library in the natural language processing algorithm NLP, and the real entities hidden in false entity objects such as phrases and short sentences are analyzed. The idioms, synonyms, and antonyms of the vocabulary corresponding to the entities after the segmentation are replaced and rewritten to enrich and improve the semantic information of the sentences. However, at this time, the entities and the corresponding attribute information of the entities obtained from the original semi-structured data are fragmented, and the relationship between these entities is flat, lacking hierarchy and logic. Therefore, the entities are first analyzed for similarity to avoid overlapping errors, and the semantic vectors are reconstructed based on the similarity analysis results, that is, the relationship between the entities is reconstructed. Then, the original energy industry semantic data is used to match the entities, and a three-dimensional entity set is re-established, so that the entities and the corresponding attribute information of the entities have a three-dimensional sense of hierarchy and corresponding logic.

[0035] S3. Perform knowledge reasoning based on the three-dimensional entity set to obtain the knowledge ontology, and perform quality assessment on the knowledge ontology to obtain the confidence level.

[0036] In one embodiment, the specific process of performing knowledge reasoning based on the three-dimensional entity set to obtain the knowledge ontology is as follows: S311, extracting entity parallel relationships and entity hierarchical relationships based on the three-dimensional entity set, and performing parallel similarity analysis based on the entity parallel relationships to obtain entity parallel features; S312: Based on the entity parallel characteristics and entity hierarchical relationships and the corresponding energy industry material supply rules, establish data definitions and structure definitions for the materials corresponding to the entities in the three-dimensional entity collection; S313: Construct a knowledge ontology based on the data definition and structure definition.

[0037] In this embodiment, although there is already a three-dimensional entity set containing three-dimensional logic, there are some overlaps among these logical relationships. Therefore, similarity analysis is performed on the parallel relationships of entities to further reduce unnecessary entity data, and data definitions and structure definitions of materials corresponding to entities in the three-dimensional entity set are established according to the corresponding energy industry material supply rules, and various business scenarios of the entity corresponding to the material supply chain are clarified, and the association relationship between the entity and the external knowledge base and relational database is obtained.

[0038] In one embodiment, the specific process of performing quality assessment on the knowledge ontology to obtain confidence is as follows: S321. Perform consistency analysis based on semantic entities and semantic vectors corresponding to the ontology in the knowledge ontology to obtain an ontology consistency degree, and perform overlap analysis based on the knowledge ontology and energy industry semantic data to obtain an ontology overlap degree; S322: Calculate the confidence of the ontology in the knowledge ontology according to the ontology consistency and ontology overlap.

[0039] In this embodiment, although a semi-manual editing method combining ontology editing software and automation is used in the process of knowledge reasoning to discover and improve the three-dimensional entity set to obtain the knowledge ontology, the quality of the ontology in the knowledge ontology cannot be guaranteed. Therefore, a statistical method is used to count the consistency of entities, entity attributes and relationships related to entities in the ontology in different data sources or different representations, and the overlap of the corresponding ontology is obtained according to the ratio of the intersection and union of entities, entity attributes and relationships related to entities. Finally, a weighted calculation is performed based on the consistency and overlap to obtain the confidence of the ontology, where the weights of the consistency and overlap are set according to the corresponding energy industry semantic data.

[0040] S4. Building a semantic library based on the confidence level and knowledge ontology, and obtaining real-time statements from staff members; S411, screening ontologies in the knowledge ontology based on the confidence to obtain ontologies with qualified confidence, and updating the knowledge ontology according to the ontologies with qualified confidence; S412: Establish a vocabulary based on the semantic entities corresponding to the ontology in the updated knowledge ontology, and establish a semantic library based on the vocabulary and the semantic vectors corresponding to the semantic entities in the vocabulary.

[0041] In this solution, the semantic library at least includes material supply chain business terms, business term relationships, and business domain management, and the semantic library can quickly find corresponding material supply chain business terms based on material slang according to business term relationships.

[0042] S5, analyzing the real-time sentence according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentence; S511, inputting the real-time sentence into the target recognition model for analysis to obtain real-time semantic entities and real-time semantic vectors, and establishing a knowledge graph based on the semantic library; S512: Perform similarity analysis between the real-time semantic entity and the semantic entities in the semantic library, and fill the real-time semantic entity and the real-time semantic vector into the knowledge graph based on the similarity analysis result.

[0043] On the other hand, a technical solution also provided in an embodiment of the present invention is an electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the electronic device executes any one of the above methods.

[0044] On the other hand, a technical solution also provided in an embodiment of the present invention is a storage medium, which stores computer-executable instructions. When the computer-executable instructions are loaded and executed by a processor, the computer executes any one of the above methods.

[0045] This embodiment has at least the following substantial effects: This embodiment introduces a bidirectional long short-term memory network into the entity recognition model to improve its bidirectional recognition capability and enhance its contextual application capability. A soft word list is introduced into the entity recognition model to establish annotation specifications, thereby improving the entity recognition model's ability to process complex text data. This results in a target recognition model with strong sequence information capture capabilities. Semantic entities and semantic vectors are extracted from energy industry semantic data to more deeply reflect the essential characteristics of the data. Entity linking technology is used to link fragmented semantic entities and semantic vectors to obtain a three-dimensional entity set. Knowledge reasoning is performed on the three-dimensional entity set to further explore potential relationships and information hidden in the data to obtain a more complete knowledge ontology. The knowledge ontology is then quality assessed, and reliable ontologies within the knowledge ontology are identified using the ontology's confidence level. Finally, based on the semantic library constructed from the reliable ontology and the target recognition model, real-time statements of staff members are analyzed to obtain the names of materials corresponding to the colloquial terms in the real-time statements. This effectively improves the efficiency of identifying the materials of interest to staff members and eliminates the need to repeatedly confirm the same colloquial term, significantly improving the efficiency of reporting requirements corresponding to the colloquial term.

[0046] The above specific embodiments are preferred embodiments of the present invention and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the scope of protection of the present invention.

Claims

1. A method for identifying colloquialisms in the energy industry, characterized by: The method comprises the following steps: S1. Obtain the semantic data of the energy industry, and improve the entity recognition model based on the bidirectional long short-term memory network and the soft word list to obtain a target recognition model; S2. Extracting semantic entities and semantic vectors based on the semantic data and the target recognition model, and performing entity linking based on the semantic entities and semantic vectors to construct a three-dimensional entity set; S3. Perform knowledge reasoning based on the three-dimensional entity set to obtain the knowledge ontology, and perform quality assessment on the knowledge ontology to obtain the confidence level; S4. Building a semantic library based on the confidence level and knowledge ontology, and obtaining real-time statements from staff members; S5. Analyze the real-time sentence according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentence.

2. The method for identifying colloquialisms in the energy industry and materials field according to claim 1 is characterized in that: In S1, the specific process of improving the entity recognition model based on the bidirectional long short-term memory network and the soft word list to obtain the target recognition model is as follows: S111, extracting an embedding module and a recognition module based on the structure of the entity recognition model; S112, improving the recognition module based on a bidirectional long short-term memory network to obtain a target recognition module, and adding a calibration entity and a calibration object to the embedding module based on a soft word list to obtain a logical embedding module; S113. Integrate the target recognition module and the logic embedding module based on the structure of the entity recognition model to obtain a target recognition model.

3. The method for identifying colloquialisms in the energy industry and materials field according to claim 1 is characterized in that: In S2, the specific process of extracting semantic entities and semantic vectors based on semantic data and target recognition model is as follows: S211, inputting the semantic data into a target recognition model for entity calibration to obtain entities and objects, and classifying the entities, objects and linear classifiers to obtain entity relationships and entity attributes; S212: Arrange the entities, objects, entity relationships and entity attributes to obtain semantic entities and semantic vectors.

4. The method for identifying colloquialisms in the energy industry and materials field according to claim 3 is characterized in that: In S2, the specific process of constructing a three-dimensional entity set by performing entity linking based on the semantic entities and semantic vectors is as follows: S221. Segment the semantic entities based on a natural language processing algorithm and organize them to obtain an entity set; Simultaneously, performing similarity analysis on entities in the entity set, and reconstructing semantic vectors based on the results of the similarity analysis; S222. Extract candidate entities from the energy industry semantic data based on the entity set, and pair the entities in the entity set with the candidate entities based on the semantic vectors; S223: Construct a three-dimensional entity set based on the pairing result, the entity, and the candidate entity.

5. The method for identifying colloquialisms in the energy industry and materials field according to claim 1 is characterized in that: In S3, the specific process of performing knowledge reasoning based on the three-dimensional entity set to obtain the knowledge ontology is as follows: S311, extracting entity parallel relationships and entity hierarchical relationships based on the three-dimensional entity set, and performing parallel similarity analysis based on the entity parallel relationships to obtain entity parallel features; S312: Based on the entity parallel characteristics and entity hierarchical relationships and the corresponding energy industry material supply rules, establish data definitions and structure definitions for the materials corresponding to the entities in the three-dimensional entity collection; S313: Construct a knowledge ontology based on the data definition and structure definition.

6. The method for identifying colloquialisms in the energy industry and materials field according to claim 5 is characterized in that: In S3, the specific process of performing quality assessment on the knowledge ontology to obtain the confidence level is as follows: S321. Perform consistency analysis based on semantic entities and semantic vectors corresponding to the ontology in the knowledge ontology to obtain an ontology consistency degree, and perform overlap analysis based on the knowledge ontology and energy industry semantic data to obtain an ontology overlap degree; S322: Calculate the confidence of the ontology in the knowledge ontology according to the ontology consistency and ontology overlap.

7. The method for identifying colloquialisms in the energy industry and materials field according to claim 1 is characterized in that: In S4, the specific process of constructing the semantic library based on the confidence level and knowledge ontology is as follows: S411, screening ontologies in the knowledge ontology based on the confidence to obtain ontologies with qualified confidence, and updating the knowledge ontology according to the ontologies with qualified confidence; S412: Establish a vocabulary based on the semantic entities corresponding to the ontology in the updated knowledge ontology, and establish a semantic library based on the vocabulary and the semantic vectors corresponding to the semantic entities in the vocabulary.

8. The method for identifying colloquialisms in the energy industry and materials field according to claim 1 is characterized in that: In S5, the specific process of analyzing the real-time sentence according to the semantic library and the target recognition model to obtain the material names corresponding to the colloquialisms in the real-time sentence is as follows: S511, inputting the real-time sentence into the target recognition model for analysis to obtain real-time semantic entities and real-time semantic vectors, and establishing a knowledge graph based on the semantic library; S512: Perform similarity analysis between the real-time semantic entity and the semantic entities in the semantic library, and fill the real-time semantic entity and the real-time semantic vector into the knowledge graph based on the similarity analysis result.

9. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of a colloquialism semantic recognition method based on the energy industry material field as described in any one of claims 1 to 8 are implemented.

10. A storage medium, characterized in that: The storage medium stores computer-executable instructions, which, when loaded and executed by the processor, implement the steps of a colloquialism semantic recognition method based on the energy industry material field as described in any one of claims 1 to 8.

Images