Transmission solution generation method and system based on transmission knowledge graph

Abstract

The application discloses a transmission solution generation method and system based on a transmission knowledge graph. The method comprises the following steps: collecting a transmission solution, and constructing a transmission knowledge information corpus based on the transmission solution; extracting core elements of the transmission solution, constructing a pattern layer based on the core elements, constructing a data layer by using entity information, relationship information and attribute information in the transmission knowledge information corpus, establishing an associated mapping between the data layer and the pattern layer, and generating a transmission knowledge graph based on the associated mapping; collecting transmission solution expected factors, identifying a solution demand based on the solution expected factors, determining entity logical association based on the solution expected factors and the transmission knowledge graph, and generating the transmission solution according to the solution demand and the entity logical association. The method can simplify the generation process of the transmission solution based on the transmission knowledge graph.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, specifically to a method and system for generating transmission solutions based on a transmission knowledge graph. Background Technology

[0002] A knowledge graph is a semantic network that describes conceptual entities, events, and the relationships between them in the objective world. Compared to traditional databases that simply store data, knowledge graphs use entity concepts as nodes and relationships as edges. Through data mining, knowledge reasoning, and other techniques, they extract potential semantic connections between knowledge and present them intuitively in the form of triples. Knowledge graphs transform massive amounts of unstructured or semi-structured knowledge into standardized, reliable structured data. Through data processing and reasoning, they uncover semantic connections, forming a highly interconnected semantic network that supports data mining and intelligent services. The construction of knowledge graphs includes key technologies such as knowledge modeling, knowledge extraction, knowledge representation, knowledge fusion, and knowledge reasoning.

[0003] Transmission solutions are tailored to specific applications and scenarios. Based on factors such as power transmission requirements, environmental conditions, application duration, application intensity, and space conditions, the transmission method is determined, and then integrated solutions are provided, including electric motors, reducers, various accessories and their performance parameters, selection and matching, assembly methods, lubrication and sealing methods, etc.

[0004] The generation of transmission solutions requires specialized skills and knowledge to design transmission methods, select equipment, preset performance parameters, and calculate and verify parameters such as power, torque, and transmission ratio. Generally, a transmission solution needs to be designed separately for each specific case, which requires a high level of expertise from the participants and consumes a lot of time.

[0005] Therefore, how to simplify the generation process of transmission solutions based on transmission knowledge graphs, improve generation efficiency, and reduce the degree of specialization dependence is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] In view of the above problems, the purpose of this invention is to solve the problem that the generation of transmission solutions requires specialized skills and knowledge, has high requirements for the professional level of the participants, and consumes a lot of time.

[0007] This invention provides a method for generating transmission solutions based on a transmission knowledge graph, including:

[0008] Collect transmission solutions and build a transmission knowledge information corpus based on the transmission solutions;

[0009] The core elements of the transmission solution are extracted, a pattern layer is constructed based on the core elements, and a data layer is constructed using entity information, relation information and attribute information in the transmission knowledge information corpus. An association mapping between the data layer and the pattern layer is established, and a transmission knowledge graph is generated based on the association mapping.

[0010] Collect expected factors for transmission solutions, identify solution requirements based on these expected factors, determine entity logical associations based on these expected factors and the transmission knowledge graph, and generate the transmission solution based on the solution requirements and the entity logical associations.

[0011] In one embodiment, the acquisition of transmission solutions and the construction of a transmission knowledge information corpus based on the transmission solutions include:

[0012] The original description files of the transmission solution are collected, and the original description files are preprocessed to generate transmission knowledge corpus.

[0013] According to predetermined rules, the transmission knowledge corpus is tagged and labeled to generate corpus annotations. The transmission knowledge information corpus is then constructed based on the transmission knowledge corpus and the corpus annotations.

[0014] In one embodiment, the process of extracting the core elements of the transmission solution, constructing a pattern layer based on the core elements, defining a triplet framework using entity information, relation information, and attribute information from the transmission knowledge information corpus, generating a data layer, establishing an association mapping between the data layer and the pattern layer, and generating a transmission knowledge graph based on the association mapping includes:

[0015] Extract the core elements of the transmission solution and construct the triplet framework based on the core elements to generate the pattern layer; the core elements of the transmission solution include, but are not limited to: the motor, reducer, various accessories, transmission requirements, environmental conditions, spatial conditions, transmission method, performance parameters, assembly relationship, lubrication and sealing, application duration, and application intensity included in the transmission solution; furthermore, analyze the entities, relationships, attributes, and logical associations of these core elements in the existing transmission solution to construct a triplet framework of "entity-relationship-entity" or "entity-attribute-attribute value";

[0016] Based on the corpus annotation, the transmission knowledge corpus is extracted, and knowledge extraction is used to obtain entity information, relation information and attribute information in the transmission knowledge corpus;

[0017] The triplet framework is filled based on the entity information, the relation information, and the attribute information, and the filled triplet framework is associated based on semantic relations.

[0018] Entity fusion is performed using a knowledge fusion algorithm to generate data entities, and the filled triplet framework is associated with the data entities to generate the data layer;

[0019] Establish the association mapping between the data layer and the pattern layer, and combine the data layer and the pattern layer based on the association mapping to generate a transmission knowledge graph.

[0020] In one embodiment, the step of using a knowledge fusion algorithm to perform entity fusion and generate data entities includes:

[0021] The entity's ontology concept is compared with its hierarchy. Ontology alignment is determined based on the comparison results. Semantic similarity of the entity is calculated based on the determination results, and a similarity threshold is set. Entity fusion is performed based on the semantic similarity to generate data entities.

[0022] In one embodiment, the process of collecting expected factors for the transmission solution, identifying solution requirements based on these expected factors, determining entity logical associations based on the expected factors and the transmission knowledge graph, and generating the transmission solution according to the solution requirements and the entity logical associations includes:

[0023] Collect the expected factors of the transmission solution, and preprocess the expected factors of the transmission solution to generate keywords;

[0024] Based on the keywords, the solution requirements are identified, and a requirement classifier is used to classify the solution requirements.

[0025] Based on the keywords, entities in the transmission knowledge graph are identified. Taking the entities as core nodes, associated entities and attribute values ​​are determined according to entity relationships. Entity logical associations are generated based on the associated entities and attribute values.

[0026] Based on the solution requirements and the logical association with the entities, the transmission solutions are filtered using the entity relationships, and then the transmission solutions are sent to the user.

[0027] Secondly, the present invention also provides a transmission solution generation system based on a transmission knowledge graph, comprising:

[0028] A corpus construction module for transmission knowledge information is used to collect transmission solutions and construct a corpus of transmission knowledge information based on the transmission solutions.

[0029] The transmission knowledge graph generation module is used to extract the core elements of the transmission solution, construct a pattern layer based on the core elements, construct a data layer using entity information, relationship information and attribute information in the transmission knowledge information corpus, establish an association mapping between the data layer and the pattern layer, and generate a transmission knowledge graph based on the association mapping.

[0030] The transmission solution generation module is used to collect expected factors of transmission solutions, identify solution requirements based on the expected factors, determine entity logical associations based on the expected factors and the transmission knowledge graph, and generate the transmission solution according to the solution requirements and the entity logical associations.

[0031] In one embodiment, the transmission knowledge information corpus construction module includes:

[0032] The transmission knowledge corpus generation unit is used to collect the original description files of transmission solutions and perform text preprocessing on the original description files to generate transmission knowledge corpus.

[0033] The transmission knowledge information corpus construction unit is used to tag the transmission knowledge corpus according to predetermined rules, generate corpus annotations, and construct the transmission knowledge information corpus based on the transmission knowledge corpus and the corpus annotations.

[0034] In one embodiment, the transmission knowledge graph generation module includes:

[0035] The pattern layer generation unit is used to extract the core elements of the transmission solution and construct the triplet framework based on the core elements to generate the pattern layer. The core elements of the transmission solution include, but are not limited to: the motor, reducer, various accessories, transmission requirements, environmental conditions, spatial conditions, transmission method, performance parameters, assembly relationship, lubrication and sealing, application duration, and application intensity of the transmission solution. Furthermore, based on the core elements of the transmission solution, the unit analyzes the entities, relationships, attributes, and logical associations of these core elements in existing transmission solutions to construct a triplet framework of "entity-relationship-entity" or "entity-attribute-attribute value".

[0036] The knowledge extraction unit is used to extract the transmission knowledge corpus based on the corpus annotation, and to obtain entity information, relation information and attribute information in the transmission knowledge corpus by knowledge extraction;

[0037] The triplet frame filling unit is used to fill the triplet frame based on the entity information, the relation information and the attribute information, and associate the filled triplet frame based on semantic relations.

[0038] The data layer generation unit is used to perform entity fusion using a knowledge fusion algorithm, generate data entities, and associate the filled triplet framework with the data entities to generate the data layer;

[0039] The association mapping unit is used to establish the association mapping between the data layer and the pattern layer, and to combine the data layer and the pattern layer based on the association mapping to generate a transmission knowledge graph.

[0040] In one embodiment, the step of using a knowledge fusion algorithm to perform entity fusion and generate data entities includes:

[0041] The entity's ontology concept is compared with its hierarchy. Ontology alignment is determined based on the comparison results. Semantic similarity of the entity is calculated based on the determination results, and a similarity threshold is set. Entity fusion is performed based on the semantic similarity to generate data entities.

[0042] In one embodiment, the transmission solution generation module includes:

[0043] The keyword generation unit is used to collect expected factors of the transmission solution, preprocess the expected factors of the transmission solution, and generate keywords;

[0044] A solution requirement identification unit is used to identify solution requirements based on the keywords and to classify the solution requirements using a requirement classifier.

[0045] An entity logical association generation unit is used to determine entities in the transmission knowledge graph based on the keywords, take the entities as core nodes, determine associated entities and attribute values ​​according to entity relationships, and generate entity logical associations based on the associated entities and attribute values.

[0046] A transmission solution generation unit is used to logically associate the transmission solutions with the entity based on the solution requirements, filter the transmission solutions using the entity relationships, and send the transmission solutions to the user.

[0047] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0048] The present invention provides a method for generating transmission solutions based on a transmission knowledge graph. This method generates a transmission knowledge graph based on existing transmission solutions, defines entity, relationship, and attribute information for each transmission solution, generates a triplet framework, and uses the triplet framework as the basic building block to associate data entities. This realizes the process of constructing a transmission knowledge graph for the core elements of the transmission solution. Based on the logical relationships of entities in the transmission knowledge graph, it enables the rapid matching of corresponding transmission solutions to user expectations, simplifies the generation process of transmission solutions, improves generation efficiency, and reduces the degree of specialization dependence.

[0049] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0050] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0051] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 A flowchart illustrating the method for generating a transmission solution based on a transmission knowledge graph, as provided in an embodiment of the present invention;

[0053] Figure 2 This is a flowchart of step S101 provided in an embodiment of the present invention;

[0054] Figure 3 This is a flowchart of step S102 provided in an embodiment of the present invention;

[0055] Figure 4 A schematic diagram of the BERT-BiLSTM-CRF model provided in an embodiment of the present invention;

[0056] Figure 5 This is a flowchart of step S103 provided in an embodiment of the present invention;

[0057] Figure 6 A block diagram of a transmission solution generation system based on transmission knowledge graph provided in an embodiment of the present invention. Detailed Implementation

[0058] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0059] Reference Figure 1 As shown, the transmission solution generation method based on transmission knowledge graph provided in this embodiment of the invention includes steps S101 to S103.

[0060] S101. Collect transmission solutions and build a transmission knowledge information corpus based on the transmission solutions.

[0061] Specifically, the original description files of existing transmission solutions collected and summarized are preprocessed to generate structured transmission knowledge corpus. According to predetermined rules, the preprocessed structured transmission knowledge corpus is tagged to generate corpus annotations. Based on the transmission knowledge corpus and the corpus annotations, the transmission knowledge information corpus is constructed.

[0062] S102. Extract the core elements of the transmission solution, construct a pattern layer based on the core elements, and construct a data layer using entity information, relationship information, and attribute information in the transmission knowledge information corpus. Establish an association mapping between the data layer and the pattern layer, and generate a transmission knowledge graph based on the association mapping. The core elements of the transmission solution include, but are not limited to: the motor, reducer, various accessories, transmission requirements, environmental conditions, spatial conditions, transmission method, performance parameters, assembly relationship, lubrication and sealing, application duration, and application intensity included in the transmission solution.

[0063] Specifically, the transmission knowledge graph includes a schema layer and a data layer. The schema layer defines its conceptual ontology, attribute types, relationships, and structural hierarchy. The data layer stores entity data and uses triples as the basic building blocks to associate various types of data entities, thus constructing the entity network of the transmission knowledge graph.

[0064] Furthermore, the schema layer is constructed in a top-down manner. Based on the core elements of the transmission solution, it analyzes the entities, relationships, attributes, and logical connections of these core elements in the existing transmission solution, and constructs a triple framework of "entity-relationship-entity" or "entity-attribute-attribute value". It also sets the triple framework of "entity-relationship-entity" or "entity-attribute-attribute value" for the transmission knowledge graph, thereby providing a conceptual basis for the construction of the data layer, defining the hierarchical relationships, semantic relationships, and attribute relationships between various ontologies, and thus defining the schema layer.

[0065] Furthermore, the data layer is constructed in a bottom-up manner, starting from the transmission knowledge information corpus. Using knowledge extraction, based on the corpus annotations in the transmission knowledge information corpus, information such as entities, relations, and attributes in the transmission knowledge corpus is obtained and filled into triples of the type "entity-relation-entity" and "entity-attribute-attribute value". The triples are interconnected based on semantic relations.

[0066] S103. Collect expected factors of the transmission solution, identify solution requirements based on the expected factors of the solution, determine the entity logical association based on the expected factors of the solution and the transmission knowledge graph, and generate the transmission solution according to the solution requirements and the entity logical association.

[0067] Specifically, users input the expected application products and scenarios of the transmission solution, as well as expected factors such as power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions. Based on these expected factors, the solution requirements are identified and a transmission solution is generated.

[0068] In this embodiment, a transmission knowledge graph is generated based on existing transmission solutions. Entities, relationships, and attribute information are defined for each transmission solution, and a triplet framework is generated. Using the triplet framework as the basic building block, data entities are associated, realizing the process of constructing a transmission knowledge graph for the core elements of the transmission solution. Based on the logical relationships of entities in this transmission knowledge graph, the corresponding transmission solution can be quickly matched to the user's expected factors, simplifying the generation process of the transmission solution, improving generation efficiency, and reducing the degree of specialization dependence.

[0069] In one embodiment, refer to Figure 2 As shown, the step S101 above, which involves collecting transmission solutions and constructing a transmission knowledge information corpus based on those solutions, includes:

[0070] S1011. Collect the original description file of the transmission solution, and perform text preprocessing on the original description file to generate transmission knowledge corpus.

[0071] Specifically, text preprocessing includes standardizing the text encoding format of the original description file, followed by corpus deduplication and corpus cleaning. Corpus deduplication involves calculating the similarity between corpora and deleting duplicate content to prevent multiple repetitions of a single corpus. Corpus cleaning involves cleaning the text of the deduplicated corpus by matching the deduplicated corpus content with a stop word dictionary to filter out irrelevant or meaningless corpus content.

[0072] S1012. According to predetermined rules, the transmission knowledge corpus is tagged and labeled to generate corpus annotations. The transmission knowledge information corpus is constructed based on the transmission knowledge corpus and the corpus annotations.

[0073] Specifically, according to predetermined rules, the transmission knowledge corpus is tagged, including: for the corpus describing application products and scenarios, each corpus is tagged with power transmission requirements, environmental conditions, application duration, application intensity, spatial conditions, etc.; for the corpus describing transmission solutions, tags such as transmission method, motor selection, reducer selection, accessory selection, motor / reducer / accessory performance parameters, assembly method, lubrication and sealing method, etc. can be added.

[0074] In one embodiment, refer to Figure 3 As shown, step S102 above, which involves extracting the core elements of the transmission solution, constructing a pattern layer based on these core elements, defining a triplet framework using entity information, relational information, and attribute information from the transmission knowledge information corpus, generating a data layer, establishing an association mapping between the data layer and the pattern layer, and generating a transmission knowledge graph based on the association mapping, includes:

[0075] S1021. Extract the core elements of the transmission solution, and construct the triplet framework based on the core elements to generate the mode layer.

[0076] Specifically, the core elements of the transmission solution include, but are not limited to: the motor, reducer, various accessories, transmission requirements, environmental conditions, space conditions, transmission method, performance parameters, assembly relationship, lubrication and sealing, application duration, and application intensity.

[0077] Specifically, for transmission solutions, the motor, reducer, various accessories, transmission requirements, environmental conditions, and spatial conditions included in the transmission solution can be defined as entities. The transmission method, assembly relationships, lubrication and sealing structural relationships can be defined as relationships. Performance parameters, application duration, and application intensity included in the transmission solution can be defined as attributes, with their specific values ​​serving as attribute values. Furthermore, based on the core elements of the transmission solution—transmission method, performance parameters, assembly relationships, lubrication and sealing, application duration, and application intensity—the entities, relationships, attributes, and their logical connections within existing transmission solutions can be analyzed to construct a "entity-relationship-entity" or "entity-attribute-attribute value" triplet framework. For example, "motor-assembly-reducer" and "reducer-lubrication-lubricant" are "entity-relationship-entity" triplet frameworks, while "reducer-torque-torque value" is an "entity-attribute-attribute value" triplet framework.

[0078] S1022. Extract the transmission knowledge corpus based on the corpus annotation, and use knowledge extraction to obtain entity information, relation information and attribute information in the transmission knowledge corpus.

[0079] Specifically, knowledge extraction includes entity extraction, relation extraction, and attribute extraction. Entity extraction employs named entity recognition (NER) algorithms. Based on prior training, natural language processing (NLP) techniques are used to automatically collect entity content from the corpus. The collected data is then added to the model for further training to improve accuracy. Finally, an entity database is built by organizing and summarizing the entity information in the text. Relation extraction algorithms essentially obtain specific descriptions of the logical relationships between entities or between entities and attributes in the text corpus. Through this, they extract potential meanings and construct semantic connections between entities. Current relation extraction algorithms are mainly divided into: manual annotation combined with entity extraction, machine learning-based semantic relation recognition, deep learning-based relation recognition, and joint extraction algorithms combined with named entity recognition. Attribute extraction obtains the attribute information of entities, enriching the content of the entities.

[0080] Furthermore, since named entity recognition algorithms are highly efficient in labeling, but require a large amount of data for training and have relatively low accuracy, the choice of entity labeling algorithm often depends on the data situation and work requirements. Therefore, the BERT-BiLSTM-CRF model is used for entity extraction. The BERT-BiLSTM-CRF model is a named entity recognition model composed of the BERT pre-trained language model and the bidirectional long short-term memory network-conditional random field model (BiLSTM-CRF).

[0081] Among them, reference Figure 4As shown, the BERT-BiLSTM-CRF model consists of three layers. The first layer is the BERT layer, which uses a large amount of general Chinese corpus and extensive computation to train the BERT general language model, extracting word vector features of transmission knowledge information to obtain low-dimensional word vectors. The second layer is the BiLSTM layer, which, based on the understanding of word vector features of transmission knowledge information from the pre-trained layer, trains the bidirectional long short-term memory network using a large amount of annotated transmission knowledge corpus. Based on the model training results, it uses contextual semantic information to infer and label entity sequences in the corpus, and further filters entity types by setting weights through an attention mechanism. The third layer is the CRF layer, which, based on the entity sequences output by the BiLSTM layer, uses a probabilistic model to predict and output the optimal expression for sequence labels, thereby achieving automatic sequence labeling of the corpus and completing named entity recognition.

[0082] S1023. Fill the triplet framework based on the entity information, the relationship information and the attribute information, and associate the filled triplet framework based on semantic relationships.

[0083] S1024. Entity fusion is performed using a knowledge fusion algorithm to generate data entities, and the filled triplet framework is associated with the data entities to generate the data layer.

[0084] Specifically, knowledge fusion is divided into two parts: ontology alignment and entity matching. Ontology alignment refers to comparing and confirming the ontology concepts and levels of the entities to be confirmed. If the similarity of levels and concepts is high, the ontology is considered to be aligned. Entity matching refers to confirming the degree of similarity between two entities in terms of entity names, entity relationships, and attributes.

[0085] Furthermore, the step of using a knowledge fusion algorithm to perform entity fusion and generate data entities includes: comparing the entity's ontology concept with its hierarchy, determining ontology alignment based on the comparison result, calculating the semantic similarity of the entity based on the determination result, setting a similarity threshold, and performing entity fusion based on the semantic similarity to generate data entities.

[0086] Furthermore, the semantic similarity between two entities is calculated in terms of entity name, hierarchical category, and attributes. A comprehensive similarity vector is derived by combining the similarity of individual attributes to determine whether entity redundancy exists. The formula for calculating the similarity of individual attributes is as follows:

[0087]

[0088] In the above formula, S AmBm This represents the similarity of entities A and B based on a certain attribute m. A mLet m represent a certain attribute of entity A. B m Let m represent a certain attribute of entity B. a i Represents attributes A m Word frequency of segmentation b i Represents attributes B m The word frequency of the segmented words, where n represents the number of segments.

[0089] Furthermore, A m , B m There are n word segments in the semantic space, which are determined by statistical attributes. A m , B m Word frequency of each segment a i , b i The word frequency vector (i.e., the comprehensive similarity vector) is constructed, and the similarity between two sentences is determined by calculating the cosine value of the vector.

[0090] Furthermore, by calculating the ratio of semantic similarity between entities A and B on each attribute to the number of attributes, the similarity ratio of entities A and B is generated, and the calculation formula is as follows:

[0091]

[0092] In the above formula, S A,B This represents the similarity ratio between entities A and B, where 's' represents the number of attributes. S A,B The value ranges from 0 to 1, with the value closer to 1 indicating a higher semantic similarity between the two entities.

[0093] S1025. Establish the association mapping between the data layer and the pattern layer, and combine the data layer and the pattern layer based on the association mapping to generate a transmission knowledge graph.

[0094] Specifically, by mapping the ontology and its relationships in the schema layer to numerous data entities in the data layer, a mapping is established between the data layer and the schema layer. By combining the data layer and the schema layer, the transmission knowledge graph can be constructed.

[0095] In one embodiment, refer to Figure 5As shown, step S103 above, which involves collecting expected factors for the transmission solution, identifying solution requirements based on these expected factors, determining entity logical associations based on the expected factors and the transmission knowledge graph, and generating the transmission solution according to the solution requirements and the entity logical associations, includes:

[0096] S1031. Collect the expected factors of the transmission solution, and preprocess the expected factors of the transmission solution to generate keywords.

[0097] Specifically, for the expected factors of the transmission solution input by the user, if the user input is in natural language format, preprocessing of the expected factors of the user input is performed, including word segmentation, text disambiguation, and keyword recognition. Among them, word segmentation refers to breaking down the user input question into multiple groups of words for representation; text disambiguation refers to removing meaningless components in the word segmentation results and eliminating ambiguity in the text to prevent polysemy from affecting the accuracy of applicability recognition. Text disambiguation mainly includes stop word removal and part-of-speech tagging; keyword recognition is performed on the disambiguated text to filter out the core entities, relationships, or descriptions of the user's question content.

[0098] Furthermore, a TF-IDF (term frequency–inverse document frequency) weighted model is used for keyword identification and extraction. The steps include: inputting the text into the TF-IDF weighted model to determine the words. t i Word frequency in text tf i The calculation formula is as follows:

[0099]

[0100] in, n ij Words t i In the text d j The number of times it appears in Indicated in text d j The sum of the number of occurrences of all words in the text.

[0101] Calculate the frequency of reverse files IDF i The calculation formula is as follows:

[0102]

[0103] In the above formula, |D| represents the total number of texts in the corpus. Indicates words containingt i The number of texts.

[0104] Word frequency calculation based on term frequency and reverse document frequency t i TF-IDF value tf idfi The calculation formula is as follows:

[0105]

[0106] As can be seen from the above formula, the high frequency of a word in a specific file and the low frequency of that word in the entire file set can generate a high-weight TF-IDF value. Then, the TF-IDF values ​​of each word in the text are counted, and the words corresponding to the TF-IDF values ​​that are combined with the preset weight thresholds are selected as keywords.

[0107] S1032. Identify solution requirements based on the keywords, and classify the solution requirements using a requirement classifier.

[0108] Specifically, the solution requirements in the expected factors of the transmission solution input by the user are analyzed, so as to obtain a solution that matches the solution requirements in the subsequent matching process.

[0109] S1033. Based on the keywords, determine the entities in the transmission knowledge graph, take the entities as core nodes, determine the associated entities and attribute values ​​according to the entity relationships, and generate entity logical associations based on the associated entities and attribute values.

[0110] S1034. Based on the solution requirements and the logical association with the entity, filter the transmission solutions using the entity relationships, and send the transmission solutions to the user.

[0111] Based on the same inventive concept, this invention also provides a transmission solution generation system based on transmission knowledge graph. Since the principle of the problem solved by this system is similar to the aforementioned transmission solution generation method based on transmission knowledge graph, the implementation of this system can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0112] This invention provides a transmission solution generation system based on a transmission knowledge graph, referring to... Figure 6 As shown, it includes:

[0113] The transmission knowledge information corpus construction module 61 is used to collect transmission solutions and construct a transmission knowledge information corpus based on the transmission solutions.

[0114] Specifically, the original description files of existing transmission solutions collected and summarized are preprocessed to generate structured transmission knowledge corpus. According to predetermined rules, the preprocessed structured transmission knowledge corpus is tagged to generate corpus annotations. Based on the transmission knowledge corpus and the corpus annotations, the transmission knowledge information corpus is constructed.

[0115] The transmission knowledge graph generation module 62 is used to extract the core elements of the transmission solution, construct a pattern layer based on the core elements, construct a data layer using entity information, relationship information and attribute information in the transmission knowledge information corpus, establish an association mapping between the data layer and the pattern layer, and generate a transmission knowledge graph based on the association mapping.

[0116] Specifically, the transmission knowledge graph includes a schema layer and a data layer. The schema layer defines its conceptual ontology, attribute types, relationships, and structural hierarchy. The data layer stores entity data and uses triples as the basic building blocks to associate various types of data entities, thus constructing the entity network of the transmission knowledge graph.

[0117] Furthermore, the schema layer is constructed in a top-down manner, setting up a triple framework of "entity-relationship-entity" or "entity-attribute-attribute value" for the transmission knowledge graph based on the core elements of the transmission solution. This provides a conceptual basis for the construction of the data layer, defining the hierarchical relationships, semantic relationships, and attribute relationships between various ontologies, thereby defining the schema layer.

[0118] Furthermore, the data layer is constructed in a bottom-up manner, starting from the transmission knowledge information corpus. Using knowledge extraction, based on the corpus annotations in the transmission knowledge information corpus, information such as entities, relations, and attributes in the transmission knowledge corpus is obtained and filled into triples of the type "entity-relation-entity" and "entity-attribute-attribute value". The triples are interconnected based on semantic relations.

[0119] The transmission solution generation module 63 is used to collect expected factors of transmission solutions, identify solution requirements based on the expected factors, determine entity logical associations based on the expected factors and the transmission knowledge graph, and generate the transmission solution according to the solution requirements and the entity logical associations.

[0120] Specifically, users input the expected application products and scenarios of the transmission solution, as well as expected factors such as power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions. Based on these expected factors, the solution requirements are identified and a transmission solution is generated.

[0121] In one embodiment, the transmission knowledge information corpus construction module 61 includes:

[0122] The transmission knowledge corpus generation unit 611 is used to collect the original description file of the transmission solution and perform text preprocessing on the original description file to generate the transmission knowledge corpus.

[0123] Specifically, text preprocessing includes standardizing the text encoding format of the original description file, followed by corpus deduplication and corpus cleaning. Corpus deduplication involves calculating the similarity between corpora and deleting duplicate content to prevent multiple repetitions of a single corpus. Corpus cleaning involves cleaning the text of the deduplicated corpus by matching the deduplicated corpus content with a stop word dictionary to filter out irrelevant or meaningless corpus content.

[0124] The transmission knowledge information corpus construction unit 612 is used to tag the transmission knowledge corpus according to predetermined rules, generate corpus annotations, and construct the transmission knowledge information corpus based on the transmission knowledge corpus and the corpus annotations.

[0125] Specifically, according to predetermined rules, the transmission knowledge corpus is tagged, including: for the corpus describing application products and scenarios, each corpus is tagged with power transmission requirements, environmental conditions, application duration, application intensity, spatial conditions, etc.; for the corpus describing transmission solutions, tags such as transmission method, motor selection, reducer selection, accessory selection, motor / reducer / accessory performance parameters, assembly method, lubrication and sealing method, etc. can be added.

[0126] In one embodiment, the transmission knowledge graph generation module 62 includes:

[0127] The pattern layer generation unit 621 is used to extract the core elements of the transmission solution and construct the triplet framework based on the core elements to generate the pattern layer.

[0128] Specifically, for transmission solutions, motors, reducers, various accessories, transmission requirements, environmental conditions, and spatial conditions can be defined as entities. Then, based on the transmission method, performance parameters, assembly relationships, lubrication and sealing, application duration, and application intensity of the transmission solution, a triplet framework of "entity-relationship-entity" or "entity-attribute-attribute value" can be constructed.

[0129] The knowledge extraction unit 622 is used to extract the transmission knowledge corpus based on the corpus annotation, and to obtain entity information, relation information and attribute information in the transmission knowledge corpus by using knowledge extraction.

[0130] Specifically, knowledge extraction includes entity extraction, relation extraction, and attribute extraction. Entity extraction employs named entity recognition (NER) algorithms. Based on prior training, natural language processing (NLP) techniques are used to automatically collect entity content from the corpus. The collected data is then added to the model for further training to improve accuracy. Finally, an entity database is built by organizing and summarizing the entity information in the text. Relation extraction algorithms essentially obtain specific descriptions of the logical relationships between entities or between entities and attributes in the text corpus. Through this, they extract potential meanings and construct semantic connections between entities. Current relation extraction algorithms are mainly divided into: manual annotation combined with entity extraction, machine learning-based semantic relation recognition, deep learning-based relation recognition, and joint extraction algorithms combined with named entity recognition. Attribute extraction obtains the attribute information of entities, enriching the content of the entities.

[0131] Furthermore, since named entity recognition algorithms are highly efficient in labeling, but require a large amount of data for training and have relatively low accuracy, the choice of entity labeling algorithm often depends on the data situation and work requirements. Therefore, the BERT-BiLSTM-CRF model is used for entity extraction. The BERT-BiLSTM-CRF model is a named entity recognition model composed of the BERT pre-trained language model and the bidirectional long short-term memory network-conditional random field model (BiLSTM-CRF).

[0132] The BERT-BiLSTM-CRF model consists of three layers. The first layer is the BERT layer, which uses a large amount of general Chinese corpus and extensive computation to train a general language model, extracting word vector features of transmission knowledge information to obtain low-dimensional word vectors. The second layer is the BiLSTM layer, which, based on the understanding of word vector features of transmission knowledge information from the pre-trained layer, trains a bidirectional long short-term memory network using a large amount of annotated transmission knowledge corpus. Based on the model training results, it uses contextual semantic information to infer and label entity sequences in the corpus, and further filters entity types by setting weights through an attention mechanism. The third layer is the CRF layer, which, based on the entity sequences output by the BiLSTM layer, uses a probabilistic model to predict and output the optimal expression for sequence labels, thereby achieving automatic sequence labeling of the corpus and completing named entity recognition.

[0133] The triple frame filling unit 623 is used to fill the triple frame based on the entity information, the relationship information and the attribute information, and associate the filled triple frame based on semantic relationship.

[0134] The data layer generation unit 624 is used to perform entity fusion using a knowledge fusion algorithm, generate data entities, and associate the filled triplet framework with the data entities to generate the data layer.

[0135] Specifically, knowledge fusion is divided into two parts: ontology alignment and entity matching. Ontology alignment refers to comparing and confirming the ontology concepts and levels of the entities to be confirmed. If the similarity of levels and concepts is high, the ontology is considered to be aligned. Entity matching refers to confirming the degree of similarity between two entities in terms of entity names, entity relationships, and attributes.

[0136] Furthermore, the step of using a knowledge fusion algorithm to perform entity fusion and generate data entities includes: comparing the entity's ontology concept with its hierarchy, determining ontology alignment based on the comparison result, calculating the semantic similarity of the entity based on the determination result, setting a similarity threshold, and performing entity fusion based on the semantic similarity to generate data entities.

[0137] Furthermore, the semantic similarity between two entities is calculated in terms of entity name, hierarchical category, and attributes. A comprehensive similarity vector is derived by combining the similarity of individual attributes to determine whether entity redundancy exists. The formula for calculating the similarity of individual attributes is as follows:

[0138]

[0139] In the above formula, This represents the similarity of entities A and B based on a certain attribute m. A m Let m represent a certain attribute of entity A. B m Let m represent a certain attribute of entity B. a i Represents attributes A m Word frequency of segmentation b i Represents attributes B m The word frequency of the segmented words, where n represents the number of segments.

[0140] Furthermore, A m , B m There are n word segments in the semantic space, which are determined by statistical attributes. A m , B m Word frequency of each segment a i , b i The word frequency vector (i.e., the comprehensive similarity vector) is constructed, and the similarity between two sentences is determined by calculating the cosine value of the vector.

[0141] Furthermore, by calculating the ratio of semantic similarity between entities A and B on each attribute to the number of attributes, the similarity ratio of entities A and B is generated, and the calculation formula is as follows:

[0142]

[0143] In the above formula, S A,B This represents the similarity ratio between entities A and B, where 's' represents the number of attributes. S A,B The value ranges from 0 to 1, with the value closer to 1 indicating a higher semantic similarity between the two entities.

[0144] The association mapping unit 625 is used to establish the association mapping between the data layer and the pattern layer, and to combine the data layer and the pattern layer based on the association mapping to generate a transmission knowledge graph.

[0145] Specifically, by mapping the ontology and its relationships in the schema layer to numerous data entities in the data layer, a mapping is established between the data layer and the schema layer. By combining the data layer and the schema layer, the transmission knowledge graph can be constructed.

[0146] In one embodiment, the transmission solution generation module 63 includes:

[0147] The keyword generation unit 631 is used to collect expected factors of the transmission solution, preprocess the expected factors of the transmission solution, and generate keywords.

[0148] Specifically, for the expected factors of the transmission solution input by the user, if the user input is in natural language format, preprocessing of the expected factors of the user input is performed, including word segmentation, text disambiguation, and keyword recognition. Among them, word segmentation refers to breaking down the user input question into multiple groups of words for representation; text disambiguation refers to removing meaningless components in the word segmentation results and eliminating ambiguity in the text to prevent polysemy from affecting the accuracy of applicability recognition. Text disambiguation mainly includes stop word removal and part-of-speech tagging; keyword recognition is performed on the disambiguated text to filter out the core entities, relationships, or descriptions of the user's question content.

[0149] Furthermore, a TF-IDF (term frequency–inverse document frequency) weighted model is used for keyword identification and extraction. The steps include: inputting the text into the TF-IDF weighted model to determine the words. t i Word frequency in text tf i The calculation formula is as follows:

[0150]

[0151] in, n ij Words t i In the text d j The number of times it appears in Indicated in text d j The sum of the number of occurrences of all words in the text.

[0152] Calculate the frequency of reverse files IDF i The calculation formula is as follows:

[0153]

[0154] In the above formula, |D| represents the total number of texts in the corpus. Indicates words containing t i The number of texts.

[0155] Word frequency calculation based on term frequency and reverse document frequency t i TF-IDF value tf idfi The calculation formula is as follows:

[0156]

[0157] As can be seen from the above formula, the high frequency of a word in a specific file and the low frequency of that word in the entire file set can generate a high-weight TF-IDF value. Then, the TF-IDF values ​​of each word in the text are counted, and the words corresponding to the TF-IDF values ​​that are combined with the preset weight thresholds are selected as keywords.

[0158] The solution requirement identification unit 632 is used to identify solution requirements based on the keywords and to classify the solution requirements using a requirement classifier.

[0159] Specifically, the solution requirements in the expected factors of the transmission solution input by the user are analyzed, so as to obtain a solution that matches the solution requirements in the subsequent matching process.

[0160] The entity logical association generation unit 633 is used to determine the entities in the transmission knowledge graph based on the keywords, take the entities as core nodes, determine the associated entities and attribute values ​​according to the entity relationships, and generate entity logical associations based on the associated entities and attribute values.

[0161] The transmission solution generation unit 634 is used to logically associate the solution requirements with the entity, filter the transmission solutions using the entity relationships, and send the transmission solutions to the user.

[0162] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating transmission solutions based on transmission knowledge graphs, characterized in that, include: Step 101: Collect transmission solutions and construct a transmission knowledge information corpus based on the transmission solutions; Step 102: Extract the core elements of the transmission solution, construct a pattern layer based on the core elements, construct a data layer using entity information, relationship information and attribute information in the transmission knowledge information corpus, establish an association mapping between the data layer and the pattern layer, and generate a transmission knowledge graph based on the association mapping. The core elements include at least: the motor, reducer, various accessories, transmission requirements, environmental conditions, spatial conditions, transmission method, performance parameters, assembly relationship, lubrication and sealing, application duration and application intensity included in the transmission solution. Step 103: Collect expected factors of the transmission solution in natural language format input by the user, identify solution requirements based on the expected factors, determine entity logical associations based on the expected factors and the transmission knowledge graph, and generate the transmission solution according to the solution requirements and the entity logical associations. The expected factors include: expected application products and scenarios, power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions. Step 101 includes: Collect the original description files of existing transmission solutions, perform text preprocessing on the original description files, and generate a structured transmission knowledge corpus; According to predetermined rules, the pre-processed and structured transmission knowledge corpus is tagged to generate corpus annotations, and a transmission knowledge information corpus is constructed based on the transmission knowledge corpus and the corpus annotations. The step of tagging the pre-processed and structured transmission knowledge corpus according to predetermined rules includes: For the corpus of transmission knowledge that describes application products and scenarios, each corpus of transmission knowledge is tagged with power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions based on its content; for the corpus of transmission knowledge that describes transmission solutions, the corpus of transmission methods, motor selection, reducer selection, accessory selection, motor / reducer / accessory performance parameters, assembly method, and lubrication and sealing method are tagged. Step 102, which involves extracting the core elements of the transmission solution and constructing a pattern layer based on these core elements, includes: The transmission solution includes motors, reducers, various accessories, transmission requirements, environmental conditions, and spatial conditions, which are defined as entities. The transmission method, assembly relationship, and lubrication and sealing structure relationship included in the transmission solution are defined as relationships. The performance parameters, application duration, and application intensity included in the transmission solution are defined as attributes, and their specific values ​​are used as attribute values. The entities, relationships, attributes, and their logical associations of the core elements in the existing transmission solution are analyzed to construct a triplet framework of "entity-relationship-entity" or "entity-attribute-attribute value" to generate the pattern layer. The step 102 of constructing a data layer using entity information, relation information, and attribute information from the transmission knowledge information corpus includes: Starting from the aforementioned transmission knowledge information corpus, knowledge extraction is applied to obtain entity, relation, and attribute information from the transmission knowledge corpus based on the corpus annotations in the transmission knowledge information corpus. Triples of type "entity-relation-entity" and "entity-attribute-attribute value" are filled in, and entity fusion is performed using a knowledge fusion algorithm to generate data entities. The filled triple framework is then associated with the data entities, and a data layer is constructed in a bottom-up manner. Step 103 includes: Preprocessing of the expected factors of the transmission solution involves word segmentation, text disambiguation, and keyword recognition to generate keywords; Based on the keywords, the solution requirements are identified, and a requirement classifier is used to classify the solution requirements. Based on the keywords, entities in the transmission knowledge graph are identified. Taking the entities as core nodes, associated entities and attribute values ​​are determined according to entity relationships. Entity logical associations are generated based on the associated entities and attribute values. Based on the solution requirements and the logical association with the entities, the transmission solutions are filtered using the entity relationships, and then the transmission solutions are sent to the user. The TF-IDF weighted model is used for keyword identification and extraction. The steps include: inputting the text into the TF-IDF weighted model to determine the words. Word frequency in text The calculation formula is as follows: in, Words In the text The number of times it appears in Indicated in text The sum of the occurrences of all words in the text; calculate the frequency of the reverse-engineered file. The calculation formula is as follows: In the above formula, This represents the total number of texts in the corpus. Indicates words containing The number of texts; word count based on term frequency and reverse document frequency. TF-IDF value The calculation formula is as follows: Then, the TF-IDF values ​​of each word in the text are counted, and the words corresponding to the TF-IDF values ​​that are combined with the preset weight thresholds are selected as keywords.

2. The method as described in claim 1, characterized in that, The method of using knowledge fusion algorithms to perform entity fusion and generate data entities includes: The entity's ontology concept is compared with its hierarchy. Ontology alignment is determined based on the comparison results. Semantic similarity of the entity is calculated based on the determination results, and a similarity threshold is set. Entity fusion is performed based on the semantic similarity to generate data entities.

3. A transmission solution generation system based on transmission knowledge graph, characterized in that, include: A corpus construction module for transmission knowledge information is used to collect transmission solutions and construct a corpus of transmission knowledge information based on the transmission solutions, including: The process involves collecting original description files of existing transmission solutions, performing text preprocessing on these files to generate structured transmission knowledge corpus, and tagging the preprocessed structured transmission knowledge corpus according to predetermined rules to generate corpus annotations. A transmission knowledge information corpus is then constructed based on the transmission knowledge corpus and the corpus annotations. The tagging of the preprocessed structured transmission knowledge corpus according to predetermined rules includes: for corpus describing application products and scenarios, tagging each corpus with labels such as power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions; and for corpus describing transmission solutions, tagging with labels such as transmission method, motor selection, reducer selection, accessory selection, motor / reducer / accessory performance parameters, assembly method, and lubrication / sealing method. A transmission knowledge graph generation module is used to extract the core elements of the transmission solution, construct a pattern layer based on the core elements, and construct a data layer using entity information, relationship information, and attribute information from the transmission knowledge information corpus. An association mapping is established between the data layer and the pattern layer, and a transmission knowledge graph is generated based on the association mapping. The core elements of the transmission solution include: the motor, reducer, various accessories, transmission requirements, environmental conditions, spatial conditions, transmission method, performance parameters, assembly relationships, lubrication and sealing, application duration, and application intensity. Furthermore, the transmission knowledge graph generation module includes a pattern layer generation unit, which defines the motor, reducer, various accessories, transmission requirements, environmental conditions, and spatial conditions included in the transmission solution as entities, and defines the transmission method, assembly relationships, lubrication and sealing, application duration, and application intensity as entities. The assembly and assembly relationships and the structural relationships of lubrication and sealing are defined as relationships, and the performance parameters, application duration, and application intensity included in the transmission solution are defined as attributes, with their specific values ​​serving as attribute values. Then, the entities, relationships, attributes, and their logical associations of these core elements in the existing transmission solution are analyzed to construct a "entity-relationship-entity" or "entity-attribute-attribute value" triple framework, generating the pattern layer. Furthermore, starting from the transmission knowledge information corpus, knowledge extraction is used to obtain entity, relationship, and attribute information from the transmission knowledge corpus based on the corpus annotations. Triples of the "entity-relationship-entity" and "entity-attribute-attribute value" types are filled in, and entity fusion is performed using a knowledge fusion algorithm to generate data entities. The filled triple framework is then associated with the data entities, constructing the data layer in a bottom-up manner. The transmission solution generation module is used to collect expected factors of transmission solutions in natural language format input by the user, identify solution requirements based on the expected factors, determine entity logical associations based on the expected factors and the transmission knowledge graph, and generate the transmission solution according to the solution requirements and the entity logical associations; wherein, the expected factors of the transmission solution include: expected application products and scenarios, power transmission requirements, environmental conditions, application duration, application intensity, and spatial conditions; Furthermore, the transmission solution generation module includes: a keyword generation unit, used to collect expected factors of the transmission solution, and perform preprocessing such as word splitting, text disambiguation, and keyword recognition on the expected factors to generate keywords; a solution requirement identification unit, used to identify solution requirements based on the keywords, and classify the solution requirements using a requirement classifier; an entity logical association generation unit, used to determine entities in the transmission knowledge graph based on the keywords, using the entities as core nodes, determining associated entities and attribute values ​​according to entity relationships, and generating entity logical associations based on the associated entities and attribute values; and a transmission solution generation unit, used to filter the transmission solutions based on the solution requirements and entity logical associations, using the entity relationships, and sending the transmission solutions to the user; wherein, the TF-IDF weighted model is used for keyword identification and extraction, the steps of which include: inputting text into the TF-IDF weighted model to determine words. Word frequency in text The calculation formula is as follows: in, Words In the text The number of times it appears in Indicated in text The sum of the occurrences of all words in the text; calculate the frequency of the reverse-engineered file. The calculation formula is as follows: In the above formula, This represents the total number of texts in the corpus. Indicates words containing The number of texts; word count based on term frequency and reverse document frequency. TF-IDF value The calculation formula is as follows: Then, the TF-IDF values ​​of each word in the text are counted, and the words corresponding to the TF-IDF values ​​that are combined with the preset weight thresholds are selected as keywords.

4. The system as described in claim 3, characterized in that, The method of using knowledge fusion algorithms to perform entity fusion and generate data entities includes: The entity's ontology concept is compared with its hierarchy. Ontology alignment is determined based on the comparison results. Semantic similarity of the entity is calculated based on the determination results, and a similarity threshold is set. Entity fusion is performed based on the semantic similarity to generate data entities.

Images