A steel resource retrieval method, system and device based on a knowledge graph

Abstract

This invention relates to a knowledge graph-based method, system, and apparatus for steel resource retrieval. The method includes: constructing and storing a retrieval knowledge graph for searching steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph establishes the association between commodity codes and entities of steel resources, where the entities include uses and characteristics; obtaining the steel uses and / or characteristics to be retrieved; and querying all commodity codes associated with those uses and / or characteristics from the retrieval knowledge graph. Compared with existing technologies, this invention has advantages such as high retrieval efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of information retrieval technology, and in particular to a method, system, and apparatus for retrieving steel resources based on knowledge graphs. Background Technology

[0002] The stock keeping unit (SKU) currently used by steel e-commerce platforms for retrieval includes the following six dimensions: grade, coating amount, coating type, surface texture, surface treatment, and edge morphology. This chain can be continuously expanded to refine the granularity of steel products, providing users with more accurate results even for fuzzy searches.

[0003] Steel SKUs can be stored using various methods. However, using a relational database such as MySQL for storage has the following drawbacks:

[0004] 1. Overly fine granularity leads to an excessively large amount of data. For example, if there are 100 values ​​for each of the current 6 dimensions, the length of the table would be 6 to the power of 100. As more dimensions are added later, the amount of data will grow exponentially, greatly reducing search efficiency.

[0005] 2. The large amount of data also means that the steel expert responsible for maintaining this SKU data needs to spend a lot of time checking it;

[0006] 3. The clustering relationship between SKUs can only be displayed in numerical form by querying using SQL, which is not intuitive enough.

[0007] Therefore, using traditional relational databases for storage has significant limitations in terms of query efficiency, maintenance costs, and clustered relationship queries. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a steel resource retrieval method, system, and device based on knowledge graphs, which greatly improves retrieval efficiency and accuracy.

[0009] The objective of this invention can be achieved through the following technical solutions:

[0010] A knowledge graph-based steel resource retrieval method, comprising:

[0011] Construct and store a retrieval knowledge graph for retrieving steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph constructs the association between commodity codes and entities of steel resources, and the entities include uses and characteristics.

[0012] Obtain the intended use and / or characteristics of the steel to be searched, and query all product codes associated with that intended use and / or characteristic from the search knowledge graph.

[0013] Preferably, the method for constructing the retrieval knowledge graph includes:

[0014] S1. Extract entities from steel standards, including their uses, characteristics, standard numbers, and standard templates, and determine the relationships between these entities;

[0015] S2. Based on the relationship between steel standards and grades, construct a knowledge graph of steel application characteristics with the standard number as the core. The knowledge graph of steel application characteristics associates the standard number, grade, application, characteristics, and standard template.

[0016] S3. Obtain a steel SKU knowledge graph, wherein the SKU knowledge graph associates a unique product code with an SKU, and the SKU includes the grade;

[0017] S4. Based on the grade, integrate the steel application characteristic knowledge graph and SKU knowledge graph, and associate the standard number, grade, application, characteristics, standard template with the unique commodity code to form a retrieval knowledge graph.

[0018] Preferably, step S1 includes:

[0019] S11. Collect steel standards, and use the standard models, standard scopes, and normative references as raw data;

[0020] S12. The original data in the standard range column is segmented and labeled with part of speech. This data is used as input data to train the named entity recognition model. The named entity recognition model is used to extract short words. The short words include features and uses. Then, an entity table of short words is constructed based on the short words. Some short words are concatenated into long words to obtain an entity table of long words. The association between long words and short words is constructed.

[0021] S13. Extract the standard template and version number from the standard model and normative reference documents, combine the standard template and version number to form the standard number, and obtain the entity table of the standard number.

[0022] S14. Link the entity tables of short words, long words, and standard numbers to construct the association relationship between standard numbers, long words, and short words.

[0023] S15. Cluster the standard numbers that have the same standard template and establish the association between the standard number and the standard template.

[0024] Preferably, the named entity recognition model is a BERT-based named entity recognition model.

[0025] Preferably, the steel SKU knowledge graph in step S3 is a pre-stored knowledge graph, and the SKU also includes coating amount, coating type, surface structure, surface treatment, and edge morphology.

[0026] Preferably, the Cypher graph query language is used when querying the knowledge graph.

[0027] Preferably, the method further includes: after the query is completed, displaying all steel resources with the queried product codes.

[0028] A knowledge graph-based steel resource retrieval system, comprising:

[0029] Storage module: Stores a retrieval knowledge graph for retrieving steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph constructs the association between commodity codes and entities of steel resources, and the entities include uses and characteristics.

[0030] Search module: Obtain the intended use and / or characteristics of the steel to be searched, find all product codes associated with the intended use and / or characteristics from the search knowledge graph, and display resources tagged with this product code.

[0031] Preferably, the system further includes a display module that displays all steel resources with queried product codes.

[0032] A knowledge graph-based steel resource retrieval device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the knowledge graph-based steel resource retrieval method.

[0033] Compared with the prior art, the present invention has the following advantages:

[0034] (1) Based on the extraction of uses, features, standard numbers, standard templates, grades and the relationships between these entities from the standard, this invention constructs a steel use feature knowledge graph with the standard number as the core, and then integrates the steel SKU knowledge graph of Ouye Cloud Commerce itself, and associates uses, features and commodity types, so that the model can find the corresponding commodity type immediately after the user inputs the use and features, and return all steel resources that match these commodity types on the front end, thereby greatly improving the retrieval accuracy of steel commodities and the user's retrieval satisfaction;

[0035] (2) This invention incorporates knowledge graph technology. Compared to deep learning models in traditional recommendation algorithms, knowledge graphs offer high interpretability and the ability to quickly query many-to-many relationships and penetrate deep layers. While improving user experience, it also provides strong support for the platform's operational analysis.

[0036] (3) This invention combines rule models with deep learning models in natural language processing to ensure that the extracted entities are accurate and complete.

[0037] (4) This invention clusters steel SKUs in terms of use and features, allowing users to search for target steel resources by features or uses, thereby improving the user search function and increasing the user search efficiency. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the specific process of a knowledge graph-based steel resource retrieval method according to the present invention. Detailed Implementation

[0039] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Note that the following description of the embodiments is merely illustrative and is not intended to limit its applicability or use, nor is the present invention limited to the following embodiments.

[0040] Example 1

[0041] like Figure 1 As shown in the figure, this embodiment provides a steel resource retrieval method based on knowledge graphs, which includes:

[0042] A retrieval knowledge graph is constructed and stored for retrieving steel commodity codes based on steel uses and / or characteristics. This retrieval knowledge graph establishes the association between commodity codes and entities of steel resources. Entities include uses and characteristics. The retrieval knowledge graph is constructed through the following steps:

[0043] S1. Extract entities from steel standards, including their uses, characteristics, standard numbers, and standard templates, and determine the relationships between these entities;

[0044] S2. Based on the relationship between steel standards and grades, construct a knowledge graph of steel application characteristics with standard number as the core. The knowledge graph of steel application characteristics links standard number, grade, application, characteristics, and standard template.

[0045] S3. Obtain the steel SKU knowledge graph. The SKU knowledge graph associates a unique product code with an SKU, which includes the grade.

[0046] S4. Based on the grade, integrate the steel application characteristic knowledge graph and SKU knowledge graph, and associate the standard number, grade, application, characteristics, standard template with the unique commodity code to form a retrieval knowledge graph.

[0047] Based on the aforementioned construction of the retrieval knowledge graph, the uses and / or characteristics of the steel to be retrieved are obtained, and all product codes associated with the uses and / or characteristics are queried from the retrieval knowledge graph.

[0048] As a preferred implementation, after the query is completed, all steel resources with the obtained product codes are displayed.

[0049] In a preferred embodiment, step S1 includes:

[0050] S11. Collect steel standards and compile the information on standard models, standard scope, and normative references into a table as raw data.

[0051] S12. The original data in the "Standard Range" column is segmented and labeled with parts of speech. This data is used as input to train the named entity recognition model. The named entity recognition model is used to extract short words, which include features and uses. Then, an entity table for the short words is constructed. Some short words are concatenated into long words to obtain an entity table for the long words. The relationship between long words and short words is constructed. The named entity recognition model is a BERT-based named entity recognition model. Specifically:

[0052] The raw data in the 'Standard Range' column is segmented and part-of-speech tagged, serving as input data for training a BERT-based named entity recognition model. This model is then used to extract short words, constructing a 'short word entity table'. Based on the positions of the judgment words and specialized terms, some short words are concatenated into longer words, resulting in a 'long word entity table' (e.g., the three short words 'construction', 'structural use', and 'thick steel plate' can be concatenated into a single long word 'building structure thick steel plate'). Relationships are established in the knowledge graph: long word - [contains] -> short word.

[0053] S13. Extract the standard template and version number from the standard model and normative reference documents. Combine the standard template and version number to obtain the standard number, thus obtaining the entity table of the standard number. Specifically:

[0054] Entities are extracted from 'standard model' and 'normative reference documents' according to rules such as separators and Chinese / English representation, yielding: standard template (e.g., Q / BQB 203), version number (e.g., 2018), standard name, publication date, and implementation date. Then, the extracted entities are corrected, supplemented, and name-disambiguated using the official website of the 'National Standards Full-Text Disclosure System' (when the name of the same standard differs between the local file and the official website, the official website name prevails). Finally, the 'standard template' and 'version number' are combined to form the standard number, resulting in the 'Standard Number Entity Table'.

[0055] S14. Link the entity tables of short words, long words, and standard numbers to construct the association relationship between standard numbers, long words, and short words.

[0056] S15. Since the 'standard number' is obtained by merging the 'standard template' and the 'version number', standard numbers with the same standard template are clustered to establish the association between standard number and standard template: standard number - [clustering] -> standard template.

[0057] In a preferred embodiment, the following steps may also be included: combining the 'normative reference document' field in the original data with the 'entity table of standard number' to establish a relationship in the knowledge graph: standard number - [reference document] -> standard number, so that the steel application characteristic knowledge graph can also associate a certain standard number with other standard numbers.

[0058] Step S2 combines Ouye Cloud Commerce's proprietary 'standard and grade association' data to obtain a knowledge graph of steel application characteristics centered on the standard number and interconnected with grade, standard template, long words, short words, and other standard numbers.

[0059] Step S3 obtains a steel SKU knowledge graph consisting of six entities: grade, coating amount, coating type, surface structure, surface treatment, and edge morphology, along with their corresponding unique product codes. Long / short terms are then associated with standard numbers, grade numbers, and product codes to fuse the two graphs. The steel SKU knowledge graph is a pre-stored knowledge graph.

[0060] Finally, the user enters the purpose or keywords in the search box, and the model uses the graph query language Cypher to find all the product codes (i.e., the unique key values ​​of SKUs) associated with that purpose or keywords from the knowledge graph, and displays the resources with these product code tags to the customer on the front end.

[0061] This invention first collects steel standards covering three aspects: standard model, standard scope, and normative references. Next, it performs word segmentation and part-of-speech tagging on the raw data in the 'standard scope' column to train a BERT-based named entity recognition model. This model is then used to extract short words, constructing a 'short word entity table'. Based on the positions of judgment words and specialized terms, some short words are concatenated into longer words, resulting in a 'long word entity table'. Relationships are established in the knowledge graph: long word - [contains] -> short word. Then, a rule-based model is used to extract entities from 'standard model' and 'normative references', obtaining the standard template, version number, standard name, publication date, and implementation date. The extracted entities are then corrected, supplemented, and disambiguated through the official website of the 'National Standards Full-Text Disclosure System'. The 'standard template' and 'version number' are combined to form the standard number, resulting in the 'entity table of standard numbers'. Then, the 'entity table of short words', 'entity table of long words', and 'entity table of standard numbers' are combined to establish a relationship in the knowledge graph: standard number - [extraction] -> long / short words. Next, standard number entities with the same standard template are clustered to establish a relationship in the knowledge graph: standard number - [clustering] -> standard template. Furthermore, combining Ouye Cloud Commerce's proprietary 'standard and brand association' data, 'long / short words' are associated with 'standard number', 'brand name', and 'product code', achieving the fusion of the two graphs. Finally, users input their purpose or feature words through the search box, and the model uses the graph query language Cypher to search the knowledge graph for all product codes associated with that purpose or feature word, displaying resources with these product code tags to the customer on the front end.

[0062] The aforementioned method combines rule-based models with deep learning models in natural language processing to extract entities such as standard terms and feature words from steel resources, making the extracted entities more accurate and complete. Simultaneously, steel SKUs are clustered along the dimensions of usage and features, allowing users to retrieve target steel resources by feature or usage, thereby improving user search functionality and efficiency. Furthermore, knowledge graph technology is integrated into the platform's recommendation algorithm. Compared to deep learning models in traditional recommendation algorithms, knowledge graphs offer high interpretability and the ability to quickly query many-to-many relationships and penetrate deeper layers. While enhancing user experience, this also provides strong support for the platform's operational analysis. In summary, through optimization in these three dimensions, the business platform can construct a knowledge graph centered on the standard number by extracting the uses, features, standard numbers, standard templates, grades, and relationships between these entities from the standard. This knowledge graph is then integrated with Ouye Cloud Commerce's own steel SKU knowledge graph, linking uses and features with product types. This allows the model to immediately find the corresponding product type after the user inputs the uses and features, and return all steel resources that match these product types on the front end. This significantly improves the accuracy of steel product retrieval and user satisfaction.

[0063] Example 2

[0064] This embodiment provides a knowledge graph-based steel resource retrieval system, which includes:

[0065] Storage module: Stores the retrieval knowledge graph used for retrieving steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph constructs the association between commodity codes and entities of steel resources. Entities include uses and characteristics.

[0066] Search module: Obtain the intended use and / or characteristics of the steel to be searched, find all product codes associated with the intended use and / or characteristics from the search knowledge graph, and display resources tagged with this product code.

[0067] The system also includes a display module that shows all steel resources with searchable product codes.

[0068] This embodiment provides a knowledge graph-based steel resource retrieval system that implements the knowledge graph-based steel resource retrieval method in Embodiment 1. This retrieval method has been specifically described in Embodiment 1 and will not be repeated here.

[0069] Example 2

[0070] This embodiment provides a steel resource retrieval device based on a knowledge graph, including a memory and a processor. The memory is used to store a computer program, and the processor is used to implement the steel resource retrieval method based on a knowledge graph as described in Embodiment 1 when the computer program is executed. This retrieval method has been specifically described in Embodiment 1, and will not be repeated in this embodiment.

[0071] The above embodiments are merely illustrative and do not constitute a limitation on the scope of the present invention. These embodiments can also be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the technical spirit of the present invention.

Claims

1. A steel resource retrieval method based on knowledge graphs, characterized in that, The method includes: Construct and store a retrieval knowledge graph for retrieving steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph constructs the association between commodity codes and entities of steel resources, and the entities include uses and characteristics. Obtain the intended use and / or characteristics of the steel to be retrieved, and query all product codes associated with that intended use and / or characteristic from the retrieval knowledge graph; the construction method of the retrieval knowledge graph includes: S1. Extract entities from steel standards, including their uses, characteristics, standard numbers, and standard templates, and determine the relationships between these entities; S2. Based on the relationship between steel standards and grades, construct a knowledge graph of steel application characteristics with the standard number as the core. The knowledge graph of steel application characteristics associates the standard number, grade, application, characteristics, and standard template. S3. Obtain a steel SKU knowledge graph, wherein the SKU knowledge graph associates a unique product code with an SKU, and the SKU includes the grade; S4. Based on the grade, integrate the steel application characteristic knowledge graph and SKU knowledge graph, and associate the standard number, grade, application, characteristics, standard template with the unique product code to form a retrieval knowledge graph; Step S1 includes: S11. Collect steel standards, and use the standard models, standard scopes, and normative references as raw data; S12. The original data in the standard range column is segmented and labeled with part of speech. This data is used as input data to train the named entity recognition model. The named entity recognition model is used to extract short words. The short words include features and uses. Then, an entity table of short words is constructed based on the short words. Some short words are concatenated into long words to obtain an entity table of long words. The association between long words and short words is constructed. S13. Extract the standard template and version number from the standard model and normative reference documents, combine the standard template and version number to form the standard number, and obtain the entity table of the standard number. S14. Link the entity tables of short words, long words, and standard numbers to construct the association relationship between standard numbers, long words, and short words. S15. Cluster the standard numbers that have the same standard template and establish the association between the standard number and the standard template.

2. The steel resource retrieval method based on knowledge graphs according to claim 1, characterized in that, The named entity recognition model is a BERT-based named entity recognition model.

3. The steel resource retrieval method based on knowledge graphs according to claim 1, characterized in that, The steel SKU knowledge graph in step S3 is a pre-stored knowledge graph. The SKU also includes coating amount, coating type, surface structure, surface treatment, and edge morphology.

4. The steel resource retrieval method based on knowledge graphs according to claim 1, characterized in that, When querying the knowledge graph, the Cypher graph query language is used.

5. The steel resource retrieval method based on knowledge graphs according to claim 1, characterized in that, The method also includes: after the query is completed, displaying all steel resources with the obtained product codes.

6. A steel resource retrieval system based on knowledge graphs, characterized in that, The system employs the knowledge graph-based steel resource retrieval method described in claim 1, and the system comprises: Storage module: Stores a retrieval knowledge graph for retrieving steel commodity codes based on steel uses and / or characteristics. The retrieval knowledge graph constructs the association between commodity codes and entities of steel resources, and the entities include uses and characteristics. Search module: Obtain the intended use and / or characteristics of the steel to be searched, find all product codes associated with the intended use and / or characteristics from the search knowledge graph, and display resources tagged with this product code.

7. A steel resource retrieval system based on knowledge graphs according to claim 6, characterized in that, The system also includes a display module that shows all steel resources with searchable product codes.

8. A steel resource retrieval device based on knowledge graph, characterized in that, The system includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the knowledge graph-based steel resource retrieval method as described in any one of claims 1 to 5 when the computer program is executed.

Images