A method for constructing enterprise material classification catalog based on artificial intelligence

Abstract

The present invention provides a method for constructing an enterprise material classification catalog based on artificial intelligence. The present invention uses artificial intelligence technology to extract image features or text information from enterprise materials and related external information that need to be classified by the enterprise, obtain an image recognition model or a text classification model, and completely grasp the structure, working principle, category and other information of internal equipment and materials. By clustering and modeling the labels obtained by clustering, the existing equipment and materials of the enterprise are classified and associated to establish a catalog. The method can quickly, accurately and comprehensively connect with external related information, automatically deduplicate and normalize, and finally confirm through a manual review link to form a complete catalog of enterprise equipment and materials, avoiding duplication in the catalog sorting process.

Description

Technical Field

[0001] The present invention belongs to the technical field of enterprise material classification, and specifically relates to a method for constructing an enterprise material classification catalog based on artificial intelligence. Background Art

[0002] At present, most of the equipment and material management systems used by enterprises are traditional MIS, which are in the stage of realizing electronic business processes and replacing traditional paper documents. Although compared with the traditional paper document management method, MIS is more convenient for finding and transmitting the location of the input directory, it still lacks intelligence. The technical level of managers and their understanding of the business are limited. They are not clear about the structure, working principle and category of equipment and materials within each enterprise, and cannot quickly, accurately and comprehensively connect with external relevant information, resulting in low work efficiency and easy errors, omissions and duplications in the process of sorting the directory. Summary of the invention

[0003] Therefore, in order to overcome the above technical problems, the present invention provides a method for constructing an enterprise material classification catalog based on artificial intelligence, which is characterized by comprising:

[0004] S1: Collect the internal equipment and material information of the enterprise, and store the collected internal equipment and material information of the enterprise as a computer file, legally and compliantly collect external information related to the internal equipment and material information of the enterprise and establish a database to store the collected external information;

[0005] S2: Using artificial intelligence technology to extract image features or text information from the computer documents and databases in step S1, and obtaining an image recognition model or a text classification model and archiving them;

[0006] S3: Cluster the image recognition model or text classification model in the computer documents and databases, obtain cluster labels and archive them;

[0007] S4: Modeling is performed based on cluster labels to classify and associate the company’s existing equipment and materials and create a catalog;

[0008] S5: Automatically deduplicate the data and finally confirm it through manual review to form a complete catalog of the company's equipment and materials.

[0009] Furthermore, the steps of image feature extraction in step S2 include: step (1): reading equipment material image files; step (2): image preprocessing; step (3): calling an artificial intelligence image recognition algorithm to train the image preprocessing result obtained in step (2) to extract features.

[0010] Furthermore, the artificial intelligence image recognition algorithm is ResNet or InceptionNet.

[0011] Furthermore, the steps of text information extraction in step S2 include, in sequence: step (1): reading the equipment and material text file; step (2): text preprocessing, such as word segmentation, part-of-speech tagging, reference disambiguation, removal of stop words, keyword extraction, named entity recognition, semantic role labeling, word vector and word embedding; step (3): calling the artificial intelligence text classification algorithm to train the text preprocessing result obtained in step (2).

[0012] Furthermore, the artificial intelligence text classification algorithm is LSTM, RCNN or BERT.

[0013] Furthermore, the step of clustering the image recognition model in step S3 includes: step (1): reading the image recognition model document; step (2): image preprocessing; step (3): calling the clustering algorithm, the clustering algorithm is K-Means, and clustering the image preprocessing result obtained in step (2); step (4): manually reviewing the clustering result, establishing clustering labels and descriptions and associating them with the equipment and material names and archiving them;

[0014] The step of clustering the text classification model in step S3 includes: step (1): reading the text classification model; step (2): text preprocessing; step (3): calling the clustering algorithm, the clustering algorithm is K-Means or DBSCAN, and clustering the text preprocessing results obtained in step (2); step (4): manually reviewing the clustering results, establishing clustering labels and descriptions and associating them with equipment and material names, and archiving them.

[0015] Furthermore, the specific steps of step S4 are as follows: step (1): collecting and organizing the alternative catalog information of equipment and materials, including the enterprise's existing equipment and materials catalog, relevant catalogs of national standards, e-commerce website catalogs, and clustering labels obtained in the previous step; step (2): comprehensively analyzing and referring to the alternative catalog information, manually reviewing and forming a preliminary catalog, establishing a hierarchical association relationship with the alternative catalog and saving it; step (3): according to the position of the equipment and materials in the alternative catalog and the association relationship of step (2), each equipment and material is placed in the bottom sub-catalog of the newly created tree-like catalog. If multiple alternative catalogs are encountered, they are saved for manual review and decision; step (4): manually reviewing and deciding the catalog classification of each equipment and material, and continuing to adjust the hierarchical relationship of the newly created catalog tree to form a new catalog of enterprise equipment and materials.

[0016] Furthermore, the preliminary directory is in tree form.

[0017] Furthermore, the specific steps of step S5 are as follows: Step (1): traverse each equipment asset in the new directory, calculate the similarity based on the image recognition model and the text classification model, and form a list of suspected duplicates; Step (2): manually review the list of duplicates to confirm whether there are duplicates, and normalize the relevant information; Step (3): based on the manual review results, deduplicate and normalize the entire directory tree to form a complete directory of enterprise equipment and materials; Step (4): based on the complete directory, update the hierarchical association relationship with the above-mentioned alternative directory and save it.

[0018] Furthermore, after step S5, the following steps are further included:

[0019] S6: If the equipment and material information needs to be added, execute S7; if the equipment and material information does not need to be added, end; S7: The manager enters the equipment and material information in the operation interface. The equipment and material information can be the corresponding name, description and picture of the e-commerce website, or a photo taken on site; S8: Use artificial intelligence technology to extract image features or text information from the equipment and material information in step S7 to obtain an image recognition model or a text classification model; S9: Match and identify the extracted image recognition model or text classification model with the equipment and material information in the complete catalog of enterprise equipment and materials obtained in S5; S10: If the equipment and materials in the complete catalog of enterprise equipment and materials are matched, proceed to step S11; if no match is found, proceed to step S12; If any equipment and materials in the current complete catalog of equipment and materials of the enterprise are matched, execute step S12; S11: return the matching result to the operation interface for the manager to select, and then execute S14; S12: If any equipment and materials in the current complete catalog of equipment and materials of the enterprise are not matched, the manager collects external information based on the input equipment and materials information and supplements it to the database in S1, such as relevant information of e-commerce websites, public information of government websites, etc., and then executes S13; S13: call the artificial intelligence model algorithm to calculate the matching degree and possibility with each category or equipment and materials in the complete catalog of equipment and materials of the enterprise, and then execute S11; S14: After the manager selects one of them, the system records the relevant data and then returns to step S3.

[0020] Furthermore, the specific steps of step S9 are: step (1): matching and identifying the extracted image recognition model or text classification model with the image recognition model or text classification model of S2, and returning equipment and materials with a matching degree greater than 0.7; step (2): combining and sorting the equipment and material list of step (1) to obtain a matching result.

[0021] Furthermore, the specific steps of step S13 are as follows: step (1): compare the extracted image recognition model or text classification model with the image recognition model and text classification model obtained in S2, and return the three sub-directories with the highest matching degree; step (3): if there is information such as a national standard directory or an e-commerce directory, call the association relationship between the alternative directory and the enterprise directory for matching, return the three sub-directories with the highest matching degree, and then execute step (4); if not, directly execute step (4); step (4): combine and sort the sub-directory lists obtained in step (1) and step (2) and return them.

[0022] The beneficial effects of the present invention are as follows: the present invention uses artificial intelligence technology to extract image features or text information from enterprise materials and related external information that need to be classified by the enterprise, obtains an image recognition model or a text classification model, and completely grasps the structure, working principle, category and other information of internal equipment and materials. By clustering and modeling the labels obtained by clustering, the existing equipment and materials of the enterprise are classified and associated and a catalog is established. It can quickly, accurately and comprehensively connect with external related information, automatically deduplicate and standardize, and finally be confirmed through a manual review link to form a complete catalog of enterprise equipment and materials, avoiding duplication in the catalog sorting process. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Specific embodiments of the present invention will be described below.

[0025] This embodiment provides a method for constructing an enterprise material classification catalog based on artificial intelligence, which is characterized by comprising:

[0026] S1: Collect the internal equipment and material information of the enterprise, and store the collected internal equipment and material information of the enterprise as a computer file, legally and compliantly collect external information related to the internal equipment and material information of the enterprise and establish a database to store the collected external information;

[0027] S2: Using artificial intelligence technology to extract image features or text information from the computer documents and databases in step S1, and obtaining an image recognition model or a text classification model and archiving them;

[0028] S3: Cluster the image recognition model or text classification model in the computer documents and databases, obtain cluster labels and archive them;

[0029] S4: Modeling is performed based on cluster labels to classify and associate the company’s existing equipment and materials and create a catalog;

[0030] S5: Automatically deduplicate the data and finally confirm it through manual review to form a complete catalog of the company's equipment and materials.

[0031] The steps of image feature extraction in step S2 include: step (1): reading equipment material image files; step (2): image preprocessing; step (3): calling artificial intelligence image recognition algorithm to train the image preprocessing result obtained in step (2) to extract features.

[0032] The artificial intelligence image recognition algorithm is ResNet or InceptionNet.

[0033] The steps of text information extraction in step S2 include: step (1): reading equipment and material text files; step (2): text preprocessing, such as word segmentation, part-of-speech tagging, reference disambiguation, stop word removal, keyword extraction, named entity recognition, semantic role labeling, word vector and word embedding; step (3): calling an artificial intelligence text classification algorithm to train the text preprocessing result obtained in step (2).

[0034] The artificial intelligence text classification algorithm is LSTM, RCNN or BERT.

[0035] The step of clustering the image recognition model in step S3 includes: step (1): reading the image recognition model document; step (2): image preprocessing; step (3): calling the clustering algorithm, the clustering algorithm is K-Means, and clustering the image preprocessing result obtained in step (2); step (4): manually reviewing the clustering result, establishing clustering labels and descriptions and associating them with the equipment and material names and archiving them;

[0036] The step of clustering the text classification model in step S3 includes: step (1): reading the text classification model; step (2): text preprocessing; step (3): calling the clustering algorithm, the clustering algorithm is K-Means or DBSCAN, and clustering the text preprocessing results obtained in step (2); step (4): manually reviewing the clustering results, establishing clustering labels and descriptions, and associating them with the equipment and material names and archiving them.

[0037] Specific steps of step S4: Step (1): Collect and organize the alternative catalog information of equipment and materials, including the company's existing equipment and materials catalog, relevant catalogs of national standards, e-commerce website catalogs, and clustering labels obtained in the previous step; Step (2): Comprehensively analyze and refer to the alternative catalog information, manually review and form a preliminary catalog, establish a hierarchical association relationship with the alternative catalog and save it; Step (3): According to the position of the equipment and materials in the alternative catalog and the association relationship in step (2), each equipment and material is placed in the bottom sub-catalog of the newly created tree-like catalog. If multiple alternative catalogs are encountered, they are saved for manual review and decision; Step (4): Manually review and decide the catalog classification of each equipment and material, and continue to adjust the hierarchical relationship of the newly created catalog tree to form a new catalog of enterprise equipment and materials.

[0038] The initial directory is in tree form.

[0039] Specific steps of step S5: Step (1): traverse each equipment asset in the new directory, calculate the similarity based on the image recognition model and the text classification model, and form a list of suspected duplicates; Step (2): Manually review the list of duplicates to confirm whether there are duplicates, and standardize the relevant information; Step (3): Based on the results of manual review, deduplicate and standardize the entire directory tree to form a complete directory of enterprise equipment and materials; Step (4): Based on the complete directory, update the hierarchical association relationship with the above-mentioned alternative directory and save it.

[0040] After step S5, the following steps are also included:

[0041] S6: If the equipment and material information needs to be added, execute S7; if the equipment and material information does not need to be added, end; S7: The manager enters the equipment and material information in the operation interface. The equipment and material information can be the corresponding name, description and picture of the e-commerce website, or a photo taken on site; S8: Use artificial intelligence technology to extract image features or text information from the equipment and material information in step S7 to obtain an image recognition model or a text classification model; S9: Match and identify the extracted image recognition model or text classification model with the equipment and material information in the complete catalog of enterprise equipment and materials obtained in S5; S10: If the equipment and materials in the complete catalog of enterprise equipment and materials are matched, proceed to step S11; if no match is found, proceed to step S12; If any equipment and materials in the current complete catalog of equipment and materials of the enterprise are matched, execute step S12; S11: return the matching result to the operation interface for the manager to select, and then execute S14; S12: If any equipment and materials in the current complete catalog of equipment and materials of the enterprise are not matched, the manager collects external information based on the input equipment and materials information and supplements it to the database in S1, such as relevant information of e-commerce websites, public information of government websites, etc., and then executes S13; S13: call the artificial intelligence model algorithm to calculate the matching degree and possibility with each category or equipment and materials in the complete catalog of equipment and materials of the enterprise, and then execute S11; S14: After the manager selects one of them, the system records the relevant data and then returns to step S3.

[0042] Specific steps of step S9: Step (1): Match and identify the extracted image recognition model or text classification model with the image recognition model or text classification model of S2, and return equipment and materials with a matching degree greater than 0.7; Step (2): Combine and sort the equipment and material list of step (1) to obtain the matching result.

[0043] Specific steps of step S13: Step (1): compare the extracted image recognition model or text classification model with the image recognition model and text classification model obtained in S2, and return the three sub-directories with the highest matching degree; Step (3): if there is information such as a national standard directory or an e-commerce directory, call the association relationship between the alternative directory and the enterprise directory for matching, return the three sub-directories with the highest matching degree, and then execute step (4); if not, directly execute step (4); Step (4): combine and sort the sub-directory lists obtained in step (1) and step (2) and return them.

[0044] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be an element in the middle. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be an element in the middle. The terms "upper", "lower", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0045] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for constructing an enterprise material classification catalog based on artificial intelligence, characterized in that: include: S1: Collecting the internal equipment and material information of the enterprise, and storing the collected internal equipment and material information of the enterprise as a computer document, collecting external information related to the internal equipment and material information of the enterprise and establishing a database to store the collected external information; S2: Using artificial intelligence technology to extract image features or text information from the computer documents and databases in step S1, obtain an image recognition model or a text classification model and archive it; S3: Cluster the image recognition model or text classification model in the computer documents and databases, obtain cluster labels and archive them; S4: Modeling is performed based on cluster labels to classify and associate the company’s existing equipment and materials and create a catalog; S5: Automatically remove duplicates and finally confirm through manual review to form a complete catalog of enterprise equipment and materials; S6: If the equipment and material information needs to be increased, execute S7; if the equipment and material information does not need to be increased, end; S7: The manager inputs the equipment and material information on the operation interface; S8: Using artificial intelligence technology to extract image features or text information from the equipment and material information in step S7 to obtain an image recognition model or a text classification model; S9: Match and identify the extracted image recognition model or text classification model with the equipment and material information of the complete catalog of enterprise equipment and materials obtained in S5. The specific steps of step S9 are as follows: step (1): match and identify the extracted image recognition model or text classification model with the image recognition model or text classification model of S2, and return equipment and materials with a matching degree greater than 0.7; step (2): combine and sort the equipment and material list of step (1) to obtain a matching result; S10: If the equipment and materials in the complete catalog of equipment and materials of the enterprise are matched, then step S11 is performed; if no equipment and materials in the complete catalog of equipment and materials of the current enterprise are matched, then step S12 is performed; S11: Return the matching results to the operation interface for the administrator to select, and then execute S14; S12: If any equipment and materials in the complete catalog of equipment and materials of the current enterprise cannot be matched, the manager collects external information based on the input equipment and materials information and supplements it to the database in S1, and then executes S13; S13: Calling the artificial intelligence model algorithm to calculate the matching degree and possibility of each category or equipment and material with the complete catalog of equipment and materials of the enterprise, and then executing S11; S14: After the manager selects one of the options, the system records the relevant data and then returns to step S3; Specific steps of step S4: Step (1): Collect and organize the information of the alternative catalog of equipment and materials, including the existing equipment and materials catalog of the enterprise, the relevant catalog of national standards, the catalog of e-commerce websites and the clustering labels obtained in step S3; Step (2): Comprehensively analyze and refer to the alternative catalog information, manually review and form a preliminary catalog, establish and save the hierarchical association relationship with the alternative catalog; Step (3): According to the position of the equipment and materials in the alternative catalog and the association relationship in step (2), each equipment and material is placed in the lowest sub-catalog of the newly created catalog. If multiple alternative catalogs are encountered, they are saved for manual review and decision; Step (4): Manually review and decide the catalog classification of each equipment and material, and continue to adjust the hierarchical relationship of the newly created catalog tree to form a new catalog of enterprise equipment and materials; Specific steps of step S5: Step (1): traverse each equipment asset in the new directory, calculate the similarity based on the image recognition model and the text classification model, and form a list of suspected duplicates; Step (2): manually review the list of duplicates to confirm whether there are duplicates; Step (3): based on the manual review results, the entire directory tree is deduplicated to form a complete directory of enterprise equipment and materials; Step (4): based on the complete directory, the hierarchical association relationship with the above-mentioned alternative directory is updated and saved; Specific steps of step S13: Step (1): compare the extracted image recognition model or text classification model with the image recognition model and text classification model obtained in S2, and return the three sub-directories with the highest matching degree; Step (3): if there is information such as a national standard directory or an e-commerce directory, call the association relationship between the alternative directory and the enterprise directory for matching, return the three sub-directories with the highest matching degree, and then execute step (4); if not, directly execute step (4); Step (4): combine and sort the sub-directory lists obtained in step (1) and step (2) and return them.

2. The method for constructing an enterprise material classification catalog based on artificial intelligence according to claim 1 is characterized in that: The steps of image feature extraction in step S2 include: step (1): reading equipment material image files; step (2): image preprocessing; step (3): calling artificial intelligence image recognition algorithm, training and extracting features from the image preprocessing results obtained in step (2).

3. The method for constructing an enterprise material classification catalog based on artificial intelligence according to claim 1 is characterized in that: The steps of text information extraction in step S2 include: step (1): reading equipment and material text files; step (2): text preprocessing; step (3): calling artificial intelligence text classification algorithm to train the text preprocessing results obtained in step (2).

4. The method for constructing an enterprise material classification catalog based on artificial intelligence according to claim 1 is characterized in that: The step of clustering the image recognition model in step S3 includes: step (1): reading the image recognition model document; step (2): image preprocessing; step (3): calling the clustering algorithm to cluster the image preprocessing result obtained in step (2); step (4): manually reviewing the clustering result, establishing clustering labels and descriptions and associating them with the equipment and material names and archiving them; The step of clustering the text classification model in step S3 includes: step (1): reading the text classification model; step (2): text preprocessing; step (3): calling the clustering algorithm to cluster the text preprocessing results obtained in step (2); step (4): manually reviewing the clustering results, establishing clustering labels and descriptions and associating them with the equipment and material names and archiving them.