Product modeling method and related equipment

Abstract

The invention provides a product modeling method and related equipment, and relates to the technical field of computers.The method comprises the steps that information of any standardized product unit is obtained, and product keywords are determined; starting breadth-first search to query a product information set associated with the product keyword according to a product knowledge graph of the latest version; the product knowledge graph is a pre-constructed and dynamically updated knowledge graph; and performing screening according to the relationship weight between the product information sets to obtain target product information as template information of any standardized product unit information, and generating a product template. Through the constructed topological relation graph of the product information, the hidden relation between the products is positively mined, the flexibility and timeliness of product template configuration are improved, and the query efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer, and in particular, to a product modeling method and related equipment. BACKGROUND

[0002] The template of the traditional standardized product unit (SPU) needs to be generated by the merchant according to the industry in which the merchant is located. With the continuous diversification of transaction content and the acceleration of product update iteration, the association level of different products to attributes is more complex, the query time consumption of multiple traversals and multiple index layers is lengthened, and the response time linearly increases with the amount of existing data, which seriously limits the query efficiency of SPU / stock keeping unit (SKU) product or commodity configuration.

[0003] In addition, there are common attributes between different products, such as the attribute of "material" for outdoor equipment, and the attribute of "material" for mobile phones. The current attribute configuration is relatively independent and strongly dependent on manual configuration and expansion, and the same attribute is reused in different product categories, resulting in serious attribute data redundancy.

[0004] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The present disclosure provides a product modeling method and related equipment, which at least partially solves the problem of low query efficiency and low flexibility of product template configuration in the related art.

[0006] Other characteristics and advantages of the present disclosure will become apparent from the following detailed description, or will be learned by practice of the present disclosure.

[0007] In a first aspect, the embodiments in the present disclosure provide a product modeling method, the method comprising:

[0008] obtaining any one standardized product unit information, and determining a product keyword;

[0009] According to the latest version of the product knowledge graph, a breadth-first search query is started to obtain a product information set associated with the product keyword; the product knowledge graph is a pre-constructed and dynamically updated knowledge graph;

[0010] According to the relationship weight between the product information sets, the target product information is obtained as the template information of any one standardized product unit information, and a product template is generated.

[0011] In a possible embodiment, the method comprises:

[0012] obtaining product information;

[0013] taking the product information as graph nodes, constructing an association relationship between the product information, and determining a relationship weight between the product information through a graph neural network to construct a product knowledge graph.

[0014] In a possible embodiment, the relationship weight comprises a weight calculation dimension; the method further comprises:

[0015] every preset time period, determining an incremental value of different weight calculation dimensions according to incremental data;

[0016] dynamically updating the initial relationship weight based on the incremental value of the different weight calculation dimensions through the graph neural network to obtain an updated product knowledge graph.

[0017] In a possible embodiment, the incremental data comprises at least one of the following: operation information of a user, information statistical times, and incremental product information; and the weight calculation dimension comprises at least one of the following: a statistical frequency weight calculation dimension, a semantic similarity weight calculation dimension, and a user feedback weight calculation dimension.

[0018] every preset time period, determining an incremental value of different weight calculation dimensions according to incremental data, comprises:

[0019] determining the incremental value of the statistical frequency weight calculation dimension according to the information statistical times;

[0020] or, determining the incremental value of the semantic similarity weight calculation dimension according to a similarity between the incremental product information and a user comment in the operation information of the user;

[0021] or, determining the incremental value of the user feedback weight calculation dimension according to a search behavior history record and a click behavior history record in the operation information of the user.

[0022] In a possible embodiment, the product information is obtained by:

[0023] obtaining data information of the product through a plurality of data sources;

[0024] performing word segmentation and recognition on the data information through a natural language processing library to determine attribute information and category information corresponding to the product;

[0025] determining a similarity between different attribute information to determine attribute combination information;

[0026] performing data preprocessing on the data information through a data stream engine to obtain product information in a unified data format; the product information comprises the attribute information, the attribute combination information, and the category information.

[0027] In a possible embodiment, the method further comprises:

[0028] collecting satisfaction information of the user for the product template; the satisfaction information comprises at least one of the following: click rate, conversion rate;

[0029] According to the pre-defined reward function mechanism, the relationship weight of the target product information in the product knowledge graph is dynamically adjusted through the satisfaction information and the deep Q network.

[0030] In a possible embodiment, the queried product information set comprises product information of a direct graph node and product information of an indirect graph node;

[0031] According to the relationship weight between the product information sets, the target product information is obtained as the template information of any one standardized product unit information, and the product template is generated, comprising:

[0032] determining the relationship weight between the product information of the direct graph node and the product information of the indirect graph node;

[0033] The product information of the indirect graph node corresponding to the relationship weight before the preset position and the product information of the direct graph node are screened and sorted, and the target product information is obtained as the template information to generate the product template.

[0034] In a second aspect, the embodiments in the present disclosure provide a product modeling device, comprising:

[0035] A determination unit is configured to acquire any one standardized product unit information and determine a product keyword;

[0036] A query unit is configured to query a product information set associated with the product keyword according to a breadth-first search in a latest version of a product knowledge graph; the product knowledge graph is a pre-constructed and dynamically updated knowledge graph;

[0037] A modeling unit is configured to filter according to relationship weights between product information sets, and take product information with a relationship weight in a preset position as template information of any one standardized product unit information to generate a product template.

[0038] In a possible embodiment, the device further comprises:

[0039] A graph construction unit is configured to acquire product information;

[0040] The product information is taken as a graph node to construct an association relationship between the product information, and a graph neural network is used to determine relationship weights between the product information to construct a product knowledge graph.

[0041] In a possible embodiment, the device further comprises:

[0042] The first graph updating unit is configured to determine an incremental value of each dimension of different weight calculation at a preset time interval according to the incremental data.

[0043] The initial relationship weight is dynamically updated based on the incremental value of each dimension of different weight calculation through the graph neural network, and an updated product knowledge graph is obtained.

[0044] The second graph updating unit is configured to collect satisfaction information of a user for a product template, and the satisfaction information includes at least one of a click rate and a conversion rate.

[0045] The relationship weight of target product information in the product knowledge graph is dynamically adjusted according to a predefined reward function mechanism and the satisfaction information and the deep Q network.

[0046] In a third aspect, an electronic device is provided, including a processor and a memory for storing executable instructions of the processor, wherein the processor is configured to execute the method of the first aspect described above by executing the executable instructions.

[0047] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the method of the first aspect described above.

[0048] In a fifth aspect, according to another aspect of the present disclosure, a computer program product or a computer program is also provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the method of any one of the above aspects.

[0049] The product modeling method and related device provided by the embodiments of the present disclosure relate to the technical field of computers, and the method includes: obtaining any one of standardized product unit information and determining product keywords; performing a breadth-first search query on a product information set associated with the product keywords according to a latest version of a product knowledge graph; the product knowledge graph is a pre-constructed and dynamically updated knowledge graph; filtering target product information according to relationship weights between the product information sets to obtain the target product information as template information of the any one of standardized product unit information, and generating a product template. The hidden relationships between products are actively mined through the constructed topological relationship graph of product information, the flexibility and timeliness of product template configuration are improved, and the query efficiency is improved.

[0050] It should be understood that the general description above and the detailed description below are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0052] Figure 1 A flowchart of a product modeling method according to an embodiment of the present disclosure is shown;

[0053] Figure 2 A flowchart of constructing a product knowledge graph according to an embodiment of the present disclosure is shown;

[0054] Figure 3 A flowchart of obtaining product information according to an embodiment of the present disclosure is shown;

[0055] Figure 4 A flowchart of dynamically updating a product knowledge graph according to an embodiment of the present disclosure is shown;

[0056] Figure 5 A flow chart of dynamically adjusting relationship weights according to an embodiment of the present disclosure is shown;

[0057] Figure 6 A schematic diagram illustrating a product modeling process in an embodiment of the present disclosure;

[0058] Figure 7 A schematic diagram illustrating a method of generating a product template according to an embodiment of the present disclosure is shown;

[0059] Figure 8 A schematic structural diagram of a product modeling device according to an embodiment of the present disclosure is shown;

[0060] Figure 9 A schematic structural diagram of an electronic device in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0061] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0062] In addition, the accompanying drawings are only schematic and are non-limiting illustrative of the disclosure. Identical reference signs denote identical or similar parts throughout the figures. Some of the blocks in the drawings are functional entities that may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0063] The terms appearing in the embodiments of the disclosure are explained as follows.

[0064] SPU: refers to Standard Product Unit. It refers to a product as the core, including all attribute information of the product, such as brand, model, specification, etc. SPU is used to describe a collection of goods with the same characteristics.

[0065] SKU: Stock Keeping Unit, which is the smallest unit to identify each specific product. Each SKU represents an independent inventory entity, which can be different colors, sizes or configurations of specific goods.

[0066] Since the template of the traditional Standard Product Unit (SPU) needs to be generated by the merchant according to the industry in which it is located. With the continuous diversification and updating of current trade modes and transaction contents, the acceleration of product updates and iterations, the association level of attributes of different products becomes more complex, the query time consumption of multiple traversals and multiple index layers is lengthened, and the response time increases linearly with the amount of existing data, which seriously limits the query efficiency of SPU / Stock Keeping Unit (SKU) product or commodity configuration.

[0067] In addition, there will be common attributes between different products, such as "material" attribute for outdoor equipment, and "material" attribute for mobile phones. The current attribute configuration is relatively independent, not only cannot dynamically combine attributes, but also strongly depends on manual configuration and expansion, which greatly limits the efficiency of attribute configuration. At the same time, the same attribute is reused in different product categories, causing serious attribute data redundancy. With the increasing diversification and flexibility of trade, the trade barriers between different industries are constantly reduced, and the relatively solid traditional attribute configuration method is difficult to adapt to the cross-category commodity attribute cross-configuration scene, and has obvious defects in query timeliness and configuration flexibility.

[0068] The embodiments of the present disclosure provide a product modeling method and related equipment, which relate to the field of computer technology. The method includes: obtaining any standardized product unit information and determining product keywords; starting a breadth-first search query for a set of product information associated with the product keywords based on the latest version of the product knowledge graph; the product knowledge graph is a pre-built and dynamically updated knowledge graph; filtering based on the relationship weights between the product information sets, obtaining the target product information as the template information of any standardized product unit information, and generating a product template. By constructing a topological relationship diagram of product information, the hidden relationships between products are actively mined, the flexibility and timeliness of product template configuration are improved, and query efficiency is improved.

[0069] This exemplary implementation is described in detail below with reference to the accompanying drawings and examples.

[0070] First, a product modeling method is provided in an embodiment of the present disclosure. The method can be executed by any electronic device with computing and processing capabilities. In the following process, the electronic device is taken as a server as an example.

[0071] Figure 1 A flowchart of a product modeling method according to an embodiment of the present disclosure is shown. Figure 1 As shown, the product modeling method provided in the embodiment of the present disclosure includes the following steps:

[0072] S102: Obtain any standardized product unit information and determine product keywords.

[0073] In a possible embodiment, the present disclosure takes the smart device product line and the decorative wear product line as examples, and uses "new smart watch" as the standardized product unit information of the product to complete product modeling.

[0074] For example, the smart device product line covers multiple subcategories such as laptops, tablets, mobile phones, drones, etc.

[0075] These devices involve cross-domain attributes: electronic parameters: communication protocol (Zigbee / Wi-Fi), network access license, firmware version, etc.; functional attributes: built-in speaker, voice mode, NFC function, size specifications, etc.; energy efficiency: standard battery life, charging mode, battery capacity, etc.; user interaction: screen size, voice assistant compatibility (Alexa / Google Home), mobile app functions.

[0076] For example, the decorative wearable product line covers multiple subcategories such as watches, bracelets, and accessories.

[0077] These products involve cross-domain attributes: wearable devices: wristband material, waterproof level, sports monitoring mode, etc. Fashion accessories: style type, color, applicable population, etc.

[0078] In a possible embodiment, the standardized product unit information of the product "new smart watch" is input into the electronic device, and the product keywords are "new" + "smart" + "watch".

[0079] S104: According to the latest version of the product knowledge graph, the breadth first search query is started to query the product information set associated with the product keywords, and the product knowledge graph is a pre-constructed and dynamically updated knowledge graph.

[0080] In a possible embodiment, the breadth first search (BFS) query is started based on the established dynamic attribute graph to query the product information set associated with "new" + "smart" + "watch", and then "smart" automatically queries the product information set of "smart device product", and "watch" automatically queries the product information set of "watch".

[0081] In a possible embodiment, the keyword "new" does not belong to specific product information, but based on the keyword "new", product information that is mutually exclusive with "new" can be screened, such as retro style labels and the like.

[0082] Exemplarily, the product information set associated with the smart device product can include "smart", a telecommunications device network access license, a Zigbee / Wi-Fi communication protocol, and the like.

[0083] Exemplarily, the product information set associated with the watch can include "watch", a waterproof level, a watchband color, a watch dial shape, and the like.

[0084] S106: According to the relationship weight between the product information sets, the target product information is obtained as the template information of any one standardized product unit information, and the product template is generated.

[0085] Since the directly queried product information set includes a large number of information, the information can be screened and combined when the product template is generated. The product information has a relationship weight.

[0086] In a possible embodiment, the target product information can be obtained according to the relationship weight in other manners, and the specific screening manner is not limited.

[0087] Exemplarily, the product information corresponding to the relationship weight before the preset bit can be sorted according to the size of the relationship weight, and the product information corresponding to the relationship weight is taken as the target product information, that is, the target product information is screened in a TOPN manner based on the relationship weight, and the product template is generated.

[0088] Through the above manner, a more efficient product modeling manner can be realized, the knowledge graph constructed based on analysis of topological relations among product information avoids invalid traversal in a multi-level complex association query process, and while improving attribute configuration timeliness and flexibility, attribute query efficiency is greatly improved.

[0089] The product knowledge graph used in the above process is constructed in the following manner, as shown in Figure 2 Figure 2 A flowchart of constructing a product knowledge graph in the embodiment of the present disclosure is shown, including the following steps:

[0090] S202: Obtain product information.

[0091] In a possible embodiment, the product information can be obtained in a manner of data source integration and data preprocessing.

[0092] The data source integration can integrate user comments about entities, product detail description information, industry standard document information, competitor information and other unstructured data information and structured data information.

[0093] By way of example, product detail description text: support multi-language description. User comment information. Industry standard document: reference PDF format UL certification, energy star standard, etc.

[0094] The data preprocessing process is a process of processing the integrated product data information, Figure 3 A flowchart of obtaining product information in the embodiment of the present disclosure is shown, as shown in Figure 3 including the following steps:

[0095] S302: Obtain product data information through multiple data sources.

[0096] S304: Perform word segmentation and recognition on the data information through a natural language processing library to determine the attribute information and category information corresponding to the product.

[0097] In a possible embodiment, spaCy can be used to perform multi-language word segmentation and entity recognition on the data information to obtain the attribute information and category information corresponding to the product.

[0098] In a possible embodiment, when obtaining product information, a field dictionary can be constructed to merge information of various products in the same field to obtain a field product information set, and specific information expansion can be constructed for a professional field, for example, the professional term information "Telecommunications Equipment In-Network Permit" is expanded to an entity model under the intelligent device product.

[0099] ​S306: Determine the similarity between different attribute information, and determine the attribute combination information.

[0100] In a possible embodiment, similar attributes can be merged through a similarity algorithm, for example, "power interface" and "charging method" are considered as the same attribute.

[0101] The similarity algorithm can use a similarity algorithm based on Levenshtein distance.

[0102] S308: Through the data flow engine, data preprocessing is performed on the data information to obtain product information in a unified data format, and the product information includes attribute information, attribute combination information and category information.

[0103] In a possible embodiment, Apache NiFi data flow engine can be used to unify the data format of data information from different sources, which can be unified into JSON format to obtain product information, including attribute information, attribute combination information merged through the similarity algorithm and category information corresponding to the product.

[0104] Through Figure 3 , the attribute entities and their associated relationships of multi-source data are extracted, the industry barriers of the traditional SPU / SKU association configuration mode are broken, the cross-domain configuration of SPU / SKU is realized, the hidden relationships between different SPUs are analyzed in combination with other external associated information such as user comments and product descriptions, and product information in a unified format and directly used to build a product knowledge graph is obtained.

[0105] S204: Taking the product information as the graph node, the association relationship between the product information is constructed, and the relationship weight between the product information is determined through the graph neural network to construct the product knowledge graph.

[0106] In a possible embodiment, the attribute information, attribute combination information and category information in the product information are taken as the graph node, for example, the attribute information "voltage: 110-240V" is a single attribute information node; the attribute combination information "charging method = magnetic attraction charging line & wireless charging" is an attribute combination node; and the category information "mechanical watch", "smartphone" and the like are category nodes.

[0107] The association relationship between the graph nodes is defined, which is used to reflect the mutual influence between the graph nodes. For example, the edge relationship can reflect the mutual relationship such as inheritance / dependence / mutual exclusion.

[0108] The relationship weight between different graph nodes is dynamically calculated through the graph neural network. The graph neural network can directly process the graph structure data, and has higher accuracy and convenience.

[0109] The relationship weight includes a static weight and a dynamic weight. The static weight can be directly determined after determining the association between the two graph nodes (product information). For example, the category inheritance relationship between a watch and a decorative wearable can be a static weight, and the relationship weight can be set to 1.0. The attribute mutual exclusion relationship between battery power supply and wired charging can also be a static weight, and the relationship weight can be set to -1.0.

[0110] For example, the dynamic weight can be an attribute information dependency relationship between WiFi6 and the 5GHz frequency band. With the continuous development of communication technology, such a dependency relationship can change, and then the dynamic weight is calculated by the graph neural network.

[0111] In a possible embodiment, the relationship weight calculated by the graph neural network can be calculated according to different weight calculation dimensions, and various methods can be used.

[0112] For example, a multi-dimensional feature fusion method can be used, which uses feature vector splicing and a full connection layer.

[0113] For example, a multi-head attention mechanism can be used to design an independent attention head for each weight calculation dimension, and the outputs of each attention head are aggregated in the weighted output.

[0114] For example, a gating mechanism can be used to control the weight of the dynamic dimension.

[0115] In a possible embodiment, the queried product information set includes product information of a direct graph node and product information of an indirect graph node.

[0116] The product template can be generated in the following manner, which can include: determining the relationship weight between the product information of the direct graph node and the product information of the indirect graph node; screening and sorting the product information of the indirect graph node corresponding to the relationship weight before the preset position, and the product information of the direct graph node, to obtain target product information as template information, and generating a product template.

[0117] Based on the product keyword query and the direct graph node directly corresponding to the product keyword, the associated indirect graph node of the direct graph node is determined. For example, the product keywords are "new" + "smart" + "watch", and the direct graph node is "smart" + "watch".

[0118] Based on the direct graph node, the associated indirect graph node is determined. The direct graph node "smart" can include a telecommunications equipment network access license, a Zigbee / Wi-Fi communication protocol, and the like.

[0119] The direct graph node of the watch can include a waterproof level, a watchband color, a watch dial shape, and the like, respectively.

[0120] In this way, by extracting and mining attribute entities and their associated relationships from multi-source data, the industry barriers of the traditional SPU / SKU association configuration mode are broken, and cross-domain configuration of SPU / SKU is realized.

[0121] Figure 4 A flowchart of dynamically updating a product knowledge graph in the embodiments of the present disclosure is shown, as shown in Figure 4 The flowchart includes the following steps:

[0122] S402: At every preset time interval, determine the increment value of each weight calculation dimension according to the incremental data.

[0123] S404: Dynamically update the initial relationship weight based on the increment value of each weight calculation dimension through a graph neural network to obtain an updated product knowledge graph.

[0124] In a possible embodiment, an incremental update mechanism is used for the knowledge graph, the preset time interval can be one hour or two hours, the increment value of each weight calculation dimension is determined, and the relationship weight value is refreshed in real time and dynamically.

[0125] In a possible embodiment, the incremental data includes at least one of the following: user operation information, information statistical frequency, and incremental product information.

[0126] In a possible embodiment, the weight calculation dimension includes at least one of the following: a statistical frequency weight calculation dimension, a semantic similarity weight calculation dimension, and a user feedback weight calculation dimension.

[0127] For example, the increment value of the statistical frequency weight calculation dimension can be determined according to the information statistical frequency, and the increment value of the statistical frequency weight calculation dimension can be determined according to the number of occurrences of the product information in the unified commodity category SPU.

[0128] For example, the increment value of the semantic similarity weight calculation dimension can be determined according to the similarity between the incremental product information and the user comments in the user operation information. The cosine similarity of the user comments and the incremental product information (for example, the description text of the product information) can be calculated using Sentence-BERT, and the cosine similarity can be used as the increment value of the semantic similarity weight calculation dimension.

[0129] For example, the increment value of the user feedback weight calculation dimension can be determined according to the search behavior history record and the click behavior history record in the user operation information.

[0130] Further explanation is made by taking the information statistical times as an example. In the construction of the product knowledge graph, the relationship weight is calculated by the GNN, and the construction of the knowledge graph is completed. In the incremental updating process, if the information statistical times of a certain product information is high, the incremental value calculated can be a positive number and the value is greater than the original value. If the information statistical times of a certain product information is low, the incremental value calculated can be a negative number and the value is less than the original value, thereby affecting the statistical frequency weight calculation dimension. Therefore, when calculating the maximum weight value, the proportion of the statistical frequency weight calculation dimension is increased, and the relationship weight is dynamically updated.

[0131] In the above manner, the GNN is used to dynamically calculate the weight between attributes, the dynamic adjustment of the attribute weight is realized, a large amount of data redundancy caused by the same attribute associated with multiple products is effectively removed, the relationship weight is calculated in real time according to the incremental data, the dynamic adjustment of the attribute weight is realized, the invalid traversal in the multi-level complex association query process can be avoided, the attribute configuration timeliness and flexibility are improved, and the attribute query efficiency is greatly improved.

[0132] Figure 5 A flowchart of dynamically adjusting the relationship weight in the embodiment of the present disclosure is shown, as shown in Figure 5 The flowchart includes the following steps:

[0133] S502: Collect the satisfaction information of users for the product template. The satisfaction information includes at least one of the following: click rate, conversion rate.

[0134] S504: According to the pre-defined reward function mechanism, the relationship weight of the target product information in the product knowledge graph is dynamically adjusted by the satisfaction information and the deep Q network.

[0135] In a possible embodiment, after the product template is generated, the user click rate, conversion rate and the like are collected, the reward function mechanism is defined, and the deep Q network (DQN) is used to realize the continuous optimization of the dynamic attribute graph generation template.

[0136] For example, if the user click rate and conversion rate are high (greater than a preset value) after the product template is generated, for example, multiple users click the "like" evaluation after the product template A is generated, it indicates that the product template generated based on the relationship weight in the product knowledge graph is reasonable. Therefore, according to the reward function mechanism, the relationship weight of the target product information in the product knowledge graph can be increased by the DQN network.

[0137] Exemplarily, if the click rate and conversion rate of the user are low (greater than a preset value) after the product template is generated, for example, a plurality of users delete product information A in the target product information from the product template, it indicates that the product template corresponding to the product information A as the standardized product unit information is unreasonable, and then according to the reward function mechanism, the relationship weight of the product information A in the product knowledge graph is reduced through the dynamic adjustment of the DQN network. In the subsequent generation of the product template, the product information A will be filtered out and will not appear in the product template.

[0138] Figure 6 A schematic diagram of a product modeling process in the embodiment of the present disclosure is shown, as shown in Figure 6 The product modeling process includes three processes, which are respectively: a process of constructing a product knowledge graph, a process of dynamically updating the product knowledge graph, and a process of generating a product template.

[0139] The process of constructing the product knowledge graph is to collect data information from a plurality of data sources, and then pre-process to obtain structured product information with unified format. The product information is defined as a graph node, and the edge relationship is defined. The relationship weight between the product information is determined to complete the construction of the product knowledge graph.

[0140] The process of dynamically updating the product knowledge graph is to obtain incremental data generated in real time, and use an incremental updating mechanism to update the constructed product knowledge graph with the incremental data, and dynamically calculate the relationship weight.

[0141] The process of dynamically updating the product knowledge graph further includes: updating the relationship weight of the target product information included in the product template in the product knowledge graph according to the satisfaction information of the user to the product template.

[0142] The process of generating the product template is to obtain the standardized product unit information, determine the product keyword, query the product information through the product knowledge graph, and filter to generate the product template.

[0143] Through the above process, in the process of dynamically updating the product knowledge graph, the hidden relationship between different SPUs is analyzed in combination with other external associated information such as user comments and product descriptions, and this relationship is brought into the attribute relationship of dynamic self-adjustment of weight assignment, thereby forming the attribute dynamic association and optimization under the cross-domain condition, to realize the removal of attribute information redundancy under the cross-domain SPU / SKU configuration management, improve the query time consumption efficiency, and optimize the attribute configuration flexibility.

[0144] The product is taken as an entity, hidden relationships between product entities are mined by combining user reviews, product descriptions, click correlations and other information, barriers between product entities in the traditional mode are broken, and a basic channel for cross-domain SPU / SKU configuration is opened. It is proposed to construct a set of product information topology graph based on the attribute information under the existing various product classifications, and to realize the weight calculation of each attribute under different product conditions by combining related algorithms.

[0145] Based on the above-mentioned mining of hidden information between entities and the topology of weight assignment attributes, the dynamic generation of product correlation attribute templates can be further improved, and the consideration of time sequence dimension attribute topology weight change is introduced. The weight change is quantified through an incremental update mechanism, the change trend of product information correlation is recorded, so as to reduce attribute information redundancy and shorten attribute query time. This flexible configuration method can effectively improve the flexibility and self-adaptability of product attribute configuration.

[0146] Figure 7 A schematic diagram of generating a product template in the embodiment of the present disclosure is shown, as shown in Figure 7 As shown, by taking the product knowledge graph as a dynamic graph engine, the standardized product unit information of the new smart watch can be directly obtained by inputting the dynamic graph engine. The product template can include: the NFC function corresponding to the "smart" term, the charging method, the screen material, the screen resolution, the standby time, the applicable population, the battery capacity, the telecom equipment network access license, etc. The watchband material, watch buckle type, watchband color, watch dial shape, waterproof level, watch mirror material, transparent population, style, etc. corresponding to the "watch" term. When generating the style, since the input is a new model, the new style can be labeled, or the retro style can be removed.

[0147] The product template generated above can be generated according to the product knowledge graph of the most initial version.

[0148] After the incremental data is dynamically added and the relationship weight is adjusted through the incremental update mechanism, some product information with high relationship weight after updating can also be output. For example, the output of the product knowledge graph last year and this year will be different, and the output of the current product knowledge graph can also include health detection, motion detection, voice method, system type, etc. Figure 7 The product template output in the embodiment of the present disclosure is taken as an example.

[0149] Based on the same inventive concept, the present disclosure also provides a product modeling device, as follows. Since the principles of the device embodiments solve problems similar to the above-mentioned method embodiments, the implementation of the device embodiments can be referred to the implementation of the above-mentioned method embodiments, and the repeated parts will not be repeated.

[0150] Figure 8A structural schematic diagram of a product modeling device in an embodiment of the present disclosure is shown in FIG. 8, and the product modeling device 80 includes: Figure 8

[0151] A determining unit 801 is configured to acquire any one of standardized product unit information and determine a product keyword. A querying unit 802 is configured to query a product information set associated with the product keyword according to a breadth-first search in a latest version of a product knowledge graph. The product knowledge graph is a pre-constructed and dynamically updated knowledge graph. A modeling unit 803 is configured to filter according to relationship weights between product information sets, and take product information with a relationship weight ranked in a preset position as template information of any one of standardized product unit information, and generate a product template.

[0152] Those skilled in the art can understand that each aspect of the present disclosure can be implemented as a system, a method or a program product. Therefore, each aspect of the present disclosure can be specifically implemented as follows: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combined with hardware and software aspects, which can be collectively referred to as "circuitry", "module" or "system" here.

[0153] The electronic device 900 according to this embodiment of the present disclosure will be described below with reference to Figure 9 Figure 9 The electronic device 900 shown is merely an example, and should not impose any limitation on the functions and use range of the embodiments of the present disclosure.

[0154] As shown in Figure 9 The components of the electronic device 900 can include, but are not limited to, the above-mentioned at least one processing unit 910, the above-mentioned at least one storage unit 920, and a bus 930 connecting different system components, including the storage unit 920 and the processing unit 910.

[0155] The storage unit stores program code, which can be executed by the processing unit 910, so that the processing unit 910 performs the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of the present specification. For example, the processing unit 910 can execute the steps of any one of the above method embodiments.

[0156] The storage unit 920 can include a readable medium in the form of a volatile storage unit, such as a random access memory (RAM) 9201 and / or a cache memory unit 9202, and can further include a read-only memory (ROM) 9203.

[0157] ​​The storage unit 920 can also include the programs / utilities 9204 having a set (at least one) of program modules 9205, such as an operating system, one or more application programs, other program modules, and program data, and each or a combination thereof can include an implementation of a network environment.

[0158] The bus 930 can be representative of one or more of several types of bus structures, including a storage unit bus or bus controller, a peripheral bus, a graphics acceleration port, a processing unit bus, or a local bus using any of a variety of bus architectures.

[0159] The electronic device 900 can also communicate with one or more external devices 940 such as a keyboard or pointing device, a Bluetooth device, etc.; other devices such as a storage device or an external effects device; and / or one or more devices that enable a user to interact with the electronic device 900; and / or one or more devices that enable the electronic device 900 to communicate with one or more other computing devices. Such communication can be facilitated by an Input / Output (I / O) interface 950. Still yet, the electronic device 900 can communicate with one or more networks, such as a local area network (LAN), a wide area network (WAN), and / or the public network, such as the Internet, via a network adapter 960. As depicted, the network adapter 960 communicates with the other components of the electronic device 900 via the bus 930. It should be appreciated that although the network adapter 960 is depicted as a single component, the network adapter 960 can comprise two or more components that operate together to facilitate the communication of the electronic device 900 with one or more networks. Although not shown, it should be understood that the electronic device 900 can include other hardware and / or software components, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0160] From the above description of the embodiments, those skilled in the art will readily perceive that the example embodiments described herein can be implemented by software and / or by software in combination with the necessary hardware. Accordingly, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.) or a network, and includes a number of instructions to make a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) execute the methods according to the embodiments of the present disclosure.

[0161] In particular, according to the embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer program product or a computer program, which includes computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method in the above embodiments.

[0162] In the exemplary embodiments of the present disclosure, a computer readable storage medium is also provided, which can be a readable signal medium or a readable storage medium. A program product capable of implementing the above method of the present disclosure is stored thereon. In some possible implementations, various aspects of the present disclosure can also be implemented in the form of a program product, which includes program codes for causing the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of the specification when the program product is run on the terminal device.

[0163] More specific examples of the computer readable storage medium in the present disclosure can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any appropriate combination of the above.

[0164] In the present disclosure, the computer readable storage medium can include a data signal propagated in a baseband or as a part of a carrier wave, in which readable program codes are borne. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal, or any appropriate combination of the above. The readable signal medium can also be any readable medium other than the readable storage medium, which can send, propagate or transmit programs for use by or in connection with an instruction execution system, apparatus or device.

[0165] Optionally, the program codes contained on the computer readable storage medium can be transmitted by any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any appropriate combination of the above.

[0166] In particular embodiments, the program code utilized by the program code instructions can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++, or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computing device, partly on the user's computing device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, such as through the Internet using an Internet Service Provider. The application programming interface (API) server 106 can be used to facilitate the connection between the client device 102 and the remote computing device.

[0167] It should be noted that, although several modules or units of devices for action execution are mentioned in the foregoing detailed description, such division is not mandatory. Indeed, features and functionalities of two or more modules or units described above can be embodied in one module or unit according to embodiments of the present disclosure. Conversely, features and functionalities of one module or unit described above can be further divided into multiple modules or units.

[0168] Furthermore, although various steps of the methods in the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in that particular order, or that all of the steps must be performed to achieve the desired results. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into one step, one step can be split into multiple steps, and / or the like.

[0169] From the above description of the embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software, or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium, such as a CD-ROM, a USB flash drive, a mobile hard disk, or the like, or on a network, and includes a number of instructions to make a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) execute the methods according to the embodiments of the present disclosure.

[0170] Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the features disclosed herein. It is intended that the disclosure be construed as including any patents, patent applications, publications, publications, or other disclosure of complementary subject matter that is within the scope of the disclosure. It is intended that the specification and examples be considered exemplary only, with the true scope and spirit of the disclosure being indicated by the following claims.

Claims

1. A product modeling method, characterized in that: The method comprises: Obtain any standardized product unit information and determine product keywords; Initiate a breadth-first search to query a set of product information associated with the product keyword in the latest version of the product knowledge graph, which is a pre-built and dynamically updated knowledge graph. Screening is performed according to the relationship weights among the product information sets to obtain target product information as template information of any standardized product unit information, thereby generating a product template.

2. The method according to claim 1, characterized in that The method comprises: Get product information; Using product information as graph nodes, the association relationship between the product information is constructed, and the relationship weights between the product information are determined through a graph neural network to construct the product knowledge graph.

3. The method according to claim 2, characterized in that The relationship weight includes a weight calculation dimension; The method further comprises: At every preset time period, the incremental values ​​of different weight calculation dimensions are determined based on the incremental data; Through the graph neural network, the incremental value of the dimension is calculated based on different weights, and the relationship weight is dynamically updated to obtain an updated product knowledge graph.

4. The method according to claim 3, characterized in that The incremental data includes at least one of the following: user operation information, information statistics times, incremental product information; the weight calculation dimension includes at least one of the following: statistical frequency weight calculation dimension, semantic similarity weight calculation dimension and user feedback weight calculation dimension; The step of determining incremental values ​​of different weight calculation dimensions based on incremental data at every preset time period includes: Determine the incremental value of the statistical frequency weight calculation dimension according to the number of times the information is counted; Alternatively, determining an incremental value of a semantic similarity weight calculation dimension based on the similarity between the incremental product information and the user comments in the user's operation information; Alternatively, the incremental value of the user feedback weight calculation dimension is determined based on the search behavior history record and the click behavior history record in the user's operation information.

5. The method according to claim 2, characterized in that The obtaining of product information includes: Obtain product data information through multiple data sources; Segment and identify the data information through a natural language processing library to determine the attribute information and category information corresponding to the product; Determine the similarity between different attribute information and determine attribute combination information; The data information is pre-processed by a data flow engine to obtain product information in a unified data format; the product information includes: attribute information, attribute combination information and category information.

6. The method according to claim 1, characterized in that The method further comprises: Collecting user satisfaction information for the product template; the satisfaction information includes at least one of the following: click-through rate, conversion rate; According to a predefined reward function mechanism, the relationship weight of the target product information in the product knowledge graph is dynamically adjusted through the satisfaction information and the deep Q network.

7. The method according to claim 1, characterized in that The product information set retrieved includes: product information of direct graph nodes and product information of indirect graph nodes; The step of screening according to the relationship weights between the product information sets to obtain target product information as template information of any standardized product unit information and generating a product template includes: Determining a relationship weight between the product information of the direct graph node and the product information of the indirect graph node; The product information of the indirect graph nodes corresponding to the relationship weights ranked before the preset positions and the product information of the direct graph nodes are filtered to obtain the target product information as the template information and generate the product template.

8. A product modeling device, characterized in that: include: Determine unit, used to obtain any standardized product unit information and determine product keywords; A query unit, configured to initiate a breadth-first search query for a set of product information associated with the product keyword based on the latest version of the product knowledge graph; the product knowledge graph is a pre-built and dynamically updated knowledge graph; The modeling unit is used to screen according to the relationship weights between the product information sets, and use the product information with the relationship weight ranked in the front preset position as the template information of any standardized product unit information to generate a product template.

9. An electronic device, characterized in that: include: processor; as well as a memory for storing executable instructions of the processor; The processor is configured to perform the method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product comprising: A computer program or instruction, characterized in that when the computer program or instruction is executed by a processor, it implements the method according to any one of claims 1 to 7.

Images