Commodity selling point knowledge graph construction method and electronic equipment

Abstract

The embodiment of the invention provides a commodity selling point knowledge graph construction method and electronic equipment, and relates to the technical field of the Internet, and the method comprises the steps: obtaining multi-dimensional commodity information of a target commodity, and extracting a selling point keyword used for representing a commodity selling point of the target commodity from the multi-dimensional commodity information; determining a selling point intention of a commodity selling point according to the selling point keyword; generating an intention system of the target commodity according to the selling point keyword and the selling point intention; the intention system comprises an association relationship between the selling point keyword and the selling point intention; constructing a selling point knowledge graph according to the intention system; the selling point knowledge graph is used for representing a semantic relationship among the target commodity, the selling point keyword and the selling point intention. According to the application, the commodity selling points can be converted into an analyzable, associable and manageable structured system from scattered information, the credibility of the commodity selling points is ensured, and the selling point knowledge graph convenient for automatic management and updating is formed.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method for constructing a knowledge graph of product selling points and an electronic device. Background Art

[0002] Currently, a deep-seated conflict exists within online marketing between the richness of product attributes and the credibility of selling points. For example, in the catering industry, while local lifestyle platforms have built a basic ingredient knowledge graph, they lack a mechanism for verifying the evolution from product features to selling points, resulting in low credibility of the association between product attributes and selling points. Related technologies rely on manual rules to extract selling point terms from relevant text, such as using NER (Named Entity Recognition) models to extract entity terms as selling points. While this approach can extract basic attributes such as ingredients and flavors, it fails to establish a link between attributes and selling points, resulting in low credibility of selling points. Furthermore, it struggles to capture implicit selling point associations in unstructured text, such as the relationship between ingredients and their origins. Therefore, there is a need for a method that can automatically construct a selling point graph, enabling an automated evolution from basic attributes to selling point advantages. Summary of the Invention

[0003] The embodiments of the present application provide a method for constructing a product selling point knowledge graph and an electronic device to alleviate or solve one or more technical problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present application provides a method for constructing a product selling point knowledge graph, including: Acquire multi-dimensional product information of a target product, and extract selling point keywords used to represent the selling points of the target product from the multi-dimensional product information; the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information; Determining the selling point intention of the product selling point based on the selling point keywords; generating an intention system for the target product based on the selling point keywords and the selling point intentions; the intention system including the association relationship between the selling point keywords and the selling point intentions; A selling point knowledge graph is constructed based on the intention system; the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intentions.

[0005] Optionally, determining the selling point intention of the product selling point based on the selling point keyword includes: Parsing the selling point keywords to obtain a first attribute word and a first positive word of the product selling point; the first attribute word is used to represent the product attribute of the target product; the first positive word is used to represent the value of the target product; Determining, based on the first attribute word, a selling point attribute category corresponding to the selling point keyword; Clustering the selling point keywords according to the selling point attribute categories and the first positive words to obtain at least one selling point cluster; the selling point attribute categories corresponding to the selling point keywords in the same selling point cluster satisfy a preset similarity condition; Determine the selling point intentions corresponding to each of the selling point clusters.

[0006] Optionally, generating the intention system of the target product based on the selling point keywords and the selling point intentions includes: For any of the selling point clusters, semantic expansion is performed on the selling point keywords in the selling point cluster to obtain selling point rewriting words of the selling point keywords; the selling point rewriting words include synonyms and / or hyponyms; The intention system is generated according to the selling point keywords in the selling point cluster, the selling point rewriting words corresponding to the selling point keywords, and the selling point intentions corresponding to the selling point cluster.

[0007] Optionally, determining the selling point attribute category of the product selling point based on the first attribute word includes: Matching the first attribute word with a pre-configured product attribute library; the product attribute library includes associations between attribute words and attribute categories; In a case where the first attribute word is included in the commodity attribute library, determining the attribute category associated with the first attribute word as the selling point attribute category; In the case that the first attribute word is not included in the commodity attribute library, the first attribute word is classified using a pre-trained large language model to obtain the selling point attribute category.

[0008] Optionally, after generating the intention system of the target product based on the selling point keywords and the selling point intention, the method further includes: Verify the specified association relationship in the intent system; the specified association relationship includes at least one of the following: the association relationship between the selling point keyword and the industry to which the product belongs, the hierarchical relationship between multiple selling point keywords, the association relationship between the selling point keyword and the target product, and the association relationship between the selling point keyword and the selling point intent; When it is verified that the designated association relationship is an abnormal association relationship, the abnormal association relationship in the intention system is corrected.

[0009] Optionally, the selling point cluster includes multiple; After generating the target product intention system based on the selling point keywords and the selling point intention, the method further includes: Determining the intention priority of the selling point intentions corresponding to each of the selling point clusters; In the intent system, a priority tag is added to the selling point intent; the priority tag is used to indicate the intent priority of the selling point intent.

[0010] Optionally, determining the intention priority of the selling point intentions corresponding to each of the selling point clusters includes: Determining indicator data corresponding to the selling point intention; the indicator data includes at least one of the following: the supply of related products corresponding to the selling point intention, the visit purchase rate of the related products, the add-on purchase rate of the related products, and the accuracy rate of the association between the selling point intention and the related products; Determining a priority score corresponding to the selling point intention based on the indicator data; Determine the intention priority of the selling point intention according to the priority score.

[0011] Optionally, constructing a selling point knowledge graph based on the intent system includes: Semantically aligning a first entity object in the intent system to obtain an aligned second entity object; the first entity object includes the target product, the selling point keyword, and the selling point intent; Determining a semantic relationship between each of the second entity objects; the semantic relationship includes at least one of the following: an inclusion relationship, a derivation relationship, a mutual exclusion relationship, and a similarity relationship; The second entity object is converted into a graph node of the selling point knowledge graph, and directed edges are established between the graph nodes according to the semantic relationship to generate the selling point knowledge graph.

[0012] Optionally, after constructing the selling point knowledge graph according to the intention system, the method further includes: Detecting whether there is abnormal association information in the selling point knowledge graph; the abnormal association information includes at least one of the following: an abnormal association relationship between the selling point keyword and the target product, a semantic contradiction between multiple selling point keywords associated with the target product, and a semantic contradiction between multiple selling point intentions associated with the target product; The selling point knowledge graph is modified according to the abnormal association information.

[0013] Optionally, after constructing the selling point knowledge graph according to the intention system, the method further includes: In response to extracting a new selling point keyword of the target product, determining a confidence level of the new selling point keyword; When the confidence level is greater than or equal to a preset confidence threshold, the selling point knowledge graph is updated based on the new selling point keyword.

[0014] Optionally, after constructing the selling point knowledge graph according to the intention system, the method further includes: Receive a display request for the intent system of a target product; the display request carries product identification information of the target product; Based on the display request, the product identification information is matched with the selling point knowledge graph to obtain the intent system corresponding to the target product; Display the intent system corresponding to the target product.

[0015] Optionally, the display of the intention system corresponding to the target product includes: From the multiple selling point intentions corresponding to the target product, screening at least one selling point intention whose target group index meets a preset condition; The at least one selling point intention and selling point keywords associated with the at least one selling point intention are displayed.

[0016] Optionally, extracting selling point keywords for representing the selling points of the target product from the multi-dimensional product information includes: Extracting candidate texts including selling point information of the target product from the multi-dimensional product information; Inputting the candidate text into a pre-trained artificial intelligence model for text analysis, so as to extract a plurality of selling point tags from the candidate text through the artificial intelligence model; The multiple selling point tags are matched with pre-configured selling point types, and based on the matching results, the selling point tags matching the selling point types are determined as the selling point keywords.

[0017] In a second aspect, an embodiment of the present application provides a device for constructing a product selling point knowledge graph, including: an acquisition module, configured to acquire multi-dimensional product information of a target product and extract selling point keywords representing the selling points of the target product from the multi-dimensional product information; the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information; a determination module, configured to determine the selling point intention of the product selling point based on the selling point keywords; A generating module, configured to generate an intention system for the target product based on the selling point keywords and the selling point intentions; the intention system includes an association relationship between the selling point keywords and the selling point intentions; A construction module is used to construct a selling point knowledge graph based on the intention system; the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intentions.

[0018] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor implements any method of the embodiments of the present application when executing the computer program.

[0019] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method of any one of the embodiments of the present application is implemented.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements any method of the embodiments of the present application when executed by a processor.

[0021] According to the technical solution of the embodiment of the present application, by obtaining the multi-dimensional product information of the target product, the selling point keywords used to represent the product selling points of the target product are extracted from the multi-dimensional product information, and the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information. According to the selling point keywords, the selling point intention of the product selling point is determined, and according to the selling point keywords and the selling point intention, the intention system of the target product is generated, and the intention system includes the association relationship between the selling point keywords and the selling point intention. Then, a selling point knowledge graph is constructed based on the intention system, and the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intention. It can be seen that this application can not only capture the product selling points implied in the unstructured text by extracting the keywords related to the selling points in the multi-dimensional product information of the target product, and using the extracted keywords as the basis for selling point mining to generate the intention system, but also clarify the association path from the product selling points to the selling point intention, thereby realizing the transformation of the product selling points from scattered text into a structured system that can be analyzed, associated, and managed, thereby ensuring the credibility of the product selling points. In addition, by constructing a selling point knowledge graph based on the intent system, the semantic relationship between the target product, selling point keywords and selling point intentions is converted into a logically clear network graph structure, which facilitates automated management and updating of the graph and is more adaptable to scenarios with higher requirements for the timeliness of selling points in a dynamic market environment.

[0022] The technical solution of this application can be applied to the transaction and delivery services of instant e-commerce platforms, such as Taobao Flash Purchase, Taoxianda, Ele.me takeout and retail, etc.

[0023] The above description is only an overview of the technical solution of this application. In order to more clearly understand the technical means of this application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of this application more obvious and easy to understand, the specific implementation methods of this application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar components or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments according to the present application and should not be regarded as limiting the scope of the present application.

[0025] Figure 1 A flowchart of a method for constructing a product selling point knowledge graph provided by an embodiment of the present application is shown; Figure 2 An example diagram of the determination result of the selling point intention provided by an embodiment of the present application is shown; Figure 3 An example diagram showing the construction of a hierarchical relationship in the method for constructing a product selling point knowledge graph provided by an embodiment of the present application is shown; Figure 4 An interface diagram of the selling point knowledge graph provided by an embodiment of the present application is shown; Figure 5 A block diagram of a device for constructing a product selling point knowledge graph provided by an embodiment of the present application is shown; Figure 6 A block diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present application. Therefore, the drawings and description are to be regarded as illustrative in nature and not restrictive.

[0027] To facilitate understanding of the technical solutions of the embodiments of the present application, the following describes the related technologies of the embodiments of the present application. The following related technologies can be combined with the technical solutions of the embodiments of the present application as optional solutions, and all of them fall within the scope of protection of the embodiments of the present application.

[0028] The following terms will be used in the following text: FABE Principle: A product marketing methodology consisting of product features (Feature), advantages (Advantage), benefits (Benefit), and evidence (Evidence). This principle effectively stimulates users' desire to buy products through structured expression. Its core is to transform product features into user-perceivable value.

[0029] CPV, or Category-Property-Value, refers to the "category-property-value" system. The CPV system helps structure product information, making it easier for merchants to list products and for consumers to quickly understand product attributes. It also provides a foundational management and analysis dimension for product governance and operational management on e-commerce platforms.

[0030] The RoBERTa-LSTM-CRF model is a natural language processing model that integrates a pre-trained language model, a recurrent neural network, and a conditional random field. It has the ability to train basic attribute recognition across categories, such as identifying basic attributes such as ingredients, staple foods, and flavors.

[0031] This application addresses several major pain points in the field of online marketing, especially in the catering industry, such as the low standardization of selling point systems, poor manual construction efficiency, and lack of data verification. It proposes a dynamic knowledge graph construction system based on the FABE principle. In the dynamic knowledge graph construction system, by integrating multi-source heterogeneous data (including product description information, UGC (User-Generated Content), search terms, etc.) with CPV, and combining the semantic understanding capabilities and knowledge reasoning framework of a large language model, automated evolutionary verification from basic product attributes to selling point advantages is achieved. Compared with traditional manual labeling solutions, the efficiency of selling point evolution verification and the coverage of selling points have been greatly improved, providing the catering industry with an explainable selling point quantitative evaluation system.

[0032] It should be noted that the above-mentioned application scenarios or application examples provided in the embodiments of this application are for ease of understanding, and the embodiments of this application do not specifically limit the application of the technical solution. In addition, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation portals for users to choose to authorize or refuse.

[0033] The following describes in detail the technical solution of this application and how it solves the aforementioned technical problems using specific embodiments. The several specific embodiments listed can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. The following describes the embodiments of this application in detail with reference to the accompanying drawings.

[0034] Figure 1 The flowchart of the method for constructing a commodity selling point knowledge graph provided by the embodiment of the present application is shown as follows: Figure 1 As shown, the method may include step S101, step S102, step S103 and step S104.

[0035] Step S101 : acquiring multi-dimensional product information of a target product, and extracting selling point keywords for representing the product selling points of the target product from the multi-dimensional product information.

[0036] Among them, multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, product review information, and SKU specification data. Product attribute information refers to the basic attributes of the product itself. For example, a restaurant's slogan "Century-old Brand" is product attribute information, and the selling point keyword "old brand" can be extracted from this product attribute information. Product description information refers to the description of the product's characteristics. For example, the product description information corresponding to the product "shrimp dumplings" is "These shrimp dumplings are made with freshly peeled shrimp that day, chewy and juicy." Selling point keywords such as "freshly peeled," "shrimp," and "chewy" can be extracted from this product description information. Product image information refers to images related to the product. When extracting selling point keywords from product image information, any existing image recognition technology can be used to identify product images to identify visual features related to the product from the product images, and then convert the visual features into text as selling point keywords. For example, product photos, drawings, and posters can be used. For example, image recognition can be performed on a photo of "shrimp dumplings" to identify visual features such as "shrimp" and "transparent skin," and these visual features can be converted into selling point keywords such as "shrimp filling" and "crystal skin." Product review information refers to user reviews of products, or user-generated content (UGC). To extract selling point keywords from product review information, the product review text can be segmented and the keywords extracted from the segmented words can be used as selling point keywords. SKU specification data refers to structured product attribute information.

[0037] Selling point keywords can be understood as the core feature phrases of the product, such as "0 calorie sugar", "freshly picked strawberries", "Korean hot sauce", etc.

[0038] Step S102: determining the selling point intention of the product selling point based on the selling point keywords.

[0039] Selling point intent refers to a user's potential demand or perception of a product's selling point. Essentially, it's the user's interpretation of their needs for the selling point. For example, the selling point keyword "0 calorie sugar" could correspond to "low calorie" or "healthy and worry-free." The selling point intent for the selling point keyword "freshly picked strawberries" could correspond to "fresh" or "delicious taste."

[0040] The selling point intention of a product's selling point is the selling point intention corresponding to the selling point keyword. The purpose is to accurately match the selling point keyword to the user's real intention to resolve the semantic difference between the merchant description and user comments. For example, the merchant description is "handmade" and the user comment is "freshly made".

[0041] Step S103 : Generate an intention system for the target product based on the selling point keywords and the selling point intentions. The intention system includes associations between the selling point keywords and the selling point intentions.

[0042] In the target product's intent system, multiple selling point keywords can correspond to the same or different selling point intents. The purpose of generating an intent system is to integrate scattered selling point keywords into analyzable, manageable, and reusable structured information, so that merchants or platforms can more clearly understand the selling point characteristics of their products.

[0043] Step S104: construct a selling point knowledge graph based on the intent system. The selling point knowledge graph is used to represent the semantic relationship between the target product, selling point keywords, and selling point intent.

[0044] The purpose of constructing a selling point knowledge graph is to establish a network structure that can be dynamically updated and adapted to changing scenarios. This includes entities and relationships between them, clearly reflecting the connections between them. In this embodiment, an entity can include at least one of the following: product identification information, selling point keywords, and selling point intent for the target product. Product identification information can include at least one of the product name, unique product identifier, and unique product code.

[0045] According to the technical solution of the embodiment of the present application, by obtaining the multi-dimensional product information of the target product, the selling point keywords used to represent the product selling points of the target product are extracted from the multi-dimensional product information, and the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information. According to the selling point keywords, the selling point intention of the product selling point is determined, and according to the selling point keywords and the selling point intention, the intention system of the target product is generated, and the intention system includes the association relationship between the selling point keywords and the selling point intention. Then, a selling point knowledge graph is constructed based on the intention system, and the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intention. It can be seen that this application can not only capture the product selling points implied in the unstructured text by extracting the keywords related to the selling points in the multi-dimensional product information of the target product, and using the extracted keywords as the basis for selling point mining to generate the intention system, but also clarify the association path from the product selling points to the selling point intention, thereby realizing the transformation of the product selling points from scattered text into a structured system that can be analyzed, associated, and managed, thereby ensuring the credibility of the product selling points. In addition, by constructing a selling point knowledge graph based on the intent system, the semantic relationship between the target product, selling point keywords and selling point intentions is converted into a logically clear network graph structure, which facilitates automated management and updating of the graph and is more adaptable to scenarios with higher requirements for the timeliness of selling points in a dynamic market environment.

[0046] In some embodiments, when determining the selling point intention of a product selling point based on the selling point keyword (ie, executing step S102 ), the following steps A1 to A4 may be executed.

[0047] Step A1: parse the selling point keywords to obtain the first attribute word and the first positive word of the product selling point.

[0048] The first attribute word is used to indicate the attributes of the target product, such as ingredients, craftsmanship, origin, etc. The first positive word is used to indicate the value and advantages of the target product, such as "imported" or "freshly made".

[0049] For example, if we extract the selling point keyword "imported rice cakes," we can analyze it to separate the first attribute word "rice cakes" and the first positive word "imported." For another example, if we analyze the selling point keyword "freshly baked bread," we can analyze it to separate the first attribute word "bread" and the first positive word "freshly baked."

[0050] Optionally, the selling point keywords include multiple attribute words and multiple positive words. According to the occurrence frequency of each attribute word and the occurrence frequency of each positive word, the attribute word with higher occurrence frequency can be screened as the first attribute word, and the positive word with higher occurrence frequency can be screened as the first positive word.

[0051] By breaking down the positive words and attribute words from the selling point keywords, we can clearly identify the advantages and basic attributes of the product selling points, providing a basis for subsequent selling point clustering.

[0052] Step A2: Determine the selling point attribute category corresponding to the selling point keyword based on the first attribute word.

[0053] Optionally, during step A2, the first attribute word may be matched with a preconfigured product attribute library, which includes associations between attribute words and attribute categories. The product attribute library may be a CPV system, which stores multiple CPV relationships in a "category-attribute-attribute value" structure, such as "food-ingredients-rice cake" and "snacks-origin-imported."

[0054] If the first attribute word is included in the product attribute library, the attribute category associated with the first attribute word is determined to be the selling point attribute category. If the first attribute word is not included in the product attribute library, the first attribute word is classified using a pre-trained large language model to obtain the selling point attribute category.

[0055] Take the CPV system as an example of a product attribute library. If the CPV system includes a first attribute word, then the attribute category corresponding to the first attribute word in the CPV correspondence that matches the first attribute word in the CPV system is determined as the selling point attribute category corresponding to the selling point keyword. The CPV correspondence that matches the first attribute word refers to the CPV correspondence that includes the first attribute word. For example, the first attribute word "imported" can be matched to "origin-imported" in the CPV system, and the first attribute word "rice cake" can be matched to "origin-imported" in the CPV system. Through this mapping process from attribute words to CPV, the essential attributes of the product selling point can be clarified, that is, it is clear what attributes of the product the selling point keyword is expressing.

[0056] In some cases, the CPV system doesn't include primary attribute terms, such as the subjective attribute term "aftertaste." These attributes don't have standardized attribute categories, so a large language model can be used to categorize them. A large language model can be any artificial intelligence model capable of text analysis and clustering.

[0057] By matching the CPV system, we can build a CPV constraint rule library that can intercept unreasonable associations. For example, for the selling point keyword "Korean fried chicken," its description must contain semantically similar keywords such as "Korea" or "Korean style" to successfully match the CPV system.

[0058] Step A3: clustering the selling point keywords according to the selling point attribute categories and the first positive words to obtain at least one selling point cluster; the selling point attribute categories corresponding to the selling point keywords in the same selling point cluster meet a preset similarity condition.

[0059] Optionally, step A3 may be performed as follows: If the first positive word describes the target product itself, clustering is performed directly according to the selling point keywords. For example, the selling point keyword "imported beef" is split into the first positive word "imported" and the first attribute word "beef". Since the first positive word "imported" directly refers to "beef" and describes the product itself, clustering is performed directly according to the selling point keywords. For example, all selling point keywords containing "imported + ingredients" are clustered into one cluster.

[0060] If the first positive word describes a sub-item within the target product, such as "imported rice cake" in the "fried chicken" category, clustering is performed based on the combination of selling point keywords. For example, if "imported rice cake" is a sub-item within a "fried chicken combo," clustering is performed based on the combination of selling point keywords. For example, clustering the combination of "fried chicken + imported rice cake" and "fried chicken + Korean rice cake" into one cluster would be performed based on the combination of selling point keywords.

[0061] The preset similarity condition may include: the semantic similarity between the selling point attribute categories corresponding to multiple selling point keywords is greater than or equal to a preset similarity threshold. When performing clustering, the semantic similarity between the selling point attribute categories corresponding to the selling point keywords or selling point keyword combinations can be calculated, and then the selling point keywords or selling point keyword combinations can be clustered based on the semantic similarity. This step, by clustering the selling point keywords, can group selling points with similar attributes into a single category, forming "clusters," thereby regularizing scattered selling points and reducing information clutter.

[0062] Step A4: Determine the selling point intention corresponding to each selling point cluster.

[0063] A selling point cluster may correspond to one or more selling point intentions. The selling point intention corresponding to a selling point cluster refers to the selling point intention corresponding to the selling point keyword or selling point keyword combination in the selling point cluster.

[0064] Optionally, during step A4, the selling point intent corresponding to each selling point cluster is determined based on expert experience. This involves labeling the clustered selling point clusters. Each selling point cluster corresponds to a product category. By labeling each category, the core intent of each selling point can be clearly defined. For example, in the "fried chicken" category, the selling point intent corresponding to selling point clusters such as "Imported Rice Cakes" and "Korean Sauces" could be determined to be "Korean Flavor," while the selling point intent corresponding to "Freshly Fried" and "Crispy on the Outside and Tender on the Inside" could be determined to be "Fresh Taste."

[0065] Alternatively, a white-box approach can be used to determine the selling point intent corresponding to each selling point cluster, that is, directly determining the corresponding selling point intent based on the selling point keyword or selling point keyword combination. For example, the selling point intent corresponding to the selling point keyword combination "imported + rice cake" is "Korean ingredients."

[0066] Optionally, an AI model can be used to predict the selling point intent corresponding to a selling point cluster. Specifically, the selling point keywords in the selling point cluster are input into the AI ​​model for intent prediction, resulting in the selling point intent corresponding to the cluster. For example, if the selling point keyword "Korean rice cake" is input into the AI ​​model, the model will automatically output the corresponding selling point intent "Korean flavor." The AI ​​model can be a pre-trained twin tower model, a text classification model, or the like.

[0067] Figure 2 An example diagram of the determination result of the selling point intention provided by the embodiment of the present application is shown, such as Figure 2 As shown, "Coffee," "Noodles," and "Soups" are clustered selling point clusters. Each selling point cluster can correspond to one or more selling point attribute categories, and each selling point attribute category can correspond to one or more selling point intents. For example, in the selling point cluster "Coffee," there are three selling point attribute categories: "Quality," "Specialty," and "Appealing to the Eyes." The selling point attribute category "Quality" corresponds to one selling point intent: "Quality Coffee." The selling point attribute category "Specialty" corresponds to two selling point intents: "Low Caffeine" and "Healthy and Light." The selling point attribute category "Appealing to the Eyes" corresponds to one selling point intent: "Refreshing and Cool." Figure 2 The meanings of the selling point clusters "noodles" and "soups" shown in the figure are similar to the selling point cluster "coffee", and will not be explained one by one.

[0068] In some embodiments, after determining the selling point intent of the product selling point according to the above steps A1 to A4, an intent system for the target product is generated based on the selling point keywords and the selling point intent. The generation process of the intent system can be performed as follows: Steps A5 and A6.

[0069] Step A5: for any selling point cluster, semantic expansion is performed on the selling point keywords in the selling point cluster to obtain selling point rewriting words of the selling point keywords; the selling point rewriting words include synonyms and / or hyponyms.

[0070] The attribute dictionary can be used to mine synonyms for selling point keywords, and word inclusion relationships can be used to mine hyponyms and hyponyms of selling point keywords. For example, the attribute dictionary can be used to mine synonyms for the selling point keyword "Korea," such as "Korean style" and "Korean flavor." Hyponyms of the keyword "Korea" can be mined through word inclusion relationships, such as "Korean style" and "Korean flavor."

[0071] After mining the selling point rewrite words of the selling point keywords, the hierarchical relationship of the selling point cluster can be constructed based on the selling point keywords and the selling point rewrite words. Figure 3 An example diagram showing the construction of a hierarchical relationship in the method for constructing a commodity selling point knowledge graph provided in an embodiment of the present application is shown. Figure 3 As shown, the "boneless chicken fillets, crispy and spicy", "golden crispy chicken fillets", "crispy cumin chicken fillets" and "spicy cumin chicken fillets, crispy and delicious" in the bottom row are selling point keywords, and the "crispy" in the middle row is the first positive word extracted from the selling point keywords (that is, the word that indicates the value and advantage of the product), and "chicken fillets" and "cumin" are the first attribute words extracted from the selling point keywords (that is, the word that indicates the attributes of the product). By semantically expanding the selling point keywords, we can obtain selling point rewritten words, such as "cumin powder", "cumin flavor", "chicken" and so on in the top row.

[0072] By building hierarchical relationships within selling point clusters, we can sort out the hierarchical relationships (e.g., peer or hierarchical relationships) between selling point keywords within similar selling points, thereby identifying the core selling points of an industry in an unsupervised manner. Similar selling points refer to product selling points corresponding to selling point keywords that are clustered into the same selling point cluster.

[0073] Step A6: Generate an intention system based on the selling point keywords in the selling point cluster, the selling point rewriting words corresponding to the selling point keywords, and the selling point intentions corresponding to the selling point cluster.

[0074] In this embodiment, by associating selling point keywords, selling point rephrased words and selling point intentions in the intent system, selling point keywords and their rephrased words with semantic relevance can be associated with selling point intentions, thereby converting the scattered selling points of the target product into analyzable, manageable and reusable structured information, making it easier for merchants or platforms to understand the selling point characteristics of their own products more clearly.

[0075] The following uses a specific example to explain in detail the process of building the association between selling point keywords and selling point intentions.

[0076] Take the example of "Madeleine Cake" in a bakery. Suppose the store has the following four products: Product 1: Golden Banana Madeleines. Key selling points include "burnt banana jam bursting with flavor, crispy on the outside and moist on the inside, with a golden, oozing center when you bite into it."

[0077] Product 2: Earl Grey Madeleine. Key selling points include "English black tea ground into powder, with a long-lasting tea aroma and the crispy texture of a Madeleine."

[0078] Product 3: Orange and Cumin Madeleines. Key selling points include "a novel combination of orange pulp and slightly spicy cumin, with a dense and fluffy Madeleine cake."

[0079] Product 4: Peach and Rose Madeleine. Key selling points include "peach pulp + candied rose, sweet and sour fragrance, soft and delicious cake."

[0080] Based on the above selling point keywords, we can break down the following attribute words and positive words. Positive words: bursting, runny, long, novel, slightly spicy, sweet and sour, fragrant, etc. Attribute words: banana, madeleine, black tea, orange, cumin, peach, rose, crispy, dense, etc.

[0081] Next, by matching attribute words with the CPV system, the selling point attribute category corresponding to the selling point keyword is determined. For example, matching the attribute word "Madeleine" with the CPV system yields the following relationship: [Category: Pastry - Madeleine], indicating that the selling point attribute category is "Pastry." Matching the attribute words "Crispy, Fluffy" with the CPV system yields the following relationship: [Taste: Crispy / Flushy], indicating that the selling point attribute category is "Crispy / Flushy." For attribute words that cannot be directly matched with the CPV system, an artificial intelligence model can be used to classify the attribute words to determine the corresponding selling point attribute category.

[0082] Afterwards, based on the aforementioned selling point attribute categories, the selling point keywords are clustered. The logic is to group similar selling points together based on "user perception similarity." Next, the hierarchical relationships of the selling point clusters are constructed to statistically identify the core selling points of the industry. Assume that the clustered selling point clusters include the following: Selling point cluster 1 (banana lava cluster) includes the following selling point keywords: banana, bursting, lava, crispy outside, moist inside. This cluster is associated with product 1.

[0083] Selling point cluster 2 (black tea long-lasting flavor cluster) includes the following selling point keywords: black tea, long-lasting flavor, tea fragrance, and crispy. This cluster is associated with product 2.

[0084] Selling point cluster 3 (orange and cumin cluster) includes the following selling point keywords: orange, cumin, novelty, and slightly spicy. This cluster is associated with product 3.

[0085] Selling point cluster 4 (Peach and Rose cluster) includes the following selling point keywords: peach, rose, sweet and sour, fresh fragrance, and soft fragrance. This cluster is associated with product 4.

[0086] Afterwards, based on expert experience, we defined selling point intent for each selling point cluster. For example, the selling point intent for selling point cluster 1 is "classic French innovation," the selling point intent for selling point cluster 2 is "suitable for English afternoon tea," the selling point intent for selling point cluster 3 is "trying new and novel flavors," and the selling point intent for selling point cluster 4 is "light and sweet girlish feeling."

[0087] Next, we determine the relationship between selling point keywords and selling point intent, linking these keywords, selling point intent, and products to create the intent system for Madeleine. For example, as long as the selling point keywords contain "banana, bursting, lava," they can be linked to the selling point intent corresponding to selling point cluster 1, "Classic French Innovation," and thus to product 1.

[0088] It can be seen that in this embodiment, by constructing a product intention system, the product selling points are transformed from scattered texts into a structured system that can be analyzed, associated, and managed, thereby ensuring the credibility of the product selling points.

[0089] In some embodiments, after generating the intent system of the target product based on the selling point keywords and selling point intentions, the specified association relationship in the intent system can be verified, and if the specified association relationship is verified to be an abnormal association relationship, the abnormal association relationship in the intent system can be corrected.

[0090] Optionally, the specified association relationship includes at least one of the following (a1) to (a4): (a1) The relationship between the selling point keyword and the product's industry. Specifically, the individual selling point keywords are verified to determine if they are appropriate for the product's industry. For example, if the target product is fried chicken and its industry is the fried chicken industry, and the corresponding selling point keywords include "Korean flavor," then the fried chicken industry's selling point can be verified to determine if it has "Korean flavor." If so, then the selling point keyword is determined to be appropriate for the product's industry, and the relationship between the selling point keyword and the product's industry is normal. Otherwise, if the selling point keyword is determined to be inappropriate for the product's industry, then the relationship between the selling point keyword and the product's industry is determined to be abnormal.

[0091] (a2) The hierarchical relationship between multiple selling point keywords. Specifically, verify whether the multiple selling point keywords within the same selling point cluster are accurately associated. For example, verify whether the association between "Korean flavor," "Korean-style fried chicken," and "delicious Korean-style fried chicken" is reasonable. Since "Korean flavor," "Korean-style fried chicken," and "delicious Korean-style fried chicken" are all synonyms for "Korean flavor," the association can be determined to be normal.

[0092] (a3) The association between the selling point keyword and the target product. This means verifying whether the association between the selling point keyword and the target product is accurate. For example, if the selling point keyword "Korean" is associated with Korean fried chicken, the association is correct; if the selling point keyword "Korean" is associated with American fried chicken, the association is abnormal.

[0093] (a4) The relationship between selling point keywords and selling point intentions. That is, verify whether the deduction from selling point keywords to selling point intentions is valid to ensure the correctness of the selling point intention association.

[0094] Among them, (a1), (a2), and (a3) ​​are segmented relationships within the intent system, and (a4) is the head-tail relationship within the intent system. By verifying both segmented and head-tail relationships within the intent system, the accuracy of the associations from product to selling point to intent is verified. This multi-layered verification mechanism ensures the reliability of the entire association system. Furthermore, the above embodiment transforms the relationship from "product-selling point" to "product-intent" and structures the product's selling points, resulting in a one-to-many relationship between products and intents.

[0095] In some embodiments, after clustering in step A3, multiple selling point clusters are obtained. Based on this, after generating the intent system for the target product, the intent priority of the selling point intent corresponding to each selling point cluster can be determined. Then, in the intent system, priority tags are added to the selling point intents, and the priority tags are used to indicate the intent priority of the selling point intent.

[0096] Optionally, the intent priorities of the selling point intentions corresponding to each selling point cluster are determined in the following manner: First, determine the metrics corresponding to the selling point intent. These metrics include at least one of the following: the supply of related products corresponding to the selling point intent, the visit-to-purchase rate of related products, the add-to-purchase rate of related products, and the accuracy of the association between the selling point intent and related products. Related products corresponding to the selling point intent are one or more products that have a correlation with the selling point intent.

[0097] Secondly, based on the indicator data corresponding to the selling point intention, determine the priority score corresponding to the selling point intention.

[0098] When determining the priority score, a weighted calculation can be performed on multiple indicator data according to the weight corresponding to each indicator data to obtain the priority score.

[0099] Again, the intention priority of the selling point intention is determined based on the priority score corresponding to the selling point intention.

[0100] In this embodiment, selling point intentions with higher intention priorities may be pushed or displayed first.

[0101] In some embodiments, when constructing a selling point knowledge graph based on the intent system (i.e., executing step S104), the following steps B1, B2, and B3 are included: Step B1: semantically align the first entity object in the intent system to obtain an aligned second entity object; the first entity object includes the target product, selling point keywords, and selling point intent.

[0102] The purpose of semantically aligning the first entity object is to ensure that the entity objects in the selling point knowledge graph are semantically consistent, avoiding confusion caused by different statements but the same meaning.

[0103] For example, merchants and users may express the same concept differently. For example, a merchant might say "freshly handmade," while a user might say "made on site." This discrepancy can cause confusion in subsequent analysis. Since both are essentially "made and sold immediately," aligning these two entity objects to "made and sold immediately" eliminates this confusion, provides an accurate linguistic foundation for subsequent relationship building, and ensures that information from different sources can be correctly linked.

[0104] Optionally, semantic alignment of the first entity objects is performed using a large language model (such as an M6 rewriting model). The M6 ​​rewriting model can unify entity objects with the same semantics into the same basic unit as a graph node of the selling point knowledge graph.

[0105] Step B2: Determine the semantic relationship between the second entity objects; the semantic relationship includes at least one of the following: inclusion relationship, derivation relationship, mutual exclusion relationship, and similarity relationship.

[0106] For example, Kung Pao Chicken contains "peanuts," which is an inclusion relationship. "Freshly fried" can be inferred from "crispy," which is a deduction relationship. "Sugar-free" and "high in sugar" are mutually exclusive, which is a mutual exclusion relationship. "Freshly baked" and "fresh" are similarities.

[0107] Step B3: convert the second entity object into a graph node of the selling point knowledge graph, and establish directed edges between the graph nodes based on semantic relationships to generate a selling point knowledge graph.

[0108] Figure 4 The interface diagram of the selling point knowledge graph provided by the embodiment of the present application is shown as follows: Figure 4 As shown, the selling point knowledge graph includes entities, attributes and relationships. Entities are graph nodes, such as products, merchants, categories (i.e. product categories), selling points, keywords (i.e. selling point keywords), basic attributes, category attributes (including general attributes and / or vertical attributes), selling point intentions, etc. Attributes are information that describes entity characteristics, such as the keyword type of selling point keywords, product description information, product pictures, whether it is a package, the selling point type of selling point, the source of selling points, the intention type of selling point intentions, etc. Relationships refer to the way entities are associated with each other. For example, "one to many" means that one entity can correspond to multiple associated entities, such as one product can correspond to multiple associated selling points. "One to many" means that one entity can correspond to one associated entity, such as one product can correspond to multiple associated merchants. It should be noted that, Figure 4 The selling point knowledge graph is merely a partial representation of the content. In actual applications, the selling point knowledge graph includes a large number of nodes and the relationships between them.

[0109] It can be seen that the selling point knowledge graph uses entities, attributes and relationships to clearly present the complete logic from product information to selling point mining and the association between selling points and intentions, solving the problems of selling point extraction, verification and association, and building a standardized knowledge network.

[0110] In this embodiment, the selling point knowledge graph includes an entity layer, a relationship layer, and an inference layer. The entity layer includes graph nodes, the relationship layer includes the associations between entity objects, and the inference layer is equipped with an algorithm for calculating the semantic relationships between entity objects. By constructing a knowledge graph that is dynamically updated and adaptable to application scenarios, the semantic relationships between target products, selling point keywords, and selling point intent are transformed into a logically clear network graph structure, facilitating automated management and updating of the graph, and better adapting to scenarios in dynamic market environments that require high timeliness of selling points.

[0111] In some embodiments, after constructing a selling point knowledge graph based on the intent system, the selling point knowledge graph is checked for abnormal association information, and the selling point knowledge graph is modified based on the abnormal association information. Abnormal association information includes at least one of the following: an abnormal association between a selling point keyword and a target product, semantic contradictions between multiple selling point keywords associated with a target product, and semantic contradictions between multiple selling point intents associated with a target product.

[0112] The Selling Point Knowledge Graph automatically identifies contradictory relationships between entities. For example, if a dish is labeled both "fried" (high in fat) and "light and nutritious" (low in fat), the Selling Point Knowledge Graph will detect this contradiction through the "mutually exclusive" relationship at the relationship level, triggering an alert. This can prompt merchants to revise the product's selling points or hide the contradictory label on the front end.

[0113] In some embodiments, after constructing a selling point knowledge graph based on the intent system, in response to extracting new selling point keywords for the target product, the confidence level of the new selling point keywords is determined. If the confidence level is greater than or equal to a preset confidence threshold, the selling point knowledge graph is updated based on the new selling point keywords.

[0114] The confidence of the selling point keyword can be determined based on the authenticity and universality of the selling point keyword. For example, if more than a certain number of merchants have used the selling point keyword, it means that the confidence of the selling point keyword is high.

[0115] In this embodiment, when a new selling point keyword appears and the confidence of the selling point keyword is greater than or equal to the preset confidence threshold, the selling point knowledge graph will be triggered to automatically update and expand. When updating the selling point knowledge graph based on the new selling point keyword, the selling point intention corresponding to the selling point keyword can be determined, and then the selling point keyword and / or selling point intention can be added to the selling point knowledge graph as a new entity object, and the association relationship between the new entity object and the existing entity object is calculated and constructed through the reasoning layer in the selling point knowledge graph. For example, the new selling point keyword "freshly milled rice" is added to the entity layer, and the association relationship between the entity object and the existing entity object is calculated through the reasoning layer, such as establishing a deductive relationship with "fresh" and "healthy", without the need for manual update.

[0116] In some embodiments, after constructing the selling point knowledge graph based on the intent system, the following steps are also included: receiving a display request for the intent system of the target product, the display request carries the product identification information of the target product; based on the display request, matching the product identification information with the selling point knowledge graph to obtain the intent system corresponding to the target product; and displaying the intent system corresponding to the target product.

[0117] Optionally, when displaying the intent system corresponding to the target product, at least one selling point intent whose target group index meets preset conditions is selected from the multiple selling point intents corresponding to the target product. The selected at least one selling point intent and selling point keywords associated with the at least one selling point intent are then displayed.

[0118] For example, by using the target group index to filter out the top five selling point intentions that users are most concerned about, such as "freshly made and sold", "no additives", "crispy on the outside and tender on the inside", etc., the labels of these selling point intentions will be displayed first on the product details page, allowing users to quickly learn about the product selling points.

[0119] Furthermore, the selling point knowledge graph supports visual reverse tracing. When a user clicks on a selling point intent on the selling point knowledge graph, the system will trigger the display of products associated with that selling point intent. For example, if a user clicks on the selling point intent "Fresh, fragrant, refreshing and spicy," only dishes associated with that selling point intent will be displayed.

[0120] In some embodiments, extracting selling point keywords for representing the selling points of a target product from multi-dimensional product information includes the following steps C1, C2, and C3: Step C1 extracts candidate text containing the target product's selling point information from the multi-dimensional product information. Multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, product review information, and SKU specification data. This information can reflect product characteristics from different perspectives. For example, a product image may indicate the freshness of ingredients, while the SKU specification may directly indicate "0 calorie sugar."

[0121] Furthermore, multi-dimensional product information often includes implicit features. For example, image recognition technology can identify information such as ingredients and freshness in product images, thereby converting these implicit features into explicit ones, helping users quickly understand the core characteristics of the product. SKU specification data often contains attributes that are strongly correlated with user decisions, such as "sweetness" and "hotness / coldness," which directly influence user purchasing decisions. By analyzing SKU specification data, high-quality attributes such as "0 calories of sugar" and "no additives" can be discovered and used as keywords for product selling points.

[0122] In step C2, the candidate text is input into a pre-trained artificial intelligence model for text analysis, so as to extract multiple selling point tags from the candidate text through the artificial intelligence model.

[0123] Optionally, the AI ​​model can be a RoBERTa-LSTM-CRF model. The AI ​​model includes a general attribute model and a vertical attribute model. The general attribute model can identify general attributes in candidate texts, and the vertical attribute model can identify vertical attributes in candidate texts. The general attributes and / or vertical attributes are then determined as selling point labels.

[0124] For example, inputting candidate text into the general attribute model can output corresponding general attributes: main ingredients, flavor, staple food, and auxiliary ingredients, etc. Inputting candidate text into the vertical attribute model can output corresponding vertical attributes: basic tea, special tea, milk base, basic ingredients, special ingredients, auxiliary ingredients, tea topping, flavor, processing and packaging, etc.

[0125] Step C3: Match the multiple selling point tags with the pre-configured selling point types, and determine the selling point tags that match the selling point types as selling point keywords based on the matching results.

[0126] The pre-configured selling point types are used to define the scope for subsequent extraction of selling point keywords to avoid extracting irrelevant information. For example, they may include ingredients, taste, craftsmanship, health, origin, quantity, etc.

[0127] Optionally, the selling point labels and pre-configured selling point types are input into a large language model, the selling point labels are matched with the pre-configured selling point types by the large language model, and end-to-end prediction is performed to output the extracted selling point keywords.

[0128] In this embodiment, by extracting the selling point keywords of the target product from multiple aspects, the key selling points that can influence user decisions are accurately mined from complex multi-dimensional product information, forming a complete extraction link from original information to selling point keywords, making the mined selling point keywords more accurate and in line with user needs, and at the same time able to dynamically adapt to market changes. For example, when health trends are updated, new selling point keywords can be quickly extracted from multi-dimensional product information.

[0129] Corresponding to the application scenario and method of the method provided in the embodiment of the present application, the embodiment of the present application also provides a device for constructing a product selling point knowledge graph.

[0130] Figure 5 A block diagram of a device for constructing a knowledge graph of product selling points provided in an embodiment of the present application is shown. Figure 5 As shown, the device for constructing a product selling point knowledge graph includes: An acquisition module 51 is configured to acquire multi-dimensional product information of a target product and extract selling point keywords representing the selling points of the target product from the multi-dimensional product information; the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information; A determination module 52 is configured to determine the selling point intention of the product selling point based on the selling point keywords; A generating module 53 is configured to generate an intention system for the target product based on the selling point keywords and the selling point intentions; the intention system includes associations between the selling point keywords and the selling point intentions; The construction module 54 is used to construct a selling point knowledge graph based on the intention system; the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intentions.

[0131] Optionally, when determining the selling point intention of the product selling point based on the selling point keyword, the determining module 52 performs the following steps: Parsing the selling point keywords to obtain a first attribute word and a first positive word of the product selling point; the first attribute word is used to represent the product attribute of the target product; the first positive word is used to represent the value of the target product; Determining, based on the first attribute word, a selling point attribute category corresponding to the selling point keyword; Clustering the selling point keywords according to the selling point attribute categories and the first positive words to obtain at least one selling point cluster; the selling point attribute categories corresponding to the selling point keywords in the same selling point cluster satisfy a preset similarity condition; Determine the selling point intention corresponding to each of the selling point clusters.

[0132] Optionally, when generating the intention system of the target product based on the selling point keywords and the selling point intentions, the generating module 53 performs the following steps: For any of the selling point clusters, semantic expansion is performed on the selling point keywords in the selling point cluster to obtain selling point rewriting words of the selling point keywords; the selling point rewriting words include synonyms and / or hyponyms; The intention system is generated according to the selling point keywords in the selling point cluster, the selling point rewriting words corresponding to the selling point keywords, and the selling point intentions corresponding to the selling point cluster.

[0133] Optionally, when determining the selling point attribute category of the product selling point based on the first attribute word, the determining module 52 performs the following steps: Matching the first attribute word with a pre-configured product attribute library; the product attribute library includes associations between attribute words and attribute categories; In a case where the first attribute word is included in the commodity attribute library, determining the attribute category associated with the first attribute word as the selling point attribute category; In the case that the first attribute word is not included in the commodity attribute library, the first attribute word is classified using a pre-trained large language model to obtain the selling point attribute category.

[0134] Optionally, the device further comprises: a verification module configured to verify, after generating an intent system for the target product based on the selling point keywords and the selling point intent, specified associations in the intent system; the specified associations comprising at least one of the following: an association between the selling point keywords and the industry to which the product belongs, a hierarchical relationship between a plurality of the selling point keywords, an association between the selling point keywords and the target product, and an association between the selling point keywords and the selling point intent; The first correction module is used to correct the abnormal association relationship in the intention system when verifying that the specified association relationship is an abnormal association relationship.

[0135] Optionally, the selling point cluster includes a plurality of selling point clusters; and the device further includes: a second determining module configured to determine the intent priorities of the selling point intents corresponding to the respective selling point clusters after generating the intent system of the target product based on the selling point keywords and the selling point intents; An adding module is used to add a priority tag to the selling point intention in the intention system; the priority tag is used to indicate the intention priority of the selling point intention.

[0136] Optionally, when determining the intention priorities of the selling point intentions corresponding to each of the selling point clusters, the second determining module performs the following steps: Determining indicator data corresponding to the selling point intention; the indicator data includes at least one of the following: the supply of related products corresponding to the selling point intention, the visit purchase rate of the related products, the add-on purchase rate of the related products, and the accuracy rate of the association between the selling point intention and the related products; Determining a priority score corresponding to the selling point intention based on the indicator data; Determine the intention priority of the selling point intention according to the priority score.

[0137] Optionally, when constructing the selling point knowledge graph according to the intention system, the construction module 54 performs the following steps: Semantically aligning a first entity object in the intent system to obtain an aligned second entity object; the first entity object includes the target product, the selling point keyword, and the selling point intent; Determining a semantic relationship between each of the second entity objects; the semantic relationship includes at least one of the following: an inclusion relationship, a derivation relationship, a mutual exclusion relationship, and a similarity relationship; The second entity object is converted into a graph node of the selling point knowledge graph, and directed edges are established between the graph nodes according to the semantic relationship to generate the selling point knowledge graph.

[0138] Optionally, the device further comprises: a detection module configured to detect whether there is abnormal association information in the selling point knowledge graph after the selling point knowledge graph is constructed according to the intent system; the abnormal association information includes at least one of the following: an abnormal association relationship between the selling point keyword and the target product, a semantic contradiction between multiple selling point keywords associated with the target product, and a semantic contradiction between multiple selling point intents associated with the target product; The second correction module is used to correct the selling point knowledge graph according to the abnormal association information.

[0139] Optionally, the device further comprises: A third determination module is configured to determine the confidence level of new selling point keywords extracted from the target product after the selling point knowledge graph is constructed according to the intention system; An updating module is configured to update the selling point knowledge graph based on the new selling point keyword when the confidence level is greater than or equal to a preset confidence threshold.

[0140] Optionally, the device further comprises: A receiving module, configured to receive a display request for the intent system of a target product after constructing the selling point knowledge graph according to the intent system; the display request carries product identification information of the target product; A matching module, configured to match the product identification information with the selling point knowledge graph based on the display request to obtain an intent system corresponding to the target product; The display module is used to display the intention system corresponding to the target product.

[0141] Optionally, when displaying the intent system corresponding to the target product, the display module performs the following steps: From the multiple selling point intentions corresponding to the target product, screening at least one selling point intention whose target group index meets a preset condition; The at least one selling point intention and selling point keywords associated with the at least one selling point intention are displayed.

[0142] Optionally, when extracting selling point keywords for representing the selling points of the target product from the multi-dimensional product information, the acquisition module 51 performs the following steps: Extracting candidate texts including selling point information of the target product from the multi-dimensional product information; Inputting the candidate text into a pre-trained artificial intelligence model for text analysis, so as to extract a plurality of selling point tags from the candidate text through the artificial intelligence model; The multiple selling point tags are matched with pre-configured selling point types, and based on the matching results, the selling point tags matching the selling point types are determined as the selling point keywords.

[0143] According to the device of the embodiment of the present application, by obtaining the multi-dimensional product information of the target product, the selling point keywords used to represent the product selling points of the target product are extracted from the multi-dimensional product information. The multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information. According to the selling point keywords, the selling point intention of the product selling point is determined, and according to the selling point keywords and the selling point intention, the intention system of the target product is generated. The intention system includes the association relationship between the selling point keywords and the selling point intention. Then, a selling point knowledge graph is constructed based on the intention system. The selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intention. It can be seen that the present application can not only capture the product selling points implied in the unstructured text by extracting the keywords related to the selling points from the multi-dimensional product information of the target product, and generating the intention system based on the extracted keywords as the basis for selling point mining, but also clarify the association path from the product selling points to the selling point intention, thereby realizing the transformation of the product selling points from scattered text into a structured system that can be analyzed, associated, and managed, thereby ensuring the credibility of the product selling points. In addition, by constructing a selling point knowledge graph based on the intent system, the semantic relationship between the target product, selling point keywords and selling point intentions is converted into a logically clear network graph structure, which facilitates automated management and updating of the graph and is more adaptable to scenarios with higher requirements for the timeliness of selling points in a dynamic market environment.

[0144] The functions of each module in each device in the embodiments of the present application can be found in the corresponding description in the above method, and have corresponding beneficial effects, which will not be repeated here.

[0145] Figure 6 A block diagram of an electronic device for implementing the embodiments of the present application. Figure 6 As shown, the electronic device includes a memory 601 and a processor 602. The memory 601 stores a computer program executable by the processor 602. When the processor 602 executes the computer program, the method described in the above embodiment is implemented. The number of the memory 601 and the processor 602 can be one or more. In a specific implementation, the electronic device may also include a communication interface 603 for communicating with external devices and exchanging data.

[0146] In a specific implementation, if the memory 601, processor 602, and communication interface 603 are implemented independently, the memory 601, processor 602, and communication interface 603 can be connected to each other via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0147] Optionally, in a specific implementation, if the memory 601 , the processor 602 , and the communication interface 603 are integrated on a chip, the memory 601 , the processor 602 , and the communication interface 603 may communicate with each other through an internal interface.

[0148] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present application when the program is executed by a processor.

[0149] An embodiment of the present application provides a computer program product, including a computer program, which implements the method provided in the embodiment of the present application when executed by a processor.

[0150] An embodiment of the present application also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided in the embodiment of the present application.

[0151] An embodiment of the present application also provides a chip, including: an input interface, an output interface, a processor and a memory. The input interface, the output interface, the processor and the memory are connected through an internal connection path. The processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.

[0152] It should be understood that the processor described above may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor may be a microprocessor or any conventional processor. It is worth noting that the processor may be a processor that supports the Advanced RISC Machine (ARM) architecture.

[0153] Furthermore, optionally, the aforementioned memory may include read-only memory and random access memory. The memory may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache memory. By way of example and not limitation, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM) and direct memory bus random access memory (DR RAM).

[0154] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0155] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0156] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0157] Any process or method described in the flowchart or otherwise described herein can be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process. The scope of the preferred embodiments of the present application includes other implementations in which the functions may be performed in a different order than shown or discussed, including performing the functions substantially simultaneously or in reverse order depending on the functions involved.

[0158] The logic and / or steps described in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor, or other system that can fetch instructions from and execute instructions on an instruction execution system, apparatus or device), or used in conjunction with such instruction execution systems, apparatuses or devices.

[0159] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0160] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the aforementioned integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a magnetic disk, or an optical disk, etc.

[0161] The above is merely an exemplary embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various modifications or substitutions within the technical scope described in this application, and such modifications or substitutions should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for constructing a knowledge graph of product selling points, characterized in that: include: Acquire multi-dimensional product information of a target product, and extract selling point keywords used to represent the selling points of the target product from the multi-dimensional product information; the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information; Determining the selling point intention of the product selling point based on the selling point keywords; generating an intention system for the target product based on the selling point keywords and the selling point intentions; the intention system including the association relationship between the selling point keywords and the selling point intentions; A selling point knowledge graph is constructed based on the intention system; the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intentions.

2. The method according to claim 1, characterized in that Determining the selling point intention of the product selling point based on the selling point keyword includes: Parsing the selling point keywords to obtain a first attribute word and a first positive word of the product selling point; the first attribute word is used to represent the product attribute of the target product; the first positive word is used to represent the value of the target product; Determining, based on the first attribute word, a selling point attribute category corresponding to the selling point keyword; Clustering the selling point keywords according to the selling point attribute categories and the first positive words to obtain at least one selling point cluster; the selling point attribute categories corresponding to the selling point keywords in the same selling point cluster satisfy a preset similarity condition; Determine the selling point intention corresponding to each of the selling point clusters.

3. The method according to claim 2, characterized in that Generating the target product intention system based on the selling point keywords and the selling point intentions includes: For any of the selling point clusters, semantic expansion is performed on the selling point keywords in the selling point cluster to obtain selling point rewriting words of the selling point keywords; the selling point rewriting words include synonyms and / or hyponyms; The intention system is generated according to the selling point keywords in the selling point cluster, the selling point rewriting words corresponding to the selling point keywords, and the selling point intentions corresponding to the selling point cluster.

4. The method according to claim 2, characterized in that The step of determining the selling point attribute category of the product selling point based on the first attribute word includes: Matching the first attribute word with a pre-configured product attribute library; the product attribute library includes associations between attribute words and attribute categories; In a case where the first attribute word is included in the commodity attribute library, determining the attribute category associated with the first attribute word as the selling point attribute category; In the case that the first attribute word is not included in the commodity attribute library, the first attribute word is classified using a pre-trained large language model to obtain the selling point attribute category.

5. The method according to claim 1, wherein After generating the target product intention system based on the selling point keywords and the selling point intention, the method further includes: Verify the specified association relationship in the intent system; the specified association relationship includes at least one of the following: the association relationship between the selling point keyword and the industry to which the product belongs, the hierarchical relationship between multiple selling point keywords, the association relationship between the selling point keyword and the target product, and the association relationship between the selling point keyword and the selling point intent; When it is verified that the designated association relationship is an abnormal association relationship, the abnormal association relationship in the intention system is corrected.

6. The method according to claim 2, characterized in that The selling point cluster includes multiple; After generating the target product intention system based on the selling point keywords and the selling point intention, the method further includes: Determining the intention priority of the selling point intentions corresponding to each of the selling point clusters; In the intent system, a priority tag is added to the selling point intent; the priority tag is used to indicate the intent priority of the selling point intent.

7. The method according to claim 6, characterized in that The determining the intention priority of the selling point intentions corresponding to each of the selling point clusters includes: Determining indicator data corresponding to the selling point intention; the indicator data includes at least one of the following: the supply of related products corresponding to the selling point intention, the visit purchase rate of the related products, the add-on purchase rate of the related products, and the accuracy rate of the association between the selling point intention and the related products; Determining a priority score corresponding to the selling point intention based on the indicator data; Determine the intention priority of the selling point intention according to the priority score.

8. The method according to claim 1, characterized in that The step of constructing a selling point knowledge graph based on the intent system includes: Semantically aligning a first entity object in the intent system to obtain an aligned second entity object; the first entity object includes the target product, the selling point keyword, and the selling point intent; Determining a semantic relationship between each of the second entity objects; the semantic relationship includes at least one of the following: an inclusion relationship, a derivation relationship, a mutual exclusion relationship, and a similarity relationship; The second entity object is converted into a graph node of the selling point knowledge graph, and directed edges are established between the graph nodes according to the semantic relationship to generate the selling point knowledge graph.

9. The method according to claim 8, characterized in that After constructing the selling point knowledge graph according to the intention system, the method further includes: Detecting whether there is abnormal association information in the selling point knowledge graph; the abnormal association information includes at least one of the following: an abnormal association relationship between the selling point keyword and the target product, a semantic contradiction between multiple selling point keywords associated with the target product, and a semantic contradiction between multiple selling point intentions associated with the target product; The selling point knowledge graph is modified according to the abnormal association information.

10. The method according to claim 8, characterized in that After constructing the selling point knowledge graph according to the intention system, the method further includes: In response to extracting a new selling point keyword of the target product, determining a confidence level of the new selling point keyword; When the confidence level is greater than or equal to a preset confidence threshold, the selling point knowledge graph is updated based on the new selling point keyword.

11. The method according to claim 1, wherein After constructing the selling point knowledge graph according to the intention system, the method further includes: receiving a display request for the intent system of the target product; the display request carries product identification information of the target product; Based on the display request, the product identification information is matched with the selling point knowledge graph to obtain the intent system corresponding to the target product; Display the intent system corresponding to the target product.

12. The method according to claim 11, characterized in that The intention system corresponding to displaying the target product includes: From the multiple selling point intentions corresponding to the target product, screening at least one selling point intention whose target group index meets a preset condition; The at least one selling point intention and selling point keywords associated with the at least one selling point intention are displayed.

13. The method according to claim 1, wherein The step of extracting selling point keywords for representing the selling points of the target product from the multi-dimensional product information includes: Extracting candidate texts including selling point information of the target product from the multi-dimensional product information; Inputting the candidate text into a pre-trained artificial intelligence model for text analysis, so as to extract a plurality of selling point tags from the candidate text through the artificial intelligence model; The multiple selling point tags are matched with pre-configured selling point types, and based on the matching results, the selling point tags matching the selling point types are determined as the selling point keywords.

14. A device for constructing a knowledge graph of commodity selling points, characterized in that: include: an acquisition module, configured to acquire multi-dimensional product information of a target product and extract selling point keywords representing the selling points of the target product from the multi-dimensional product information; the multi-dimensional product information includes at least one of the following: product attribute information, product image information, product description information, and product evaluation information; a determination module, configured to determine the selling point intention of the product selling point based on the selling point keywords; A generating module, configured to generate an intention system for the target product based on the selling point keywords and the selling point intentions; the intention system includes an association relationship between the selling point keywords and the selling point intentions; A construction module is used to construct a selling point knowledge graph based on the intention system; the selling point knowledge graph is used to represent the semantic relationship between the target product, the selling point keywords and the selling point intentions.

15. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory, wherein the processor implements the method according to any one of claims 1 to 13 when executing the computer program.

16. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 13 is implemented.

17. A computer program product, characterized in that A computer program is included which, when executed by a processor, implements the method according to any one of claims 1 to 13.

Images