Method and device for acquiring commodity attribute information, electronic equipment and storage medium

Abstract

The present disclosure discloses a commodity attribute information acquisition method and device, electronic equipment and storage medium, relates to the field of computers, and particularly to the fields of artificial intelligence technologies such as large models, deep learning and intelligent recommendation. The method comprises: mining historical session data of a commodity and a commodity detail page to determine first attribute information of the commodity to obtain a commodity attribute library; collecting application session data related to the commodity in an intelligent service application process; supplementally mining the application session data to obtain second attribute information of the commodity; and according to the second attribute information, complementing attribute information of the commodity in the commodity attribute library.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to the fields of artificial intelligence technology such as large models, deep learning, intelligent recommendation, and intelligent customer service, specifically to methods, devices, electronic devices, and storage media for obtaining product attribute information. Background Technology

[0002] With the continuous expansion of the e-commerce industry and the increasing variety of products, user inquiries have exploded. Traditional human customer service models, limited by labor costs and response efficiency, can no longer meet users' core needs for "efficient, immediate, and accurate" inquiries. Against this backdrop, intelligent customer service systems based on large-scale models have emerged to improve customer service response efficiency and reduce manual operating costs. However, the quality and accuracy of intelligent customer service responses depend on product attribute information; therefore, accurate acquisition of product attribute information is essential. Summary of the Invention

[0003] This disclosure provides a method, apparatus, electronic device, and storage medium for obtaining product attribute information. The specific solution is as follows:

[0004] According to one aspect of this disclosure, a method for obtaining product attribute information is provided, comprising: By mining historical session data and product detail pages, the first attribute information of the product is determined to obtain a product attribute library; Collect application session data related to the product during the application of intelligent services; Further mining of the application session data yields the second attribute information of the product; Based on the second attribute information, the attribute information of the product in the product attribute library is completed.

[0005] According to another aspect of this disclosure, a device for obtaining product attribute information is provided, comprising: The mining module is used to mine historical session data and product detail pages of products to determine the first attribute information of the products in order to obtain a product attribute library; The data acquisition module is used to collect application session data related to the product during the application of intelligent services. The mining module is also used to perform supplementary mining on the application session data to obtain the second attribute information of the product, and to complete the attribute information of the product in the product attribute library based on the second attribute information.

[0006] According to another aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described in the above embodiments.

[0007] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described in the above embodiments.

[0008] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described in the above embodiments.

[0009] The method, apparatus, electronic device, and storage medium for obtaining product attribute information disclosed herein have the following beneficial effects: In this embodiment, the construction of a product attribute library by combining historical session data and product detail pages enriches product attributes from both user and merchant perspectives. This results in a product attribute library that is closer to real-world application scenarios, more comprehensive, and more timely, effectively improving the accuracy of matching user needs with products in scenarios such as product search, product recommendation, and intelligent customer service. Furthermore, after constructing the product attribute library, continuous collection of application session data and mining based on this data enables closed-loop optimization of the product attribute library, enhancing it and generating a significant flywheel effect. Moreover, mining based on application session data can continuously supplement the correlation between product attributes and user expressions, improving the applicability to different application scenarios. It can also capture trending keywords and changing demands, ensuring the timeliness of the product attribute library and continuously improving the user experience.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0011] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A flowchart illustrating a method for obtaining product attribute information according to an embodiment of this disclosure; Figure 2 A flowchart illustrating a method for obtaining product attribute information according to another embodiment of this disclosure; Figure 3 A flowchart illustrating a method for obtaining product attribute information according to another embodiment of this disclosure; Figure 4 A schematic diagram of the framework for a method for obtaining product attribute information provided in another embodiment of this disclosure; Figure 5 A schematic diagram of the structure of a device for obtaining product attribute information provided in an embodiment of this disclosure; Figure 6 This is a block diagram of an electronic device used to implement the method for obtaining product attribute information according to embodiments of the present disclosure. Detailed Implementation

[0012] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0013] The embodiments disclosed herein relate to the fields of artificial intelligence technology, such as large models, deep learning, intelligent recommendation, and intelligent customer service.

[0014] Artificial Intelligence (AI) is a new technological science that studies, develops, and applies theories, methods, technologies, and application systems to simulate, extend, and expand human intelligence.

[0015] Large models, also known as Foundation Models, are models that extract knowledge from hundreds of millions of corpora or images, learn, and then produce large models with hundreds of millions of parameters.

[0016] Deep learning (DL) learns the inherent patterns and hierarchical representations of sample data. The information gained during this learning process greatly aids in interpreting data such as text, images, and sound. The ultimate goal of deep learning is to enable machines to possess analytical and learning capabilities similar to humans, allowing them to recognize data such as text, images, and sound.

[0017] Intelligent recommendation, as an important means of information filtering and personalized services, is based on integrating multi-dimensional information such as user behavior data, interests and preferences, and contextual environment, and using algorithmic models to automatically analyze and predict users' potential needs or interests, and then recommend content, products, services or social relationships that users may be interested in.

[0018] Intelligent customer service refers to a system that uses artificial intelligence (AI) technology, especially natural language processing (NLP), machine learning (ML) and big data technology, to replace or assist traditional human customer service, so as to automatically and intelligently answer user inquiries, handle business needs and provide personalized services.

[0019] It should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of any type of information, such as user personal information, involved in this technical solution comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0020] The following description, with reference to the accompanying drawings, outlines a method, apparatus, electronic device, and storage medium for obtaining product attribute information according to embodiments of this disclosure.

[0021] Figure 1 This is a flowchart illustrating a method for obtaining product attribute information according to an embodiment of this disclosure.

[0022] like Figure 1 As shown, the methods for obtaining the product attribute information include: It should be noted that the application scenarios of the product attribute information acquisition method proposed in this disclosure can be determined according to actual needs. For example, it can be applied to application scenarios that include multiple heterogeneous but related scenarios, such as e-commerce, content recommendation, advertising systems, and intelligent customer service, and this disclosure does not limit it in this regard.

[0023] S101, Mining the historical session data and product details page of the product to determine the first attribute information of the product, so as to obtain the product attribute library.

[0024] In some embodiments, the dialogue process between the user and the intelligent service can be collected to obtain historical conversation data of the product. Intelligent services may include, but are not limited to, intelligent customer service, shopping guide robots, and customer service mini-programs.

[0025] In some embodiments, the product’s historical sessions may include session data for different users from the start to the end of the session at a historical moment, that is, including the user’s session context during the session.

[0026] In some embodiments, the historical session data of a product may include, but is not limited to, conversation texts between the user and the intelligent service at different service stages. For example, in an intelligent customer service scenario, historical session data may include conversation texts from all scenarios, such as pre-sales consultations, after-sales feedback, and usage questions.

[0027] In some embodiments, the specific implementation method for collecting historical session data can be set according to actual needs.

[0028] In some embodiments, user conversations on the website or application corresponding to the intelligent service can be collected using event tracking techniques. For example, a software development kit (Salesforce Interactions SDK) can be embedded in the page code of the website or application corresponding to the intelligent service to record the user's input information in the conversation interface.

[0029] In some embodiments, historical session data of a user during a session can be collected through a built-in behavior collection component.

[0030] In some embodiments, historical session data may include user-input query text, speech-to-text, uploaded text and / or images, etc.

[0031] In some embodiments, historical conversation data may also include response text from large language models, multimodal models, or intelligent customer service outputs, product descriptions, or product information.

[0032] In some embodiments, the collected historical conversation text can be processed to obtain a complete, ordered, and logical conversation data, which can then be associated with the product's identification information.

[0033] In some embodiments, the historical session data of a product can be formatted to obtain structured data, which facilitates the processing of subsequent large models.

[0034] In some embodiments, the product details page may carry multimodal information about the product, which may include, but is not limited to: text descriptions of the product, real-life photos of the product, and video clips demonstrating its features.

[0035] In some embodiments, the multimodal information on the product details page can be used to extract visual attribute information such as the product's appearance, material, and specifications.

[0036] In some embodiments, the product details page can be retrieved from a merchant database based on the product's identification information.

[0037] In some embodiments, historical session data can be understood and analyzed to obtain identification information of products that the user is interested in, and then the product details page can be obtained based on the product identification information.

[0038] In some embodiments, the product details page can be pre-generated based on the product information submitted by the product provider.

[0039] In some embodiments, after obtaining the historical session data and product details page of a product, data mining can be performed on the historical session data and product details page to extract the customer attributes of the product and obtain the first attribute information of the product. Furthermore, a product attribute library can be constructed based on the first attribute information of the product.

[0040] In some embodiments, multimodal offline mining can be performed on the historical session data and product detail pages of the product based on a large model to extract the customer attributes of the product and obtain the first attribute information of the product.

[0041] In some embodiments, the first attribute information includes multiple attribute identifiers (the attribute identifiers may include, but are not limited to, attribute indexes, attribute keys) and attribute values ​​for each attribute identifier. Optionally, the first attribute information includes multiple key-value pairs, where the key represents the attribute identifier of a certain attribute and the value is the value of a certain attribute.

[0042] S102, collect application session data related to products during the application of intelligent services.

[0043] In some embodiments, after constructing the product attribute database, services can be provided to users based on the database. Users can submit questions to the intelligent service through a session page, and further, obtain response information matching the question information based on the product attribute database. In this embodiment, application session data related to products can be collected during the application of the intelligent service. It is understood that the method for collecting application session data can refer to the method for collecting historical session data, and will not be repeated here.

[0044] In some embodiments, a user's interaction with the intelligent service may include multiple sessions, and information on each question and its corresponding response during the interaction may be collected as application session data for the product.

[0045] S103, further mine the application session data to obtain the second attribute information of the product.

[0046] In some embodiments, after obtaining application session data, the application session data can be used as a new data source for mining. By supplementing the application session data with mining, the second attribute information of the product can be extracted. In this embodiment of the disclosure, the continuous enrichment and improvement of product attribute information can be achieved through a mining-application-supplementation-re-mining pattern.

[0047] In some embodiments, natural language processing can be used to extract product-related entities from application session data. These entities include product name, brand, model, and various potential attributes (such as color, size, material) and corresponding attribute values, to obtain the product's second attribute information.

[0048] In some embodiments, the first attribute information includes multiple attribute identifiers (the attribute identifiers may include, but are not limited to, attribute indexes, attribute keys) and attribute values ​​for each attribute identifier. Optionally, the first attribute information includes multiple key-value pairs, where the key represents the attribute identifier of a certain attribute and the value is the value of a certain attribute.

[0049] S104. Based on the second attribute information, complete the attribute information of the products in the product attribute library.

[0050] In some embodiments, after obtaining the second attribute information, the attribute information that needs to be supplemented for the product can be determined based on the first and second attribute information of the product in the product attribute library, and the attribute information of the product in the product attribute library can be supplemented as needed.

[0051] For example, in an intelligent recommendation scenario, historical session data can be obtained from users' product searches using a large model. Based on this historical session data, first-attribute information is mined to obtain an initial product attribute library. After the product attribute library is built, it can be applied upstream. Users can query the constructed product attribute library based on the large model to search for products, thereby obtaining application session data. This application session data is used as a supplementary data source to mine second-attribute information, completing the product attribute library and resulting in a complete product attribute library. Furthermore, in an intelligent recommendation scenario, based on the continuously iterated and complete product attribute library, accurate matching of user needs with products can be achieved, improving the user experience on the platform.

[0052] For example, in an intelligent customer service scenario, historical user conversations on the customer service chat page can be collected to obtain historical conversation data. This historical conversation data can include conversation data from different stages of the customer's interaction, encompassing, but not limited to, pre-sales consultations, after-sales feedback, and usage questions—interactive text data across all scenarios. Based on this historical conversation data, first attribute information is mined to obtain an initial product attribute library. After the product attribute library is built, it can be applied upstream. Users can query the constructed product attribute library based on a large model to search for products, thereby obtaining application conversation data. This application conversation data is used as a supplementary data source to mine second attribute information, thus completing the product attribute information in the product attribute library. In an intelligent customer service scenario, based on a continuously iterated and complete product attribute library, the goal of providing users with detailed product information can be achieved. This allows for a comprehensive and clear display of various product attribute information, helping users quickly understand the core features of the product and improving output accuracy.

[0053] In this embodiment of the disclosure, the construction of a product attribute library by combining historical session data and product details page can enrich product attributes from both the user's and the merchant's perspectives. This results in a product attribute library that is closer to real application scenarios, more comprehensive, and more timely, which can effectively improve the accuracy of matching user needs with products in scenarios such as product search, product recommendation, and intelligent customer service.

[0054] Furthermore, after building the product attribute library, continuous collection of application session data and mining based on this data enable closed-loop optimization of the product attribute library, enhancing it and generating a significant flywheel effect. Moreover, mining based on application session data can continuously supplement the relationship between product attributes and user expressions, improving the applicability to different application scenarios. It can also capture trending keywords and changing demands, making the product attribute library timely and thus continuously improving the user experience.

[0055] Figure 2 This is a flowchart illustrating another method for obtaining product attribute information provided in an embodiment of this application. Figure 2 As shown, the methods for obtaining this product attribute information may include, but are not limited to, the following steps: S201, collect historical session data and product detail page data for products.

[0056] The specific implementation of step S201 can be adopted in any embodiment of this application, and the steps are described in detail here.

[0057] S202, invoke the multimodal large model to perform multimodal parsing and feature extraction on the product details page to obtain the first local attribute information of the product.

[0058] In some embodiments, after collecting the product details page, since the product details page can carry multimodal information about the product, in order to provide accurate product attribute information, a multimodal large model can be invoked. Through the multimodal large model, multimodal parsing and feature extraction are performed on the product details page to obtain the first local attribute information of the product.

[0059] In some embodiments, the multimodal large model can be an artificial intelligence model that integrates visual understanding and text analysis. The product details page can be input into the multimodal large model, and the unstructured product details page can be transformed into structured product attribute information through the multimodal large model.

[0060] In some embodiments, the product details page is preprocessed using a multimodal large model to extract visual modal information such as the main product image and detail images, as well as textual modal information such as the title, description, and parameter table. The visual and textual modal information is then parsed and aligned. Furthermore, the multimodal large model utilizes a unified visual encoder to map images and text to the same semantic space and understands the relationships between images and text through cross-modal interaction mechanisms, such as mapping colors in an image to color codes in the text. Further, after understanding the multimodal information, the multimodal large model can extract specific product attribute information, namely the aforementioned first local attribute information.

[0061] For example, the details page of an electric kettle can be input into a multimodal large model. The multimodal large model can perform multimodal mining on the details page to obtain information such as the material being stainless steel, the maximum working power being 1200W, and the maximum capacity being 1.7L.

[0062] S203, invoke the large text model to perform semantic parsing and entity extraction on historical session data to obtain the second local attribute information of the product.

[0063] In some embodiments, the first attribute information mined from the product's historical session data and product details page includes at least first local attribute information and second local attribute information.

[0064] In some embodiments, after obtaining historical session data, the historical session data can be input into a text big data model, which extracts product attribute information from the historical session data. In this embodiment, the text big data model accurately captures the user's focus on attributes by analyzing vague descriptions, colloquial expressions, and implicit needs in the historical session data, thereby enriching the attribute dimensions of the product.

[0065] In some embodiments, the text big model performs context-aware semantic understanding and entity extraction on historical session data under the guidance of pre-configured prompts. Furthermore, the text big model can encode the context information and entity information extracted from the historical session data, and then extract the second local attribute information of the product based on the encoded information.

[0066] In some embodiments, there may be duplicate attribute information in the extracted first local attribute information and second local attribute information, and the duplicate attribute information can be deduplicated.

[0067] In some embodiments, invalid and / or redundant attribute information may also exist in the extracted first and second local attribute information, and invalid and / or redundant attribute information can be automatically removed.

[0068] In this embodiment of the disclosure, the accuracy of the mined attribute information can be ensured by automatically removing invalid, redundant, and duplicate information.

[0069] In some embodiments, the first local attribute information and the second local attribute information can be stored as key-value pairs of attribute names and attribute values. That is, after determining the first local attribute information and the second local attribute information, they can be structured to obtain the structured attribute information of the product. It can be understood that the product attribute library stores the structured attribute information of the product. In the implementation of this disclosure, the standardization and semi-structured processing of attribute information to unify the data format lays the foundation for the subsequent reuse of attribute information.

[0070] In some embodiments, in order to improve the accuracy of extracting product attribute information, at least one of the first local attribute information and the second local attribute information can be verified. By using verification prompts, a pre-trained evaluation model can evaluate at least one of the first local information and the second local attribute information and output the confidence scores of the first local attribute information and the second local attribute information.

[0071] In some embodiments, first and second partial attribute information can be sent to a manual review device for manual review to correct the first and second partial attribute information, thereby making the attribute information of products in the product attribute database more accurate. In this disclosure, the confidence level and manual review methods ensure that the structured product attributes finally entered into the database have high credibility, thus laying the foundation for subsequent product analysis and personalized recommendations.

[0072] In this embodiment, a product attribute mining mechanism combining a text-based large-scale model and a multimodal large-scale model enables deep semantic alignment and information complementarity between user demand and product supply. The text-based large-scale model accurately captures user-focused attributes by analyzing vague descriptions, colloquial expressions, and implicit needs in user conversations, while the multimodal large-scale model structurally extracts an objective and comprehensive attribute system from images, titles, and detailed descriptions on the product details page. This joint mining of the two types of large-scale models allows for the construction of a more realistic, comprehensive, and timely product attribute library. This mechanism effectively improves the matching degree between user needs and products in scenarios such as search recommendations and intelligent customer service.

[0073] S204, Collect application session data related to goods during the application of intelligent services.

[0074] In some embodiments, the product attribute database is searched based on at least one round of question information input by the user to obtain search results. The search results include at least the first attribute information and basic information of the product. Further, the response information to the question information is output based on the first attribute information and basic information, and the at least one round of question information input by the user and the corresponding response information are determined as application session data.

[0075] In some embodiments, product retrieval information can be determined based on the question information, and the product attribute database can be searched based on the product retrieval information to obtain retrieval results. The product retrieval information may include product identification information, product index information, etc.

[0076] In some embodiments, a target question-answer pair matching the question information is determined, and the response information to the question information is output based on the target question-answer pair, the first attribute information, and the basic information. Further, at least one round of question information input by the user and the corresponding response information are determined as application session data.

[0077] In this embodiment of the disclosure, by obtaining target question-answer pairs that match the question information, it is possible to output response information that matches the user's needs, improve the accuracy of the conversation response, enhance the user experience, and facilitate the accurate push of products.

[0078] In some embodiments, candidate question-answer pairs can be pre-stored in a product attribute database. Similarity calculations are performed between the input question information and candidate questions in the candidate question-answer pairs to identify the target question-answer pair that matches the question information. Optionally, the vector representation of the question information and the vector representation of the candidate questions in the candidate question-answer pairs are obtained. Based on these vector representations, the similarity between the input question information and the candidate questions is calculated. The target question-answer pair is then determined from the candidate question-answer pairs according to the similarity score. For example, the candidate question-answer pair containing the candidate question with the highest similarity score is selected as the target question-answer pair. In other words, the target question-answer pair can be determined from the candidate question-answer pairs based on vector similarity and the input question information.

[0079] In this embodiment of the disclosure, by collecting application session data, emerging user needs, long-tail expressions, and implicit feedback can be extracted, which can provide data support for the continuous optimization of the completeness and accuracy of the attribute information of products in the product attribute library.

[0080] S205, perform supplementary mining on the application session data to obtain the second attribute information of the product, and complete the attribute information of the product in the product attribute library based on the second attribute information.

[0081] In some embodiments, after obtaining application session data, the application session data can be used as a new data source for mining. By supplementing the mining of the application session data, the second attribute information of the product can be extracted.

[0082] In some embodiments, after obtaining application session data, the application session data can be input into a large text model, and the second attribute information of the product can be extracted from the application session data through the large text model.

[0083] In some embodiments, the extracted second attribute information may contain duplicate attribute information, and the duplicate attribute information can be deduplicated.

[0084] In some embodiments, the extracted second attribute information may also contain invalid and / or redundant attribute information, which can be automatically removed. By automatically removing invalid, redundant, and duplicate information, the accuracy of the mined attribute information is ensured.

[0085] In some embodiments, after the second attribute information is determined, it can be stored as key-value pairs of attribute names and attribute values. That is, the second attribute information can be structured to obtain the structured attribute information of the product.

[0086] In some embodiments, in order to improve the accuracy of extracting product attribute information, the second attribute information can be verified. By using verification prompts, a pre-trained evaluation model can evaluate the second attribute information and output the confidence level of the second attribute information.

[0087] In some embodiments, second attribute information can be sent to a manual review device for manual review to correct the second attribute information, thereby making the attribute information of the products in the product attribute library more accurate.

[0088] In this public implementation, confidence levels and manual review are used to ensure that the structured product attributes that are finally entered into the warehouse have a high degree of credibility, thereby laying the foundation for subsequent product analysis and personalized recommendations.

[0089] In some embodiments, after obtaining the second attribute information, the attribute information that needs to be supplemented for the product can be determined based on the first and second attribute information of the product in the product attribute library, and the attribute information of the product in the product attribute library can be supplemented as needed.

[0090] For example, the first attribute information includes attribute information A, attribute information B, and attribute information C, while the second attribute information includes attribute information A, attribute information D, and attribute information E. It can be determined that attribute information D and attribute E are the attribute information that needs to be supplemented for the product, and the supplemented attribute information D and attribute E are added to the attribute information of the product in the product attribute library.

[0091] In some embodiments, a union operation is performed on the first attribute information and the second attribute information to obtain an attribute information set, and a deduplication operation is performed on the attribute information set to obtain the final attribute information of the product.

[0092] In some embodiments, duplicate attribute information in the attribute information set is identified, and the duplicate attribute information is deduplicated to obtain the final attribute information of the product.

[0093] For example, the first attribute information includes attribute information A, attribute information B, and attribute information C, while the second attribute information includes attribute information A, attribute information D, and attribute information E. By performing a union operation, it can be determined that the attribute information set includes {attribute information A, attribute information B, attribute information C, attribute information A, attribute information D, and attribute information E}. It can be determined that attribute information A is a duplicate attribute information. By removing one of the attribute information A, the final attribute information of the product can be obtained.

[0094] In this embodiment of the disclosure, the attribute information of the product is finally obtained through the first attribute information and the second attribute information. By completing the information, the completeness and timeliness of the product attribute information in the merchant attribute library are achieved, which is conducive to better product search, product recommendation and intelligent customer service.

[0095] In some embodiments, the first attribute information includes multiple attribute identifiers and a first attribute value for each attribute identifier, and the second attribute information includes multiple attribute identifiers and a second attribute value for each attribute identifier. It is understood that the attribute values ​​corresponding to the same attribute identifier may be the same or different.

[0096] In some embodiments, in response to the presence of identical attribute identifiers in the attribute information set, a first attribute value for the identical attribute identifier in the first attribute information and a second attribute value for the identical attribute identifier in the second attribute information are determined. Further, a first confidence level of the first attribute information and a second confidence level of the second attribute information are obtained, and the first and second attribute values ​​are processed based on the first and second confidence levels to obtain the target attribute value for the identical attribute identifier.

[0097] In some embodiments, based on the same attribute information and the corresponding target attribute value, the same attribute identifiers in the attribute information set are deduplicated to obtain the final attribute information of the product.

[0098] In some embodiments, the first attribute value and the second attribute value are sorted according to a first confidence level and a second confidence level, so that the attribute value with a higher confidence level can be selected as the target attribute value with the same attribute identifier.

[0099] In some embodiments, the first attribute value and the second attribute value are weighted according to a first confidence level and a second confidence level to obtain a weighted attribute value, which is used as the target attribute value for the same attribute identifier. Optionally, the weight of the corresponding attribute information is determined according to the confidence level; the higher the confidence level, the higher the weight. Based on the determined weights, the first attribute information and the second attribute value are weighted to obtain the attribute value for the same attribute identifier.

[0100] In some embodiments, it can be determined whether the first attribute information and the second attribute information have the same attribute identifier based on the attribute identifier in the first attribute information and the attribute identifier in the second attribute information. In response to the existence of the same attribute identifier in the first attribute information and the second attribute information, the first attribute value in the first attribute information and the second attribute value in the second attribute information corresponding to the same attribute identifier can be obtained. If the first attribute value and the second attribute value are different, the first attribute value and the second attribute value can be processed based on the first confidence level of the first attribute information and the second confidence level of the second attribute information to determine the target attribute value of the same attribute identifier.

[0101] In some embodiments, in response to the presence of the same attribute identifier in the first attribute information and the second attribute information, the first attribute value and the second attribute value are weighted according to the first confidence level of the first attribute information and the second confidence level of the second attribute information to obtain the target attribute value with the same attribute identifier.

[0102] In this embodiment of the disclosure, when completing the attribute information of a product, confidence level can be considered to determine the weight of each of the two attribute values. The two attribute values ​​are weighted by weight to determine the attribute value of the attribute identifier, which can make the acquisition of attribute information more accurate.

[0103] In this embodiment of the disclosure, the construction of a product attribute library by combining historical session data and product details page can enrich product attributes from both the user's and the merchant's perspectives. This results in a product attribute library that is closer to real application scenarios, more comprehensive, and more timely, which can effectively improve the accuracy of matching user needs with products in scenarios such as product search, product recommendation, and intelligent customer service.

[0104] Furthermore, after building the product attribute library, continuous collection of application session data and mining based on this data enable closed-loop optimization of the product attribute library, enhancing it and generating a significant flywheel effect. Moreover, mining based on application session data can continuously supplement the relationship between product attributes and user expressions, improving the applicability to different application scenarios. It can also capture trending keywords and changing demands, making the product attribute library timely and thus continuously improving the user experience.

[0105] Figure 3 This is a flowchart illustrating another method for obtaining product attribute information provided in an embodiment of this application. Figure 3 As shown, the methods for obtaining this product attribute information may include, but are not limited to, the following steps: S301 collects historical session data and product detail page data for products.

[0106] The specific implementation of step S301 can be adopted in any embodiment of this application, and the steps are described in detail here.

[0107] S302, mine the historical session data and product details page of the product to determine the first attribute information of the product in order to obtain the product attribute library.

[0108] The specific implementation of step S302 can be carried out using any embodiment of this application, and the steps are described in detail here.

[0109] S303 outputs response information based on at least one round of question information input by the user.

[0110] In some embodiments, based on at least one round of question information input by the user, a search is performed on a product attribute database to obtain search results, wherein the search results include at least the first attribute information and basic information of the product, and a target question-answer pair matching the question information is determined. Further, based on the target question-answer pair, the first attribute information, and the basic information, response information corresponding to the question information is output.

[0111] The specific implementation of step S303 can be implemented using any embodiment of this application, and the steps are described in detail here.

[0112] S304 displays a response correction control on the display interface while outputting the response to the problem information.

[0113] In some embodiments, after the text big model determines the response information, it can output the response information on the conversation page where it interacts with the user. Optionally, in order to ensure that the output response information meets the user's needs, a response correction control can be carried on the conversation page. The response correction control can trigger a response correction process, so that the recommender or provider of the product can optimize the response information output by the text big model.

[0114] S305, in response to the response correction control being triggered, receives input information for correction and generates target response information based on the input information.

[0115] In some embodiments, the response optimization control is triggered, which may include, but is not limited to, clicking the response correction control or sliding the response correction control.

[0116] In some embodiments, after the response optimization control is triggered, the response information output by the large text model can be automatically loaded into the dialog box, allowing the error-correcting user to make corrections based on the response information output by the large text model. Furthermore, the error-correcting user can enter information in the dialog box to modify the response information.

[0117] In some embodiments, after the response optimization control is triggered, the voice assistant can also be invoked to collect the voice input of the error-correcting user, and then receive the input information for correction.

[0118] In some embodiments, after the response optimization control is triggered, a correction interface or dialog box may be displayed. Furthermore, the error correction user can enter information in the dialog box to modify the response information.

[0119] S306, Based on the input question information and its corresponding response information or target response information, determine the application session data.

[0120] In other words, the optimized response information in the application session data can be replaced based on the target response information.

[0121] S307, perform supplementary mining on the application session data to obtain the second attribute information of the product, and complete the attribute information of the product in the product attribute library based on the second attribute information.

[0122] The specific implementation of step S307 can be adopted in any embodiment of this application, and the steps are described in detail here.

[0123] In this embodiment of the disclosure, when a deviation is detected in the response information output by the large text model, a manual error correction process can be triggered to ensure that erroneous response information can be detected in a timely manner, thereby supplementing and improving the correct product attribute information.

[0124] In some embodiments, in response to the completion of correction, the correction question-answer pair formed by the target response information and the corresponding question information can be recorded and the correction question-answer pair can be stored synchronously in the correction question-answer database.

[0125] In some embodiments, subsequent question-and-answer retrieval logic can be optimized based on corrected question-and-answer pairs. In response to receiving new question-and-answer information from a subsequent user input, if the new input is the same as or similar to a previously asked question, and this same or similar question has been corrected, vector retrieval techniques can be used to preferentially retrieve target corrected question-and-answer pairs matching the new question from the corrected question-and-answer database. Furthermore, these pairs, along with standardized attribute information and basic information from a product attribute database, are input into a large text model to generate accurate response information.

[0126] In this embodiment, on the one hand, a corrective question-and-answer pair database can be established. By correcting the question-and-answer pair database, the product attribute database can be updated in real time, automatically correcting corresponding incorrect attributes and supplementing missing attributes, ensuring the consistency between the attribute database data and the actual product information. On the other hand, the priority of corrective question-and-answer pairs can be configured to be higher than the priority of basic attribute information, thereby ensuring the accuracy and relevance of the responses and avoiding the recurrence of similar incorrect responses.

[0127] Based on the above embodiments, taking an intelligent customer service scenario as an example, the process of obtaining product attribute information will be explained. Figure 4 As shown, product attribute acquisition involves a mining layer, an application layer, and a correction layer.

[0128] In some embodiments, the mining layer can automatically capture two core data sources through a mining engine. One source is historical conversation data between merchants and users, which can include interactive text from all scenarios, such as pre-sales consultations, after-sales feedback, and usage questions. This historical conversation data can be used to extract detailed attributes not previously reported by the merchant. The other source is product detail pages, which contain multimodal information, including product text descriptions, real-life product images, and feature demonstration video clips, used to extract visual attributes such as product appearance, material, and specifications. It is understood that historical conversation data can be understood and analyzed to obtain identification information of products that users are interested in, and then the product detail pages can be retrieved based on this identification information.

[0129] Optionally, historical session data can be input into a large text model for understanding, and the unreported attribute information of the products can be extracted. Further, after semi-structured processing, the data can be stored in the product attribute library.

[0130] Optionally, the product details page can be input into a multimodal big data model. This model performs deep analysis and feature extraction on the collected text, images, videos, and other data types to accurately identify the objective attribute information of the product. Simultaneously, it automatically removes invalid, redundant, and duplicate attribute information, improving the accuracy of the mined product attribute information. Furthermore, after the multimodal big data model parses the product details page, it standardizes and semi-structures the output attribute information, unifying the data format and expression specifications, and stores it in a product attribute database, laying the foundation for subsequent reuse.

[0131] In some embodiments, the application layer can provide conversational services to users based on the product attribute library built by the mining layer. The application layer obtains the user's input question information through the conversation page provided to the user, and then inputs it into the text big data model. Together with the product attribute information and basic information extracted from the product attribute library, it generates the corresponding answer information for the question information.

[0132] In some embodiments, the correction layer may provide a session page to the merchant, who can correct or optimize the response information output by the text model by triggering the response correction control on the session page.

[0133] In some embodiments, the correction layer may record the correction question-answer pair formed by the target response information and the corresponding question information, and store it synchronously in the correction question-answer library.

[0134] In some embodiments, if the new question-and-answer information entered by the user is the same as or similar to a question the user has previously asked, and that same or similar question has been corrected, vector retrieval technology can be used to preferentially retrieve a target corrected question-and-answer pair that matches the new question from the corrected question-and-answer database. Furthermore, this information, along with standardized attribute information and basic information from the product attribute database, is input into the large text model to generate accurate response information.

[0135] The method for obtaining product attribute information provided in this disclosure can achieve normalized, large-scale, and automated mining of product attribute information. It effectively supplements a large number of attribute dimensions, such as detailed attributes and scenario-based attributes, that are missing in the traditional merchant reporting model, thus significantly improving the completeness of product attribute information. Furthermore, the standardized and structured product attribute library mined in this disclosure can successfully provide a stable and high-quality data source for multiple core business operations of the platform. Specific application scenarios are as follows: 1. Intelligent Recommendation: Based on complete attributes, it achieves accurate matching between user needs and products, effectively improving the intelligence level of platform recommendations; 2. Product objective attribute card generation: Provides users with comprehensive and clear information on various product attributes, helping them quickly understand the core features of the product and improving output accuracy; 3. Merchant Managed Operations: Providing merchants with data support such as suggestions for improving attributes, user needs analysis, and popular attribute references can enhance the richness of product information, thereby improving the matching degree with user needs and further improving the user experience; 4. Live Q&A Assistance: Supports quick and accurate answers to users' real-time inquiries in live streaming scenarios, relieving the pressure on live stream hosts and improving the user viewing experience.

[0136] To achieve the above embodiments, this disclosure also proposes a device for obtaining product attribute information.

[0137] Figure 5 A schematic diagram of the structure of a device for obtaining product attribute information provided in an embodiment of this disclosure.

[0138] like Figure 5 As shown, the product attribute information acquisition device 500 includes: a mining module 501, a collection module 502, and a completion module 503.

[0139] The mining module 501 is used to mine the historical session data and product details page of the product to determine the first attribute information of the product in order to obtain the product attribute library; The data acquisition module 502 is used to collect application session data related to the product during the intelligent service application process; The mining module 501 is also used to perform supplementary mining on the application session data to obtain the second attribute information of the product; The completion module is used to complete the attribute information of the product in the product attribute library based on the second attribute information.

[0140] In some embodiments, the mining module 501 is further configured to: The multimodal large model is invoked to perform multimodal parsing and feature extraction on the product details page to obtain the first local attribute information of the product; The text big model is invoked to perform semantic parsing and entity extraction on the historical session data to obtain the second local attribute information of the product. The first attribute information includes at least the first local attribute information and the second local attribute information.

[0141] In some embodiments, the acquisition module 502 is further configured to: Based on at least one round of question information input by the user, the product attribute database is searched to obtain search results, which include at least the first attribute information and basic information of the product. Based on the first attribute information and basic information, output the response information for the question; The at least one round of question information and the corresponding response information are determined as the application session data.

[0142] In some embodiments, the acquisition module 502 is further configured to: The search information for the product is determined based on the question information; The product attribute database is searched based on the product's search information to obtain the search results.

[0143] In some embodiments, the acquisition module 502 is further configured to: Identify the target question-answer pair that matches the question information; Based on the target question-answer pair, the first attribute information, and the basic information, output the response information for the question information; The at least one round of question information and the corresponding response information are determined as the application session data.

[0144] In some embodiments, the acquisition module 502 is further configured to: While outputting the response to the question, a response correction control is displayed on the screen. In response to the triggering of the response correction control, input information for correction is received, and target response information is generated based on the input information. The target response information is used to replace the corresponding optimized response information in the application session data.

[0145] In some embodiments, the acquisition module 502 is further configured to: Based on the target response information and the corresponding question information, a corrective question-and-answer pair is generated. In response to receiving new question-and-answer information that contains similar question-and-answer information that has been corrected in the past, the system preferentially retrieves a target corrected question-and-answer pair that matches the new question-and-answer information from the corrected question-and-answer pairs.

[0146] In some embodiments, the completion module 503 is further configured to: Perform a union operation on the first attribute information and the second attribute information to obtain the attribute information set of the product; The attribute information set is deduplicated to obtain the final attribute information of the product.

[0147] In some embodiments, each type of attribute information includes an attribute identifier and an attribute value corresponding to the attribute identifier. The completion module 503 is further configured to: In response to the presence of identical attribute identifiers in the attribute information set, a first attribute value for the identical attribute identifiers in the first attribute information and a second attribute value for the identical attribute identifiers in the second attribute information are determined. Obtain the first confidence level of the first attribute information and the second confidence level of the second attribute information; Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are processed to obtain the target attribute value with the same attribute identifier; Based on the same attribute information and the corresponding target attribute value, the same attribute identifiers in the attribute information set are deduplicated to obtain the final attribute information of the product.

[0148] In some embodiments, the completion module 503 is further configured to: Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are sorted, and the attribute value with the higher confidence level is selected as the target attribute value for the same attribute identifier; or... Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are weighted to obtain a weighted attribute value, which is used as the target attribute value for the same attribute identifier.

[0149] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0150] In this embodiment of the disclosure, the construction of a product attribute library by combining historical session data and product details page can enrich product attributes from both the user's and the merchant's perspectives. This results in a product attribute library that is closer to real application scenarios, more comprehensive, and more timely, which can effectively improve the accuracy of matching user needs with products in scenarios such as product search, product recommendation, and intelligent customer service.

[0151] Furthermore, after building the product attribute library, continuous collection of application session data and mining based on this data enable closed-loop optimization of the product attribute library, enhancing it and generating a significant flywheel effect. Moreover, mining based on application session data can continuously supplement the relationship between product attributes and user expressions, improving the applicability to different application scenarios. It can also capture trending keywords and changing demands, making the product attribute library timely and thus continuously improving the user experience.

[0152] Figure 6A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0153] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 602 or a computer program loaded from storage unit 606 into RAM (Random Access Memory) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. I / O (Input / Output) interface 605 is also connected to bus 604.

[0154] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 606, such as various types of displays, speakers, etc.; storage unit 606, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0155] The computing unit 601 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the method for obtaining product attribute information. For example, in some embodiments, the method for obtaining product attribute information may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 606. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the method for obtaining product attribute information described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform a method for obtaining product attribute information by any other suitable means (e.g., by means of firmware).

[0156] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0158] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0159] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0160] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0161] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0162] According to embodiments of this disclosure, this disclosure also provides a computer program product that, when an instruction processor in the computer program product is executed, performs the method for obtaining product attribute information proposed in the above embodiments of this disclosure.

[0163] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0164] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for obtaining product attribute information, wherein, The method includes: By mining historical session data and product detail pages, the first attribute information of the product is determined to obtain a product attribute library; Collect application session data related to the product during the application of intelligent services; Further mining of the application session data yields the second attribute information of the product; Based on the second attribute information, the attribute information of the product in the product attribute library is completed.

2. The method according to claim 1, wherein, The process of mining historical session data and product detail pages to determine the product's first attribute information includes: The multimodal large model is invoked to perform multimodal parsing and feature extraction on the product details page to obtain the first local attribute information of the product; The text big model is invoked to perform semantic parsing and entity extraction on the historical session data to obtain the second local attribute information of the product. The first attribute information includes at least the first local attribute information and the second local attribute information.

3. The method according to claim 1, wherein, The application session data related to the product during the collection of intelligent service applications includes: Based on at least one round of question information input by the user, the product attribute database is searched to obtain search results, which include at least the first attribute information and basic information of the product. Based on the first attribute information and basic information, output the response information for the question; The at least one round of question information and the corresponding response information are determined as the application session data.

4. The method according to claim 3, wherein, The step of searching the product attribute database based on at least one round of user-input question information to obtain search results includes: The search information for the product is determined based on the question information; The product attribute database is searched based on the product's search information to obtain the search results.

5. The method according to claim 3, wherein, The step of outputting the response information for the question based on the first attribute information and the basic information includes: Identify the target question-answer pair that matches the question information; Based on the target question-answer pair, the first attribute information, and the basic information, output the response information for the question information; The at least one round of question information and the corresponding response information are determined as the application session data.

6. The method according to any one of claims 3-5, wherein, The method further includes: While outputting the response to the question, a response correction control is displayed on the screen. In response to the triggering of the response correction control, input information for correction is received, and target response information is generated based on the input information. The target response information is used to replace the corresponding optimized response information in the application session data.

7. The method according to claim 6, wherein, The method further includes: Based on the target response information and the corresponding question information, a corrective question-and-answer pair is generated. In response to receiving new question-and-answer information that contains similar question-and-answer information that has been corrected in the past, the system preferentially retrieves a target corrected question-and-answer pair that matches the new question-and-answer information from the corrected question-and-answer pairs.

8. The method according to any one of claims 1-5, wherein, The step of completing the attribute information of the product in the product attribute library based on the second attribute information includes: Perform a union operation on the first attribute information and the second attribute information to obtain the attribute information set of the product; The attribute information set is deduplicated to obtain the final attribute information of the product.

9. The method according to claim 8, wherein, Each type of attribute information includes an attribute identifier and the attribute value corresponding to the attribute identifier. The step of deduplicating the attribute information set to obtain the final attribute information of the product includes: In response to the presence of identical attribute identifiers in the attribute information set, a first attribute value for the identical attribute identifiers in the first attribute information and a second attribute value for the identical attribute identifiers in the second attribute information are determined. Obtain the first confidence level of the first attribute information and the second confidence level of the second attribute information; Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are processed to obtain the target attribute value with the same attribute identifier; Based on the same attribute information and the corresponding target attribute value, the same attribute identifiers in the attribute information set are deduplicated to obtain the final attribute information of the product.

10. The method according to claim 9, wherein, The step of processing the first attribute value and the second attribute value according to the first confidence level and the second confidence level to obtain the target attribute value with the same attribute identifier includes at least one of the following operations: Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are sorted, and the attribute value with the higher confidence level is selected as the target attribute value of the same attribute identifier; or, Based on the first confidence level and the second confidence level, the first attribute value and the second attribute value are weighted to obtain a weighted attribute value, which is used as the target attribute value for the same attribute identifier.

11. A device for acquiring product attribute information, comprising: The mining module is used to mine historical session data and product detail pages of products to determine the first attribute information of the products in order to obtain a product attribute library; The data acquisition module is used to collect application session data related to the product during the application of intelligent services. The mining module is also used to perform supplementary mining on the application session data to obtain the second attribute information of the product; The completion module is used to complete the attribute information of the product in the product attribute library based on the second attribute information.

12. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

14. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1-10.

Images