Method and device for constructing user portrait data and electronic equipment

Abstract

The embodiment of the invention discloses a method and device for constructing user portrait data and electronic equipment, and the method comprises the steps: obtaining UGC (User Generated Content) associated with a target user in a commodity information service system, and converting the UGC into a plurality of first vectors; receiving demand description content for expressing a data mining demand, and converting the demand description content into a second vector; the data mining demand is related to construction of user portrait data; screening out at least one piece of target UGC content related to the data mining demand from the plurality of pieces of UGC content by calculating the distance between the second vector and the plurality of first vectors; according to the at least one piece of target UGC content and the demand description content used for expressing the data mining demand, prompt information used for carrying out dialogue with an artificial intelligence AI large-scale parameter model is constructed, and the AI large-scale parameter model is used for carrying out dialogue with the target UGC content according to the target UGC content. And generating user features and / or label data for constructing the user portrait data. According to the embodiment of the invention, more comprehensive and more accurate user portrait data can be obtained.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to methods, devices and electronic devices for constructing user portrait data. Background Art

[0002] In the product information service system, whether it is to recommend personalized product information to consumers, or to perform personalized screening or sorting of recalled products in the product search scenario, etc., user portrait data will be used. The so-called user portrait data is to express user information in a labeled manner, collect data from various dimensions such as user consumption habits and preference characteristics, and then characterize user characteristics and attributes, analyze and count these characteristics, and mine potential valuable information, so as to abstract the overall picture of user information.

[0003] For example, based on the data related to the user's historical behavior records in the product information service system, the user's characteristics in multiple dimensions can be mined and then expressed through labels. Specifically, based on the user's basic attribute information (including gender, age, occupation, etc.), the number of orders in the last month or other time periods, the average order value, and other data, the user's purchasing power characteristics can be mined, and then the user can be labeled as "high purchasing power", "medium purchasing power" or "low purchasing power", etc.

[0004] In the prior art, the process of mining user portrait data mainly involves combing and mining data that can be expressed in a structured manner, thereby constructing user portrait data. For example, the order quantity and average order value in the above example are all data that can be expressed in a structured manner through statistics, queries, etc. (for example, they can be expressed in the form of key-value pairs such as attribute fields and attribute values). By processing these structured data through clustering or classification models, users can be labeled accordingly.

[0005] Although the above method can construct user portrait data, it may not be able to fully and accurately portray a specific user. Therefore, how to obtain more comprehensive and accurate user portrait data has become a technical problem that needs to be solved by technical personnel in this field. Summary of the invention

[0006] The present application provides a method, device and electronic device for constructing user portrait data, which is conducive to obtaining more comprehensive and accurate user portrait data.

[0007] This application provides the following solutions:

[0008] A method for constructing user portrait data, comprising:

[0009] Acquire multiple pieces of user-generated UGC content associated with a target user in a product information service system, and convert the multiple pieces of UGC content into multiple first vectors;

[0010] Receiving a requirement description content for expressing a data mining requirement, and converting the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data;

[0011] Filtering out at least one target UGC content related to the data mining requirement from the multiple UGC contents by calculating the distances between the second vector and the multiple first vectors respectively;

[0012] Based on the at least one target UGC content and the requirement description content for expressing data mining requirements, prompt information for communicating with an artificial intelligence (AI) large-scale parameter model is constructed. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

[0013] Among them, the target users include consumer users in the product information service system, and the multiple UGC contents include: conversation content generated by the consumer user during an online conversation with at least one merchant, evaluation content published by the consumer user on at least one product, and / or sharing content published by the consumer user based on at least one product.

[0014] Among them, the target users include merchant users in the product information service system, and the multiple UGC contents include: conversation content generated by the merchant user during an online conversation with at least one consumer user, and / or user evaluation content obtained by multiple products associated with the merchant user.

[0015] Among them, it also includes:

[0016] Contextual content related to the plurality of UGC contents is obtained, so as to construct the prompt text according to the at least one target UGC content, the contextual content, and the text description content for expressing data mining requirements.

[0017] The context content includes user behavior sequence information related to the UGC content.

[0018] The generated first vector and second vector include semantic information of the original text before conversion.

[0019] A device for constructing user portrait data, comprising:

[0020] A first vector conversion unit, configured to obtain a plurality of user-generated UGC contents associated with a target user in a commodity information service system, and convert the plurality of UGC contents into a plurality of first vectors;

[0021] A second vector conversion unit, configured to receive a requirement description content for expressing a data mining requirement, and convert the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data;

[0022] a target UGC content determination unit, configured to filter out at least one target UGC content related to the data mining requirement from the multiple pieces of UGC content by calculating the distances between the second vector and the multiple first vectors respectively;

[0023] A prompt text construction unit is used to construct prompt information for communicating with an artificial intelligence (AI) large-scale parameter model based on the at least one target UGC content and the requirement description content for expressing data mining requirements. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

[0024] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the methods described above.

[0025] An electronic device, comprising:

[0026] one or more processors; and

[0027] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of any of the methods described above.

[0028] A computer program product comprises a computer program / computer executable instructions, wherein the computer program / computer executable instructions implement the steps of any of the aforementioned methods when executed by a processor in an electronic device.

[0029] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0030] Through the embodiment of the present application, the user-associated UGC (User Generated Content) content can be analyzed by the AI ​​(Artificial Intelligence) large model for the construction of user portrait data. Among them, since the AI ​​large model has a length limit on the input word, the UGC content can be preprocessed first. The specific processing method is to convert multiple UGC contents into multiple first vectors. In addition, the demand description content used to express the data mining demand can also be converted into a second vector. After that, at least one target UGC content related to the data mining demand can be screened out from the multiple UGC contents by calculating the distance between the second vector and the multiple first vectors respectively. Finally, according to the at least one target UGC content and the demand description content used to express the data mining demand, a prompt information for dialogue with the artificial intelligence AI large-scale parameter model can be constructed, so that the AI ​​large-scale parameter model generates user features and / or label data for constructing user portrait data according to the target UGC content. In this way, the data mining process can be made to focus more on extracting valuable information from content (such as conversations, blog posts, videos, reviews, etc.) created or shared by users themselves. This process involves text analysis, sentiment analysis, topic discovery, etc., so as to deeply understand the user's views, feelings and preferences, and build more comprehensive and accurate user portrait data, which is helpful for improving product services, customizing marketing strategies and enhancing user experience.

[0031] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 It is a schematic diagram of the system architecture provided by the embodiment of the present application;

[0034] Figure 2 is a flow chart of the method provided in the embodiment of the present application;

[0035] Figure 3 is a schematic diagram of a device provided in an embodiment of the present application;

[0036] Figure 4 It is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0038] First of all, it should be noted that user portraits are a marketing and product design tool that refers to the creation of fictional characters that represent a group of users. These characters are built based on the common characteristics, behaviors, interests, and needs of the user group. Through user portraits, you can better understand your target customers and provide them with more customized services or products. Based on user portraits, you can also perform user modeling, which involves the process of analyzing and abstracting user behaviors, characteristics, and preference labels, aiming to build a model that can predict user behavior or responses. This model can be used for recommendation systems, personalized content display, user experience optimization, etc. User modeling can help the system better understand user needs and provide more tailored services.

[0039] In the embodiment of the present application, it is considered that in addition to the behavioral sequence content generated by users in the product information service system, some UGC (User Generated Content) will also be generated. Among them, the behavioral sequence content can include information related to the user's browsing, adding to shopping carts, collecting, purchasing and other behaviors of the product when using the e-commerce platform, including related products, product categories, prices, the number of purchases within a certain period of time, etc. UGC content refers to the conversation content generated by the user in the process of using the product information service system, the conversation content generated in the process of communicating with other users, the comment content on the product, the shared "buyer show", the content of product evaluation, etc.

[0040] Among them, the content related to the above-mentioned behavior sequence belongs to the content that can be expressed in a structured manner. For these contents, as described in the background technology section, they can be processed through classification models, clustering models, etc., so as to mine user characteristics, or directly add tags to users, and then build user portrait data; however, for UGC content, it is usually difficult to express it in a structured way, which makes it difficult to perform data mining processing through existing classification models, clustering models, etc. However, this kind of UGC content actually contains a lot of information, and relevant information can be mined from it, which is helpful for building a more comprehensive and accurate user portrait.

[0041] With the emergence of AI (Artificial Intelligence) large-scale parameter models (AI big models for short), they have been applied in multiple scenarios and also provide a new idea for user portrait modeling. The so-called AI big model refers to a deep learning model containing a large number of parameters. Due to its large scale, this AI big model can store and process a large amount of information, thereby achieving higher performance in various tasks. Therefore, relying on the powerful natural language text understanding and logical thinking ability of the pre-trained AI big model, the unstructured UGC content can be analyzed. The analysis results can be used as a supplement to the user portrait data constructed based on the user behavior sequence, or the AI ​​big model can cross-analyze the UGC content and the user behavior sequence content to build a more accurate and comprehensive user portrait data. In short, through this UGC content analysis based on the AI ​​big model, the data mining process focuses more on extracting valuable information from the content created or shared by users themselves (such as conversations, blog posts, videos, reviews, etc.). This process involves technologies such as text analysis, sentiment analysis, and topic discovery, so as to deeply understand the user's views, feelings, and preferences, which is very helpful for improving product services, customized marketing strategies, and enhancing user experience.

[0042] However, when using the AI ​​big model to process the above UGC content, one problem that may be encountered is that in order to perform user portrait analysis, it is usually necessary to aggregate the conversation content and comment content of a user with multiple businesses or other users, and then analyze them by the AI ​​big model. As a result, the UGC content to be analyzed for each user may be very long, for example, it may include a lot of conversation content, etc.; however, the input token length supported by the AI ​​big model is usually limited. Even if the UGC content of a single user is analyzed, it is impossible to directly input all the UGC content of a single user into the AI ​​big model for analysis.

[0043] In view of the above problems, the embodiments of the present application provide corresponding solutions. In this solution, the user's UGC content can first be mathematically expressed through processing methods such as Embedding to obtain multiple vectors (referred to as "first vectors" here for the sake of distinction), wherein the UGC content of the same user can be expressed as multiple first vectors. For example, assuming that a user's UGC content includes 1,000 sentences, it can be converted into 1,000 vectors. Of course, the number of sentences and the number of vectors may be different. In addition, when interacting with the AI ​​big model, specific needs can usually be expressed through dialogue and other methods. For example, if the AI ​​big model is required to find content related to "bargaining" from the UGC content, the following text can be input into the AI ​​big model: "Please filter out the content belonging to bargaining from the following content", etc. It should be noted here that in addition to expressing data mining needs through text, it can also be expressed through pictures, voices, etc. Of course, in the specific implementation, for content such as pictures and voices, it can also be converted into text content first, and then subsequent processing, etc. In short, the requirement description content for expressing data mining requirements can be determined first (the requirement description content can also be called Query), and then the requirement description content can be expressed as a vector (which can be called "second vector") through Embedding and other methods.

[0044] Through the above method, multiple first vectors can be obtained, which are related to the user's UGC content, and a second vector can be obtained, which is related to specific data mining needs. Afterwards, the second vector can be used to recall at least one first vector related to the current data mining needs from multiple first vectors. For example, the distance between the second vector and multiple first vectors can be calculated, and the similarity between vectors can be expressed by this distance. The top N first vectors with the highest similarity, or the first vector with a similarity higher than a certain threshold, can be recalled. Afterwards, this recalled first vector can also be converted back into text content (which can be called "target UGC content"). Since the first vector has been screened in the process of vector recall, the content irrelevant to the current data mining needs has been filtered out, and only the first vector related to the current data mining needs has been retained. Therefore, after converting this first vector back to text content, the length of the target UGC content obtained will be shorter than the original UGC content, and can be controlled within the maximum input token length supported by the AI ​​big model. Therefore, this target UGC content can be used together with the aforementioned designed data mining demand text description content to construct prompt text (prompts) for dialogue with the AI ​​big model and input it into the AI ​​big model. Of course, the specific data mining demand description text can be related to the features or labels required to construct user portrait data, so that the AI ​​big model can generate user features and / or label data for constructing user portrait data based on the target UGC content, and then construct specific user portrait data.

[0045] In addition, since the user's UGC content may also be associated with some related contextual content, this contextual content may also include some content related to the user behavior sequence. For example, during the conversation between a user and a merchant or after the conversation is completed, the user may place an order for a certain product. At this time, the product information, order behavior information, etc. belong to the contextual content of this conversation, and the order behavior belongs to the content related to the user behavior sequence, and so on. This contextual content combined with the UGC content can more completely and comprehensively reflect the user's characteristics and other information. Therefore, in the preferred implementation of the present application, the AI ​​big model can also be used to cross-analyze the above-mentioned recalled target UGC content and related contextual content, so as to obtain more accurate user portrait data. Specifically, since the contextual content is usually not very long, it is not necessary to perform the aforementioned conversion into a vector and then perform operations such as recall. Instead, after the target UGC content is screened out through the aforementioned method, the target UGC content, the contextual content, the data mining demand description text, etc. can be used together to construct prompts and input them into the AI ​​big model for analysis. In this way, the AI ​​big model can combine the target UGC content with contextual content such as associated user behavior sequences to generate more accurate user portrait data.

[0046] From the perspective of system architecture, see Figure 1 , an embodiment of the present application provides a user portrait data analysis system, in which user portrait data can be generated based on the user's UGC content. In the generation process, an AI big model is needed. Considering the limitation of the AI ​​big model on the input token length, the UGC content can first be vectorized and expressed as multiple first vectors. In addition, the demand description content used to express the data mining demand can also be converted into a second vector, so that the specific UGC content can be judged whether it is related to the current data mining demand by calculating the distance between vectors, and the relevant part of the target UGC content can be screened out, and then input into the AI ​​big model for analysis. In a preferred manner, the screened target UGC content can also be input into the AI ​​big model together with the relevant context content for cross-analysis. This context content can include content related to the user behavior sequence, thereby realizing the combination of the user behavior sequence content and the UGC content, and jointly constructing a more comprehensive and accurate user portrait data.

[0047] The specific implementation scheme provided in the embodiments of the present application is described in detail below.

[0048] First, the present application embodiment provides a method for constructing user portrait data, see Figure 2 , the method may include:

[0049] S201: Acquire user-generated UGC content associated with a target user in a product information service system, and convert the plurality of UGC content pieces into a plurality of first vectors.

[0050] In the embodiment of the present application, user portrait data can be mainly constructed through the UGC content associated with the user, wherein the "user" can be a consumer user in the commodity information service system, a merchant user in the commodity information service system, etc. That is to say, under normal circumstances, user portrait data is mainly constructed for consumer users in order to make some personalized commodity recommendations to users, etc., but in actual applications, there is also a demand for constructing user portrait data for merchant users, for example, some marketing strategy suggestions can be recommended to merchant users based on merchant user portrait data, etc.

[0051] Specifically, for different users, different aggregation methods may be used when aggregating UGC content. For example, if the target user is a consumer user in a commodity information service system, the multiple pieces of UGC content that are aggregated may include: the conversation content generated by the consumer user during an online conversation with at least one merchant, the evaluation content published by the consumer user on at least one commodity, and / or the sharing content published by the consumer user based on at least one commodity, etc. Multiple different types of UGC content associated with the same user may be aggregated together to generate features or tags related to the user profile of the user.

[0052] If the target user is a merchant user in the product information service system, the specific multiple UGC contents may include: the conversation content generated by the merchant user in the process of online conversation with at least one consumer user, and / or the user evaluation content obtained by multiple products associated with the merchant user. In other words, the conversation content between the same merchant and multiple different users and the evaluation content of multiple products associated with the merchant can be aggregated to perform user portrait data analysis on the merchant, etc.

[0053] After obtaining the UGC content associated with the same user, since its content length may exceed the maximum length of the input token supported by the AI ​​large model, it can be preprocessed first. Specifically, when preprocessing, the UGC content can be first converted into a vector of mathematical expression. Specifically, when converting, it can be achieved through Embedding.

[0054] Among them, Embedding usually refers to the process of mapping high-dimensional data (such as text, pictures, audio) to a low-dimensional space. An Embedding vector is usually a vector composed of real numbers, which represents the input data as a point in a continuous numerical space. In other words, an Embedding vector is an N-dimensional real-valued vector, which can be used to represent almost any object, such as text, music, video, etc. As for the Embedding vector of text, this vector can also represent the semantics of words or sentences. This is because these Embedding vectors are learned based on the appearance pattern of words or sentences in the language context.

[0055] Based on the above characteristics, specific UGC content can be converted into Embedding vectors for expression. In specific implementation, there can be a variety of specific conversion methods. For example, the UGC content can be first segmented to obtain multiple words or phrases, and then each word or phrase can be converted into an Embedding vector. The Embedding vectors corresponding to multiple words or phrases in the same sentence can then be summed or averaged to obtain the Embedding vector corresponding to each sentence, etc. In this way, multiple sentences in the UGC content can be converted into Embedding vectors to obtain multiple Embedding vectors. Here, the vector obtained by converting the UGC content is called the first vector.

[0056] It should be noted that, regarding the above process of converting UGC content into vectors, in a specific implementation method, the vectorization processing of UGC content can be realized through the RAG (Retrieval-Augmented Generation) framework, and the generated multiple first vectors can be saved in the database. After the second vector is generated according to the query, the target first vector that is close to the second vector can be recalled from the database. Of course, in practical applications, the above conversion process can also be implemented in other ways.

[0057] In short, there are usually a lot of UGC contents associated with the same user. However, for a specific data mining requirement, only some of the UGC contents may be relevant to the data mining requirement, while the others are irrelevant. These irrelevant UGC contents are useless for the data mining requirement and can be filtered out. This can remove the interference of this useless data on the one hand, and shorten the length of the input content to make it within the maximum length range of the input token supported by the AI ​​large model on the other hand. When filtering these useless data, it can be achieved through mathematical calculations. Therefore, the UGC content needs to be converted into multiple first vectors first.

[0058] S202: Receive a requirement description content for expressing a data mining requirement, and convert the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data.

[0059] In addition to converting UGC content into multiple first vectors, data mining needs can also be expressed, and the corresponding demand description content can be called a query. Specifically, this query can be text content expressed in natural language, or it can also be pictures, voice content, or a combination of pictures and text, and so on. Of course, for demand description content such as pictures and voices, they can also be converted into text content first and then processed, or, in another way, subsequent vector generation and other processing can be performed directly based on the pictures and voice content. After determining the demand description content, this demand description content (that is, query) can also be converted into a second vector.

[0060] Among them, the demand description content corresponding to the above data mining demand, that is, the query, will also participate in the construction of the prompt information (prompt) later. In an embodiment of the present application, the specific data mining demand may be a demand related to the construction of user portrait data. For example, a query may be "Please summarize the user's shopping concerns based on the user's bargaining content." From the text description content, it can be seen that the main focus is on UGC content related to "bargaining." Therefore, the UGC content related to "bargaining" can be filtered out from all the user's UGC content. In order to complete the screening task, it can be performed by mathematical calculation, so the UGC content and the above text description content can be converted into vectors of mathematical expression.

[0061] S203: Filter out at least one target UGC content related to the data mining requirement from the multiple UGC contents by calculating the distances between the second vector and the multiple first vectors respectively.

[0062] After converting the UGC content into multiple first vectors and converting the demand description content used to describe the data mining demand into a second vector, the distance between the second vector and each first vector can be calculated. As mentioned above, since the specific vector can be realized by Embedding and other methods, and the Embedding vector can retain the semantic information in the text, the vector distance calculation based on the vector calculation process can also reflect the semantic similarity between the corresponding sentences. In other words, if the distance between a first vector and a second vector is relatively close, it also means that the semantics between the sentence corresponding to the first vector and the text description information corresponding to the aforementioned data mining demand are relatively close, and then prove that the sentence corresponding to the first vector is related to the data mining demand. Therefore, after calculating the distance between the second vector and the multiple first vectors, the N first vectors with the closest distance can be extracted, or the multiple first vectors with a distance less than the preset distance, etc., and the sentences corresponding to these first vectors belong to the target UGC content related to the data mining demand.

[0063] S204: Construct prompt information for communicating with an artificial intelligence (AI) large-scale parameter model based on the at least one target UGC content and the requirement description content for expressing data mining requirements. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

[0064] After determining at least one first vector that is closest to the second vector, these first vectors can be converted back into sentences expressed in natural language. That is, since in the aforementioned step S201, multiple sentences in the UGC content are converted into first vectors respectively, after recalling multiple first vectors with close distances through the second vector, these first vectors can be converted back into corresponding sentences in the UGC content, and these sentences belong to the screened target UGC content related to the data mining requirements.

[0065] After obtaining the at least one target UGC content mentioned above, prompt information (prompt) for communicating with the AI ​​big model can be constructed based on the at least one target UGC content and the requirement description content used to express the data mining requirements in step S202, so that the AI ​​big model can generate user features and / or label data for constructing user portrait data based on these target UGC contents.

[0066] Among them, the AI ​​big model can generate the above-mentioned user features or label data by natural language understanding, reasoning, summarizing and summarizing the target UGC. Of course, in practical applications, in order to make the features or labels generated by the AI ​​big model more suitable for e-commerce scenarios, the AI ​​big model can also be fine-tuned and trained in advance using relevant data in the e-commerce scenario. For example, some UGC content in historical records can be manually labeled, and the labeling results can include what kind of user features or label data can be generated from these UGC contents, etc. These data can be used as training samples to train the AI ​​big model.

[0067] In addition, since UGC content is usually associated with contextual content, for example, a user may be browsing a certain product before initiating a conversation with a merchant, and may place an order for the product after initiating the conversation. At this time, the product information being browsed before the conversation and the product ordered after the conversation belong to the contextual information of the conversation content. This contextual information will include content related to the user behavior sequence, so this contextual information can also be input into the AI ​​big model for analysis. In this way, the AI ​​big model can cross-analyze contextual content such as UGC content and behavior sequence content to obtain more accurate user portrait data.

[0068] In summary, through the embodiments of the present application, the UGC content associated with the user can be analyzed by the AI ​​large model for the construction of user portrait data. Among them, since the AI ​​large model has a length limit on the input word, the UGC content can be preprocessed first. The specific processing method is to convert multiple UGC contents into multiple first vectors. In addition, the demand description content used to express the data mining needs can also be converted into a second vector. After that, at least one target UGC content related to the data mining needs can be screened out from the multiple UGC contents by calculating the distance between the second vector and the multiple first vectors. Finally, according to the at least one target UGC content and the demand description content used to express the data mining needs, prompt information for dialogue with the artificial intelligence AI large-scale parameter model can be constructed, so that the AI ​​large-scale parameter model generates user features and / or label data for constructing user portrait data according to the target UGC content. In this way, the data mining process can be made more focused on extracting valuable information from content (such as conversations, blog posts, videos, reviews, etc.) created or shared by users themselves. This process involves text analysis, sentiment analysis, topic discovery, etc., so as to deeply understand users' opinions, feelings and preferences, which is helpful for improving product services, customizing marketing strategies and enhancing user experience.

[0069] In a preferred implementation, the contextual content of the UGC content can also be input into the AI ​​big model for analysis. This contextual content can include content related to the user behavior sequence, so that the AI ​​big model can cross-analyze the UGC content and the user behavior sequence content, which is conducive to obtaining more comprehensive and accurate user portrait data.

[0070] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0071] Corresponding to the above method embodiment, the present application embodiment also provides a device for constructing user portrait data, see Figure 3 , the device may include:

[0072] A first vector conversion unit 301 is used to obtain multiple pieces of user-generated UGC content associated with a target user in a product information service system, and convert the multiple pieces of UGC content into multiple first vectors;

[0073] A second vector conversion unit 302 is used to receive a requirement description content for expressing a data mining requirement, and convert the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data;

[0074] A target UGC content determination unit 303 is configured to filter out at least one target UGC content related to the data mining requirement from the multiple pieces of UGC content by calculating the distances between the second vector and the multiple first vectors respectively;

[0075] The prompt text construction unit 304 is used to construct prompt information for communicating with the artificial intelligence AI large-scale parameter model based on the at least one target UGC content and the requirement description content for expressing data mining requirements. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

[0076] Among them, the target users include consumer users in the product information service system, and the multiple UGC contents include: conversation content generated by the consumer user during an online conversation with at least one merchant, evaluation content published by the consumer user on at least one product, and / or sharing content published by the consumer user based on at least one product.

[0077] Alternatively, the target user includes a merchant user in a product information service system, and the multiple UGC contents include: conversation content generated by the merchant user during an online conversation with at least one consumer user, and / or user evaluation content obtained for multiple products associated with the merchant user.

[0078] In a specific implementation, the device may further include:

[0079] The context content acquisition unit is used to acquire context content related to the multiple UGC contents so as to construct the prompt text according to the at least one target UGC content, the context content, and the text description content for expressing data mining requirements.

[0080] The context content includes user behavior sequence information related to the UGC content.

[0081] In a specific implementation, the generated first vector and second vector include semantic information of the original text before conversion.

[0082] In addition, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods in the aforementioned method embodiments are implemented.

[0083] And an electronic device, comprising:

[0084] one or more processors; and

[0085] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.

[0086] A computer program product includes a computer program / computer executable instructions, which, when executed by a processor in an electronic device, implement the steps of the method described in the aforementioned method embodiment.

[0087] in, Figure 4 The architecture of the electronic device is shown as an example, which may include a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, and a memory 420. The processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420 may be communicatively connected via a communication bus 430.

[0088] Among them, the processor 410 can be implemented by a general-purpose CPU (Central Processing Unit, processor), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solution provided in this application.

[0089] The memory 420 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 420 can store an operating system 421 for controlling the operation of the electronic device 400, and a basic input and output system (BIOS) for controlling the low-level operation of the electronic device 400. In addition, a web browser 423, a data storage management system 424, and a user portrait data processing system 425, etc. can also be stored. The above-mentioned user portrait data processing system 425 can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present application. In short, when the technical solution provided in the present application is implemented by software or firmware, the relevant program code is stored in the memory 420 and is called and executed by the processor 410.

[0090] The input / output interface 413 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0091] The network interface 414 is used to connect to a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0092] The bus 430 comprises a pathway for transmitting information between the various components of the device (eg, the processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420).

[0093] It should be noted that, although the above device only shows a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, a memory 420, a bus 430, etc., in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include components necessary for implementing the solution of the present application, and does not necessarily include all the components shown in the figure.

[0094] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0095] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0096] The above is a detailed introduction to the method, device and electronic device for constructing user portrait data provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application. At the same time, for those skilled in the art, according to the ideas of the present application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. A method for constructing user portrait data, characterized in that: include: Acquire multiple pieces of user-generated UGC content associated with a target user in a product information service system, and convert the multiple pieces of UGC content into multiple first vectors; Receiving a requirement description content for expressing a data mining requirement, and converting the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data; Filtering out at least one target UGC content related to the data mining requirement from the multiple UGC contents by calculating the distances between the second vector and the multiple first vectors respectively; Based on the at least one target UGC content and the requirement description content for expressing data mining requirements, prompt information for communicating with an artificial intelligence (AI) large-scale parameter model is constructed. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

2. The method according to claim 1, characterized in that: The target users include consumer users in the product information service system, and the multiple UGC contents include: conversation content generated by the consumer users during an online conversation with at least one merchant, evaluation content published by the consumer users on at least one product, and / or sharing content published by the consumer users based on at least one product.

3. The method according to claim 1, characterized in that The target users include merchant users in the commodity information service system, and the multiple UGC contents include: conversation contents generated by the merchant user during an online conversation with at least one consumer user, and / or user evaluation contents respectively obtained by multiple commodities associated with the merchant user.

4. The method according to claim 1, characterized in that: Also includes: Contextual content related to the plurality of UGC contents is obtained, so as to construct the prompt text according to the at least one target UGC content, the contextual content, and the text description content for expressing data mining requirements.

5. The method according to claim 4, characterized in that The context content includes user behavior sequence information related to the UGC content.

6. The method according to claim 1, characterized in that The generated first vector and second vector include semantic information of the original text before conversion.

7. A device for constructing user portrait data, characterized in that: include: A first vector conversion unit, configured to obtain a plurality of user-generated UGC contents associated with a target user in a commodity information service system, and convert the plurality of UGC contents into a plurality of first vectors; A second vector conversion unit, configured to receive a requirement description content for expressing a data mining requirement, and convert the requirement description content into a second vector; the data mining requirement is related to constructing user portrait data; a target UGC content determination unit, configured to filter out at least one target UGC content related to the data mining requirement from the multiple pieces of UGC content by calculating the distances between the second vector and the multiple first vectors respectively; A prompt text construction unit is used to construct prompt information for communicating with an artificial intelligence (AI) large-scale parameter model based on the at least one target UGC content and the requirement description content for expressing data mining requirements. The AI ​​large-scale parameter model is used to generate user features and / or label data for constructing user portrait data based on the target UGC content.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

9. An electronic device, characterized in that: include: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of claims 1 to 6.

10. A computer program product comprising a computer program / computer executable instructions, characterized in that: When the computer program / computer executable instructions are executed by a processor in an electronic device, the steps of the method according to any one of claims 1 to 6 are implemented.

Images