Goods-carrying person-arriving recommendation method and device, electronic equipment and medium
By acquiring multi-dimensional feature data of the products to be promoted and profile data of sales influencers, and calculating the matching degree, the problem of time-consuming, labor-intensive and inaccurate manual screening of sales influencers in existing technologies is solved, and efficient and accurate recommendations of sales influencers are achieved.
Patent Information
- Application Number
- CN202511327045.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-17
- Publication Date
- 2026-01-09
AI Technical Summary
In existing technologies, merchants rely on manual methods to select influencers, which consumes a lot of time and energy and makes it difficult to ensure the accuracy of the selection. Especially when faced with a large and diverse pool of influencers and complex product promotion needs, it is impossible to achieve accurate matching.
By acquiring multi-dimensional product feature data of the products to be promoted and combining it with the user profile data of the influencers, the matching degree is calculated in dimensions such as product content attributes, commercial attributes and target audience. The target influencers who are highly compatible with the products to be promoted are accurately selected, and the recommendation reasons are generated using a large language model and sent to the merchant's terminal.
It improves the accuracy and efficiency of selecting influencers, reduces the subjectivity and inefficiency caused by reliance on manual methods, and achieves objective and accurate influencer recommendations.
Smart Images
Figure CN121304282A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of big data processing, and in particular to a method and device for recommending a product-promoting influencer, an electronic device and a medium. BACKGROUND
[0002] With the rapid development of the e-commerce industry, product promotion by influential product-promoting influencers has become an important way for various brands to expand their markets and increase product awareness, and is widely used in various commercial scenarios at home and abroad.
[0003] At present, when selecting a product-promoting influencer for cooperation, a merchant mainly relies on manual methods. Specifically, the merchant usually refers to the basic data provided by each platform, such as the number of fans of the influencer, the average number of video plays, the number of interactions, and limited historical product promotion records, and makes a judgment in combination with its own market experience. Some platforms provide preliminary auxiliary tools, such as classification based on a broad field of tags or simple sorting according to a single indicator (such as the number of fans), but these tools can only achieve very rough preliminary screening.
[0004] This manual approach not only consumes a lot of time and effort, but also makes it difficult to ensure the accuracy of the selection when faced with a large number of product-promoting influencer resources and complex and diverse product promotion needs. SUMMARY
[0005] The present application provides a method and device for recommending a product-promoting influencer, an electronic device and a medium, which solves the technical defect that the manual approach in the prior art not only consumes a lot of time and effort, but also makes it difficult to ensure the accuracy of the selection when faced with a large number of product-promoting influencer resources and complex and diverse product promotion needs, and enables the internal relevance between a product and a product-promoting influencer to be deeply mined, thereby improving the objectivity and accuracy of the matching of a product and a product-promoting influencer.
[0006] The present application provides a method for recommending a product-promoting influencer, which comprises: obtaining product information of a product to be promoted and user portrait data of at least one product-promoting influencer; extracting features from the product information of the product to be promoted to obtain product feature data of each preset dimension of the product to be promoted; the preset dimensions include at least one of a product content attribute dimension, a product commercial attribute dimension and a product target audience dimension; determining the matching degree of the product-promoting influencer and the product to be promoted in each preset dimension according to user feature data in the user portrait data of the product-promoting influencer that matches the product feature data of each preset dimension; determining at least one target product-promoting influencer from the at least one product-promoting influencer according to the matching degree of the product-promoting influencer and the product to be promoted in each preset dimension.
[0007] According to the method for recommending a cargo influencer provided by the application, the preset dimensions include the commodity content attribute dimension, and the determination of the matching degree of the cargo influencer and the commodity to be promoted in each preset dimension includes: determining a first group of labels under the commodity content attribute dimension of the commodity to be promoted; the first group of labels includes a first content classification label, a first occupation label, and a first content person setup label; obtaining a second group of labels in user portrait data of the cargo influencer; the second group of labels includes a second content classification label, a second occupation label, and a second content person setup label; According to the first group of labels and the second group of labels, the matching degree of the cargo influencer and the commodity to be promoted in the commodity content attribute dimension is determined.
[0008] According to the method for recommending a cargo influencer provided by the application, the preset dimensions include the commodity commercial attribute dimension, and the determination of the matching degree of the cargo influencer and the commodity to be promoted in each preset dimension includes: determining a commodity commercial attribute sub-dimension under the commodity commercial attribute dimension of the commodity to be promoted; the commodity commercial attribute sub-dimension includes at least one of a commodity name, a commodity brand, a commodity category, and a commodity competitor; According to the cargo influencer data matched with each of the commodity commercial attribute sub-dimensions in the user portrait data of the cargo influencer, a first matching degree of the cargo influencer and the commodity to be promoted in each of the commodity commercial attribute sub-dimensions is determined; the cargo influencer data includes cargo influencer times and cargo influencer media data; According to all of the cargo influencer media data and the total media data in the user portrait data of the cargo influencer, a second matching degree of the cargo influencer and the commodity to be promoted is determined; According to the second matching degree and the first matching degree of the cargo influencer and the commodity to be promoted in each of the commodity commercial attribute sub-dimensions, a matching degree of the cargo influencer and the commodity to be promoted in the commodity commercial attribute dimension is determined.
[0009] According to the method for recommending a cargo influencer provided by the application, the preset dimensions include the commodity target audience dimension, and the determination of the matching degree of the cargo influencer and the commodity to be promoted in each preset dimension includes: determining a third group of labels under the commodity target audience dimension of the commodity to be promoted; the third group of labels includes a first gender audience label, a first age audience label, a first regional audience label, and a first language audience label; obtaining a fourth group of labels in the user portrait data of the influencer with goods; the fourth group of labels includes a second gender audience label, a second age audience label, a second regional audience label, and a second language audience label; According to the third group of labels and the fourth group of labels, the matching degree of the influencer with goods and the to-be-promoted commodity in the commodity target audience dimension is determined.
[0010] According to the influencer with goods recommendation method provided by the application, the commodity information of the to-be-promoted commodity is obtained by the following method: Obtaining a commodity link of the to-be-promoted commodity; According to the commodity link, the commodity page content corresponding to the to-be-promoted commodity is accessed; The commodity information of the to-be-promoted commodity is extracted from the commodity page content.
[0011] According to the influencer with goods recommendation method provided by the application, the at least one influencer with goods is obtained by the following method: Obtaining the delivery requirements of the to-be-promoted commodity; the delivery requirements include at least one of the delivery platform, the delivery area, the delivery language, the promotion budget, the delivery time and the commodity target audience dimension; According to the delivery requirements, at least one influencer with goods is selected from the influencer database.
[0012] According to the influencer with goods recommendation method provided by the application, after determining at least one target influencer with goods from at least one influencer with goods according to the matching degree of the influencer with goods and the to-be-promoted commodity in each preset dimension, the method further comprises: According to the matching degree of each target influencer with goods in each preset dimension, at least one target influencer with goods is sorted to obtain an influencer with goods recommendation list; According to the matching degree of each target influencer with goods in each preset dimension, a large language model is used to generate a recommendation reason for each target influencer with goods; The influencer with goods recommendation list and the recommendation reason of each target influencer with goods are sent to the merchant terminal of the to-be-promoted commodity.
[0013] The application also provides an influencer with goods recommendation device, which comprises: A first recommendation module is used to obtain the commodity information of the to-be-promoted commodity and the user portrait data of at least one influencer with goods; A second recommendation module is used to extract features from the commodity information of the to-be-promoted commodity, and obtain commodity feature data of each preset dimension of the to-be-promoted commodity; the preset dimension includes at least one of the commodity content attribute dimension, the commodity commercial attribute dimension and the commodity target audience dimension. The third recommendation module is used to determine the matching degree between the influencer and the product to be promoted in each preset dimension based on the user feature data that matches the product feature data in each preset dimension of the influencer's user profile data. The fourth recommendation module is used to determine at least one target influencer from at least one influencer based on the matching degree between the influencer and the product to be promoted in each preset dimension.
[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the product recommendation method described above.
[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the product recommendation method described above.
[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the product recommendation method described above.
[0017] The influencer recommendation method provided by this invention obtains multi-dimensional product feature data of the product to be promoted and combines it with the user profile data of the influencers. It calculates the matching degree in dimensions such as product content attributes, product commercial attributes and target audience. In this way, it objectively and accurately selects target influencers who are highly compatible with the product to be promoted from multiple influencers, thereby improving the accuracy and efficiency of influencer selection and reducing the subjectivity and inefficiency caused by manual reliance. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0019] Figure 1 This is one of the flowcharts illustrating the influencer recommendation method provided in this embodiment of the invention.
[0020] Figure 2 This is the second scenario illustration of the influencer recommendation method provided in this embodiment of the invention.
[0021] Figure 3 This is the third flowchart of the influencer recommendation method provided in this embodiment of the invention.
[0022] Figure 4 This is the fourth scenario illustration of the influencer recommendation method provided in this embodiment of the invention.
[0023] Figure 5 This is a schematic diagram of the product recommendation device provided in an embodiment of the present invention.
[0024] Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0026] With the booming development of the e-commerce industry, promoting products through influential influencers has become an important way for brands to expand their markets and enhance product awareness, and it is widely used in various business scenarios. Whether in fast-moving consumer goods, electronics, or home furnishings, merchants are increasingly relying on the traffic appeal and content dissemination power of influencers to achieve increased sales and expanded brand influence.
[0027] Currently, merchants primarily rely on manual methods to select and collaborate with influencers. Specifically, merchants typically assign dedicated teams or personnel to gather basic data on influencers from various e-commerce and social media platforms. This data often includes the influencer's total number of followers, follower growth trends, average views per video, likes, comments, shares, and other interactive data, as well as limited historical sales records provided by some platforms, such as previously promoted product categories and sales volume. Staff then need to compile and organize this scattered data, combining it with their accumulated market experience, understanding of product characteristics, and judgment of past collaborations to evaluate and select influencers one by one.
[0028] While some platforms offer basic tools to simplify the process, such as tagging influencers across broad categories like beauty, apparel, and digital products, or allowing merchants to rank influencers by single metrics like follower count or engagement, these tools are extremely limited. Tagging is too coarse-grained to accurately differentiate influencers within the same field with different styles and audience characteristics; and ranking by a single metric ignores the deeper connection between influencers and products. For example, a food influencer with a huge following might not be suitable for promoting a specific type of kitchenware, as their followers might be more interested in food preparation than the utensils themselves.
[0029] This manual screening method has significant drawbacks: Firstly, it requires substantial investment of manpower and time, especially when merchants need to select suitable candidates from thousands of influencers within a short period. Manual processing is extremely inefficient and struggles to meet rapidly changing market demands. Secondly, the screening results heavily rely on the personal experience and subjective judgment of staff, lacking objective and systematic evaluation standards. When faced with a wide variety of product promotion needs and a vast pool of influencers with diverse fan bases, content styles, and sales capabilities, matching discrepancies are easily created.
[0030] Based on this, embodiments of the present invention provide a method for recommending influencers, which objectively and accurately selects target influencers from multiple influencers based on dimensions such as product content attributes, product commercial attributes, and target audience, thereby improving the accuracy and efficiency of influencer selection and reducing the subjectivity and inefficiency caused by manual reliance.
[0031] Figure 1 This is one of the flowcharts illustrating the influencer recommendation method provided by this invention, such as... Figure 1 As shown, the method includes the following steps 110, 120, 130 and 140.
[0032] Step 110: Obtain product information of the product to be promoted and user profile data of at least one influencer.
[0033] Here, "products to be promoted" includes any physical or virtual goods that merchants wish to market and promote through influencers. "Product information" refers to the various key data and attribute information related to the product that needs to be collected and organized in advance to ensure that the product can more accurately reach the target audience and improve promotional effectiveness in subsequent promotions (such as through influencers, e-commerce platform marketing, etc.).
[0034] Typically, product information includes, but is not limited to, the following key types of information: Basic product attribute information: such as product name, category, specifications, core ingredients or materials, functional selling points, pricing, brand background, etc.; Product market performance data: such as historical sales data, traffic data (number of times the product is exposed, number of users who click on it), conversion data (product click-through rate, transaction conversion rate), user evaluation data (positive review rate, negative review content and main feedback issues), return and exchange data (quality return rate, return rate due to logistics or service), etc. Product target audience dimension related information: This refers to data related to the characteristics of potential consumers who are suitable for the product, such as the gender, age distribution, city distribution, spending power, and purchasing preferences of users who have previously purchased the product.
[0035] Here, "product influencer" refers to a user who attracts followers by posting content on social media or live streaming platforms and possesses the ability to promote products. The user profile data of product influencers refers to a comprehensive data set constructed by collecting and analyzing various key information about product influencers and their followers, clearly reflecting the influencer's own capabilities, the attributes of their fan base, and their consumer behavior preferences.
[0036] Typically, user profile data includes, but is not limited to, the following key information: Basic and capability data of influencers: such as the influencer's basic identity information, core sales data (such as sales volume, total number of products promoted, number of partner stores, average sales per session, average order value), average views / reads of videos or articles, interaction rate (likes, comments, shares), cooperation price, etc. Basic attribute data of influencer fans: such as the gender distribution, age stratification, city distribution, geographical distribution, education or occupation characteristics of fans; Consumer behavior and preference data of influencers' fans: such as average purchase price, preferred product categories, purchase decision factors, purchase time habits, repurchase rate, etc.
[0037] In this embodiment, product information of the product to be promoted and user profile data of at least one influencer can be collected from data sources such as platforms, databases, or partners. The product information of the product to be promoted is used to subsequently match suitable influencers and generate promotional content; while the user profile data of the influencers is used to assess whether these influencers and their fan base are highly compatible with the target consumers of the product.
[0038] Step 120: Extract features from the product information of the product to be promoted to obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension.
[0039] It should be noted that this embodiment conducts in-depth analysis of the products to be promoted from multiple preset dimensions, including but not limited to: Product content attribute dimension: This focuses on the product's own content characteristics, such as title, description, images, videos, features, and usage scenarios. The commercial attributes dimension of a product describes its positioning in the market, such as the product's brand, detailed category, price range, brand tone, and its main competing brands or products. Product target audience dimension: Characteristics of the main consumer groups targeted by the product, such as age, gender, region, interests and preferences, spending power, purchasing decision factors, and related consumption habits.
[0040] In this embodiment, all product information related to the product to be promoted is first summarized, which mainly includes two parts: one is the product's inherent information, such as the name, specifications, ingredients, function description, brand background, etc. on the product details page; the other is market feedback information, such as sales volume, conversion rate, negative review data, and past purchase user profile reports on the e-commerce platform backend.
[0041] Next, key features are extracted according to one or more predefined dimensions (such as content attributes, commercial attributes, target audience, etc.). The extracted information is transformed into product feature data that can be used for analysis and matching. This data can be stored in a structured format for subsequent data analysis.
[0042] In one example, natural language processing analysis can be performed on information such as product details, user reviews, and marketing copy to extract keywords or tags that represent its content attributes, forming product feature data in the dimension of product content attributes.
[0043] In one example, product feature data can be obtained directly from product information, or by combining it with market database analysis to derive the product's commercial attribute dimension.
[0044] In one example, product characteristic data for the target audience dimension can be determined by analyzing the product's market positioning, user profiles in user reviews, and customer data provided by the merchant.
[0045] Here, the preset dimensions can be flexibly set according to actual promotion needs, and there are no restrictions on them.
[0046] Step 130: Based on the user feature data in the user profile data of the influencer that matches the product feature data in each preset dimension, determine the matching degree between the influencer and the product to be promoted in each preset dimension.
[0047] In this embodiment, after obtaining structured product feature data, the product to be promoted is compared with each influencer in each dimension, and the corresponding matching degree is calculated. Here, the matching degree is an indicator used to measure the degree of fit between the influencer and the product to be promoted in a specific dimension. It can be a score (e.g., from 0 to 100) or a level (e.g., high, medium, low).
[0048] For each preset dimension, user characteristic data corresponding to that dimension is found from the user profile data of the sales influencer, and then compared with the characteristic data of the product.
[0049] For example, regarding product content attributes, the product feature data in this dimension will be compared with user feature data such as historical content themes and influencer tags in the influencer's user profile data. Regarding product commercial attributes, the product feature data in this dimension will be compared with user feature data such as historical partner brands and product categories in the influencer's user profile data. Regarding the product target audience, the product feature data in this dimension will be compared with user feature data such as the influencer's fan base in the influencer's user profile data.
[0050] In this embodiment, the matching degree can be calculated using various algorithms, such as similarity calculation based on label overlap, cosine similarity calculation based on vector space model, or semantic similarity judgment based on pre-trained model, etc., and there are no restrictions on this.
[0051] Step 140: Based on the matching degree between the influencer and the product to be promoted in each preset dimension, determine at least one target influencer from at least one influencer.
[0052] After determining the matching degree between each influencer and the product to be promoted in each preset dimension, the matching degree of all preset dimensions is combined to determine the final recommended influencer.
[0053] The methods for identifying target influencers can be flexible. For example, one or more of the following strategies can be used: Weighted summation method: Assign a weight to the matching degree of each preset dimension (for example, if a merchant thinks the target audience dimension is the most important, then give it a weight of 0.5, while the content attribute dimension and the commercial attribute dimension each account for 0.25), then calculate the total weighted matching degree of each influencer, and select the influencer with the highest total weighted matching degree as the target influencer, or select one or more influencers with a total weighted matching degree higher than the threshold as the target influencer; Threshold filtering method: Set a threshold for the matching degree of each preset dimension, and select the sales influencers who exceed the threshold in all preset dimensions as target sales influencers.
[0054] The influencer recommendation method provided in this invention obtains multi-dimensional product feature data of the product to be promoted and combines it with the user profile data of the influencers. It calculates the matching degree in dimensions such as product content attributes, product commercial attributes, and target audience. In this way, it objectively and accurately selects target influencers who are highly compatible with the product to be promoted from among multiple influencers, thereby improving the accuracy and efficiency of influencer selection and reducing the subjectivity and inefficiency caused by manual reliance.
[0055] In some embodiments, reference Figure 2 , Figure 2 This is a second flowchart illustrating the influencer recommendation method provided in this embodiment of the invention, as shown below. Figure 2 As shown, the method includes the following steps 210, 220 and 230.
[0056] Step 210: Obtain the product link of the product to be promoted.
[0057] Here, the product link includes a Uniform Resource Locator (URL) or a short link containing redirection logic. It points to the product's details page on a specific e-commerce platform, the brand's official website, or any other online medium.
[0058] In this embodiment, product links can be manually entered or pasted by the merchant through the interface, or they can be automatically obtained in batches from the merchant's product management system (such as an ERP system) through an API interface, without any restrictions.
[0059] Step 220: Access the product page content corresponding to the product to be promoted based on the product link.
[0060] After obtaining the product link, simulate a browser or use an HTTP client to determine the IP address of the target server specified by the product link. For example, if the product link is a Uniform Resource Locator (URL), then parse the IP address in the URL. If the link is a shortened link, then first send a request to the shortened link service server to obtain its corresponding URL, and then perform subsequent IP address resolution.
[0061] After resolving the IP address, the client will send an access request to the target server according to the communication protocol (such as HTTPS) in the product link. Upon receiving the access request, the target server (such as the product server of an e-commerce platform) verifies the validity of the request. After successful verification, the target server will retrieve the relevant data file of the product from its own storage, and at the same time pull the dynamic data of the product (such as name, price, inventory, etc.) from the database, embed the dynamic data into the data file, and finally transmit the relevant data back to the client via the network.
[0062] After the client receives the resources returned by the server, it renders the page. Once all resources have been loaded and rendered, the client will see a complete product page, including the product name, price, detailed description, images, and videos. Users can also perform actions such as clicking and swiping on this page.
[0063] Step 230: Extract the product information of the product to be promoted from the product page content.
[0064] After successfully obtaining the complete content of the product page, a web scraping program is launched. For example, XPath paths or CSS selectors are used to locate and extract structured information such as the product's title, price, brand, and specifications. All text content in the product details description area can also be extracted as corpus for subsequent content attribute analysis. The URLs of the product's main image, details images, and videos can also be scraped for subsequent multimodal analysis. In addition, the content of the user comment section can be parsed to analyze the characteristics of the product's target audience.
[0065] The product recommendation method provided in this invention improves the efficiency and accuracy of data entry by automatically retrieving product information based on product links.
[0066] In some embodiments, reference Figure 3 , Figure 3 This is the third flowchart illustrating the influencer recommendation method provided in this embodiment of the invention, as shown below. Figure 3 As shown, the method includes the following steps 310 and 320.
[0067] Step 310: Obtain the placement requirements for the product to be promoted; the placement requirements include at least one of the following dimensions: placement platform, placement area, placement language, promotion budget, placement time, and target audience of the product.
[0068] Here, "admission requirements" refers to the selection criteria for influencers that merchants pre-set based on their marketing strategies and business goals, including but not limited to: Platforms for promotion: Specify which one or more social media or e-commerce platforms you wish to promote on; Targeting Area: Specify the regions where the main fan base of the influencer is located to match the target market of the product; Targeting Language: The primary language used by the designated influencers in creating their content; Promotion Budget: Set the budget range for this collaboration to select influencers whose bids fall within this range; Launch timing: Set the desired start and end times for the promotional campaign; Product target audience dimension: The merchant's initial profile of the target customer group, such as age group, gender, etc.
[0069] Step 320: Based on the aforementioned placement requirements, select at least one influencer from the influencer database to drive sales.
[0070] Here, the influencer database is a pre-built database that stores a massive amount of user profile data for various influencers.
[0071] In this embodiment, the above-mentioned placement requirements are used as query conditions to filter the influencer database. For example, a query command could be: "From the influencer database, filter influencers whose placement platform is A, placement region is B, promotion budget is C, and the proportion of female followers exceeds D."
[0072] The influencer recommendation method provided in this invention pre-screens a large number of influencers based on the merchant's advertising requirements, narrowing down the range of influencers that need to be analyzed in depth later, and improving the efficiency and accuracy of the entire recommendation process.
[0073] In some embodiments, the preset dimension includes the product content attribute dimension, and determining the matching degree between the influencer and the product to be promoted on each preset dimension includes: Determine the first set of tags under the product content attribute dimension of the product to be promoted; the first set of tags includes a first content category tag, a first occupation tag, and a first content persona tag; Obtain the second set of tags from the user profile data of the influencer; the second set of tags includes a second content category tag, a second occupation tag, and a second content persona tag. Based on the first set of tags and the second set of tags, determine the matching degree between the influencer and the product to be promoted in the product content attribute dimension.
[0074] In this embodiment, the product feature data of the product to be promoted is further analyzed in the dimension of product content attributes. In this dimension, the product feature data of the product to be promoted is described in detail and in a structured manner.
[0075] Specifically, by analyzing the product feature data of the products to be promoted in terms of product content attributes, the following types of tags are extracted: First content category tag: refers to the category tag of the content field to which the product belongs, such as technology, aerial photography, outdoor sports, etc. First occupational label: refers to the occupational identity of the user of this product. For example, the occupational label of a recording microphone may be podcaster, singer, voice actor, etc. The first content persona tag refers to the image or style that the content creator is most suitable to present when promoting the product; for example, a bag can be tagged as a fashionista, exquisite, and elegant; a funny board game can be tagged as humorous, playful, and silly.
[0076] Furthermore, in addition to providing a detailed and structured description of the promoted products under this dimension, the user profile data of the sales influencers are also described in the same detailed and structured way under this dimension.
[0077] Specifically, we analyzed the user characteristic data corresponding to the product characteristic data in the product content attribute dimension of the user profile data of live-streaming influencers, and extracted the following types of tags: The second content category tag indicates the field or theme of content that the influencer usually creates and publishes, such as technology, aerial photography, outdoor sports, etc. Secondary professional label: The identity and role that experts present to the outside world in their content creation, such as tech blogger, programmer, product manager, etc. The second content persona tag refers to the persona presented by the content creator; for example, professional, hardcore, rigorous in evaluation, and humorous.
[0078] In one example, a pre-built tag dictionary can be used to determine the corresponding content category tags, occupation tags, and content persona tags. Specifically, a keyword dictionary is created for each category of tags. For example, the keyword dictionary for occupation tags is: {Doctor: [medical, surgical gown, stethoscope]; Programmer: [code, keyboard, development, front-end, back-end]; Teacher: [chalk, blackboard, courseware]}. The keyword dictionary for content persona tags is: {refined lifestyle: [style, afternoon tea]; cost-effectiveness: [affordable, cheap, good value]}.
[0079] The explanation is based on the product feature data under the product content attribute dimension of the product to be promoted. When classifying the product into various tags, the product feature data under the product content attribute dimension can be scanned. If the scan finds keywords that are the same as or semantically similar to those in the pre-built tag dictionary, the corresponding tag is assigned to them.
[0080] In one example, machine learning and natural language processing techniques can also be used to determine corresponding content category tags, occupation tags, and content persona tags. For instance, a pre-annotated dataset containing a large amount of feature data and its correct labels can be used as a training set to train an initial large language model. The large language model will learn the mapping relationship from feature data to correct labels. In practical applications, prompt words corresponding to each type of label can be constructed. The corresponding feature data and prompt words can be input into the trained large language model to obtain the label prediction results output by the large language model.
[0081] After identifying the various tags corresponding to the products to be promoted and the influencers, the matching degree between the influencers and the products to be promoted in terms of product content attributes is determined based on the matching degree between the first set of tags and the second set of tags.
[0082] In one example, the overlap rate of the two sets of tags in the three sub-dimensions of content category, occupation, and content persona can be calculated separately, and the final content attribute matching degree can be obtained based on the overlap rate of the tags in the three sub-dimensions.
[0083] In one example, a word vector model can also be used to calculate the semantic similarity of two sets of tags across three sub-dimensions: content category, occupation, and content persona. The final content attribute matching degree can then be obtained based on the semantic similarity across these three sub-dimensions.
[0084] Here, we can also obtain the matching degree between the influencer and the product to be promoted in terms of product content attributes by weighted averaging across all sub-dimensions, which will not be elaborated on here.
[0085] The influencer recommendation method provided in this invention concretizes and refines the matching process of product content attributes by introducing three sub-dimensions: content category, profession, and content persona. This improves the accuracy and comprehensiveness of the matching degree between influencers and products to be promoted in terms of product content attributes.
[0086] In some embodiments, a product business attribute sub-dimension is determined under the product business attribute dimension of the product to be promoted; the product business attribute sub-dimension includes at least one of product name, product brand, product category and product competitors; Based on the influencer data matching each of the product's commercial attribute sub-dimensions in the influencer's user profile data, the first matching degree between the influencer and the product to be promoted on each of the product's commercial attribute sub-dimensions is determined; the influencer data includes the number of times the influencer has promoted products and the media data of the influencer's promotion. Based on all the media data of the influencers promoting products, and the total media data in the user profile data of the influencers, a second matching degree between the influencers and the products to be promoted is determined; Based on the second matching degree and the first matching degree between the influencer and the product to be promoted in each of the product's commercial attribute sub-dimensions, the matching degree between the influencer and the product to be promoted in the product's commercial attribute dimension is determined.
[0087] In this embodiment, the product feature data of the product to be promoted is further analyzed in the dimension of product commercial attributes. Under this dimension, the product feature data of the product to be promoted is described in detail and in a structured manner.
[0088] Specifically, the product characteristic data of the products to be promoted are analyzed in terms of their commercial attributes, and the commercial attributes dimension is further divided into the following sub-dimensions: Product Name: The specific name of the product; Product Brand: The brand to which the product belongs; Product category: The specific category to which a product belongs, such as smartphones belonging to the mobile phone and digital product category; Product competitors: The main competitors of this product in the market. For example, the competitor of the smartphone corresponding to brand A is a smartphone from another brand B with similar functions.
[0089] Furthermore, after further subdividing the commercial attribute dimension of the product to be promoted into multiple commercial attribute sub-dimensions, the user profile data of the influencer is retrieved, and influencer data related to the above sub-dimensions is found in their historical promotion data, including but not limited to: influencer promotion frequency, influencer promotion media data (such as influencer promotion media browsing data and influencer promotion media interaction data).
[0090] For smartphone A to be promoted, the system will collect the following types of data from a certain influencer's user profile: The number of times the A-series smartphone products have been promoted in the past, media views, and interaction volume; The number of times smartphone A's brand has promoted other products of the same brand in the past, media views, and engagement; The number of times smartphones have been promoted in the past, media views, and interaction volume; The number of times it has promoted its competitors (smartphones from other brands B with similar functions to smartphone A), media views, and engagement.
[0091] In one example, taking the product name as a representative sub-dimension of the product's commercial attribute, the calculation method for its corresponding first match degree is as follows: ; in, This indicates the number of times the corresponding product name has been promoted. This represents the average number of page views for the corresponding product name across various media platforms. This represents the average number of page views per product across all media platforms over the past year. This indicates the average interaction rate of the corresponding product across various media platforms. This represents the average interaction rate of all products across various media platforms over the past year.
[0092] Here, the calculation method for the first matching degree of each product brand, product category, and product competitor's commercial attribute sub-dimension is the same as the calculation method of the first matching degree above, and will not be repeated here.
[0093] In addition, in this embodiment, besides calculating the first matching degree between the influencer and the product to be promoted on each product's commercial attribute sub-dimension, a second matching degree will also be calculated based on the influencer's historical order performance data.
[0094] In one example, the formula for calculating the second matching degree is as follows: ; in, This represents the average number of page views per product across all media platforms over the past year. This represents the average media engagement rate for all products sold over the past year. This represents the average number of views across all media platforms used by this influencer over the past year (including both non-commercial and commercial media views). This refers to the average interaction rate of commercial media, which represents the total media data (including non-commercial media interaction rate and commercial media interaction rate) of this influencer over the past year.
[0095] Finally, based on the first and second matching scores, the matching degree between the influencer and the product to be promoted in terms of the product's commercial attributes is determined. For example, a weighted average of the first and second matching scores can be used to determine the matching degree between the influencer and the product in terms of the product's commercial attributes.
[0096] The influencer recommendation method provided in this invention concretizes and refines the matching process of product commercial attributes. By introducing five sub-dimensions—product name, product brand, product category, and product competitors—it improves the accuracy and comprehensiveness of the matching degree between influencers and products to be promoted in terms of product commercial attributes.
[0097] In some embodiments, the preset dimension includes the target audience dimension of the product, and determining the matching degree between the influencer and the product to be promoted in each preset dimension includes: Determine a third set of tags under the target audience dimension of the product to be promoted; the third set of tags includes a first gender audience tag, a first age audience tag, a first geographic audience tag, and a first language audience tag; Obtain the fourth set of tags from the user profile data of the influencer; the fourth set of tags includes a second gender audience tag, a second age audience tag, a second regional audience tag, and a second language audience tag. Based on the third set of tags and the fourth set of tags, determine the matching degree between the influencer and the product to be promoted in the dimension of the product's target audience.
[0098] In this embodiment, the product feature data of the product to be promoted is further analyzed in terms of the target audience dimension. In this dimension, the product feature data of the product to be promoted is described in detail and in a structured manner.
[0099] Specifically, by analyzing the product characteristic data of the promoted product in terms of the target audience, the following types of tags are extracted: First gender audience tag: refers to the gender group to whom this product is suitable; First age group audience tag: refers to the age group of people who use this product; First geographic audience tag: refers to the geographic range in which this product can be used; First Language Audience Tag: Refers to the language group to which this product is suitable.
[0100] Furthermore, in addition to providing a detailed and structured description of the promoted products under this dimension, the user profile data of the sales influencers are also described in the same detailed and structured way under this dimension.
[0101] Specifically, we analyzed the user characteristic data corresponding to the product characteristic data in the user profile data of influencers and the target audience dimension of the products, and extracted the following types of tags: First gender audience tag: refers to the gender group of the influencer's fans; First age audience tag: refers to the age group of the influencer's fans; First geographic audience tag: refers to the geographic range of the influencer's fans; First Language Audience Tag: Refers to the language group of the influencer's fans.
[0102] In one example, a pre-built tag dictionary can be used to determine the corresponding gender audience tags, age audience tags, geographic audience tags, and language audience tags. Specifically, a keyword dictionary is created for each type of tag. For example, the keyword dictionary for the gender audience tag is: {Male: [razor, gaming keyboard, tie, shaving cream...]; Female: [lipstick, dress...]}; the keyword dictionary for the age audience tag is: {Infant: [diapers, baby food, crib...]; Children: [children's watches, school uniforms...] etc.}.
[0103] The explanation is based on the product feature data under the product content attribute dimension of the product to be promoted. When classifying the product into various tags, the product feature data under the product content attribute dimension can be scanned. If the scan finds keywords that are the same as or semantically similar to those in the pre-built tag dictionary, the corresponding tag is assigned to them.
[0104] In one example, machine learning and natural language processing techniques can also be used to determine corresponding gender, age, geographic, and language audience labels. For instance, a pre-annotated dataset containing a large amount of feature data and its correct labels can be used as a training set to train an initial large language model. The large language model will learn the mapping relationship from feature data to correct labels. In practical applications, prompt words corresponding to each type of label can be constructed. The corresponding feature data and prompt words can be input into the trained large language model to obtain the label prediction results output by the large language model.
[0105] After identifying the various tags corresponding to the products to be promoted and the influencers, the matching degree between the influencers and the products to be promoted in terms of product content attributes is determined based on the matching degree between the third group of tags and the fourth group of tags.
[0106] In one example, the overlap rate of the two sets of tags can be calculated separately in five sub-dimensions: gender audience, age audience, geographic audience, and language audience. The final content attribute matching degree can be obtained based on the overlap rate of the tags in the five sub-dimensions.
[0107] In one example, word vector models can also be used to calculate the semantic similarity of two sets of tags across five sub-dimensions: gender audience, age audience, geographic audience, and language audience. The final matching score is then obtained based on the semantic similarity across these five sub-dimensions.
[0108] Here, we can also obtain the matching degree between the influencer and the product to be promoted in the dimension of the target audience by weighted averaging the matching degree of all sub-dimensions, which will not be elaborated on here.
[0109] The influencer recommendation method provided in this invention specifies and refines the matching process of product target audience dimensions. By introducing five sub-dimensions—gender audience, age audience, regional audience, and language audience—it improves the accuracy and comprehensiveness of the matching degree between influencers and products to be promoted in terms of product target audience dimensions.
[0110] In some embodiments, reference Figure 4 , Figure 4 This is the fourth flowchart illustrating the influencer recommendation method provided in this embodiment of the invention, as follows: Figure 4 As shown, the method includes the following steps 410, 420 and 430.
[0111] Step 410: Based on the matching degree of each target influencer in each preset dimension, sort at least one target influencer to obtain a recommended list of influencers.
[0112] In this step, after identifying several target influencers, these influencers are further ranked to facilitate merchants' decision-making.
[0113] In one example, the total score of matching across all preset dimensions can be combined. For instance, the matching scores of content attributes, commercial attributes, and target audience can be weighted and summed according to certain weights to obtain the final comprehensive score for each target influencer. This final comprehensive score can then be used as a benchmark to sort the influencers in descending order, forming a recommended list of influencers.
[0114] In one example, the influencers can be sorted in descending order according to their matching degree across various preset dimensions, generating influencer recommendation lists for three dimensions: content attributes, commercial attributes, and target audience.
[0115] Step 420: Based on the matching degree of each target influencer in each preset dimension, use a large language model to generate the recommendation reason for each target influencer.
[0116] Furthermore, in this embodiment, a structured prompt message will be automatically constructed for each recommended target influencer. This prompt message includes key information about the product, key information about the target influencer, and the matching degree of the target influencer in various preset dimensions.
[0117] For example, the prompt message is: "Please generate a recommendation reason for influencer A for the promoted product 'Smart Fitness Mirror'. The matching data is as follows: Product content attribute matching score 95 points (sub-dimensions: fitness tutorials, professional persona), Product commercial attribute matching score 80 points (sub-dimensions: multiple times promoting smart hardware, high conversion rate), Product target audience matching score 92 points (sub-dimensions: fans are concentrated in the 25-35 age group, focusing on healthy living). Please summarize its core advantages in concise and professional marketing language." Then, the prompt message is input into a pre-trained large language model. The large language model will understand the data and generate a natural, fluent, and persuasive recommendation reason, such as "Influencer A is the ideal candidate to promote the Smart Fitness Mirror. Her content is highly compatible with the product (95 points) and can professionally demonstrate the product's functions; her fan profile accurately covers the target consumer group (92 points); at the same time, her rich experience in promoting smart hardware also indicates good commercial conversion prospects (80 points)."
[0118] Step 430: Send the list of recommended influencers and the reasons for recommending each of the target influencers to the merchant terminal of the product to be promoted.
[0119] Finally, the sorted list of top-selling influencers and the reasons for recommending each target influencer are integrated into a comprehensive recommendation report. This report can be presented in the form of web pages, documents, icons, etc., and sent to the merchant's user terminal devices for review and decision-making.
[0120] The product recommendation method provided in this invention improves the human-computer interaction experience by transforming the analysis results into a product recommendation list and automatically generating recommendation reasons by calling a large language model.
[0121] Based on any of the above embodiments, the present invention also provides a live-streaming influencer recommendation device, for reference. Figure 5 The device includes: a first recommendation module 510, a second control module 520, a third recommendation module 530, and a fourth control module 540.
[0122] The first recommendation module 510 is used to obtain product information of the products to be promoted and user profile data of at least one sales influencer; The second recommendation module 520 is used to extract features from the product information of the product to be promoted and obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension. The third recommendation module 530 is used to determine the matching degree between the influencer and the product to be promoted in each preset dimension based on the user feature data that matches the product feature data in each preset dimension of the influencer's user profile data. The fourth recommendation module 540 is used to determine at least one target influencer from at least one influencer based on the matching degree between the influencer and the product to be promoted in each preset dimension.
[0123] The influencer recommendation device provided by this invention acquires multi-dimensional product feature data of the product to be promoted and combines it with the user profile data of the influencers. It calculates the matching degree in dimensions such as product content attributes, product commercial attributes, and target audience. In this way, it objectively and accurately selects target influencers who are highly compatible with the product to be promoted from among multiple influencers, thereby improving the accuracy and efficiency of influencer selection and reducing the subjectivity and inefficiency caused by manual reliance.
[0124] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a method for recommending products by livestreaming influencers, the method including: Obtain product information for the products to be promoted and user profile data of at least one influencer. Feature extraction is performed on the product information of the product to be promoted to obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension; Based on the user feature data in the user profile data of the sales influencer that matches the product feature data in each preset dimension, the matching degree between the sales influencer and the product to be promoted in each preset dimension is determined; Based on the matching degree between the influencer and the product to be promoted in each preset dimension, at least one target influencer is determined from at least one influencer.
[0125] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0126] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to execute the influencer recommendation method provided by the above methods, the method comprising: Obtain product information for the products to be promoted and user profile data of at least one influencer. Feature extraction is performed on the product information of the product to be promoted to obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension; Based on the user feature data in the user profile data of the sales influencer that matches the product feature data in each preset dimension, the matching degree between the sales influencer and the product to be promoted in each preset dimension is determined; Based on the matching degree between the influencer and the product to be promoted in each preset dimension, at least one target influencer is determined from at least one influencer.
[0127] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the influencer recommendation method provided by the above methods, the method comprising: Obtain product information for the products to be promoted and user profile data of at least one influencer. Feature extraction is performed on the product information of the product to be promoted to obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension; Based on the user feature data in the user profile data of the sales influencer that matches the product feature data in each preset dimension, the matching degree between the sales influencer and the product to be promoted in each preset dimension is determined; Based on the matching degree between the influencer and the product to be promoted in each preset dimension, at least one target influencer is determined from at least one influencer.
[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0129] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for recommending products by live-streaming influencers, characterized in that, The method includes: Obtain product information for the products to be promoted and user profile data of at least one influencer. Feature extraction is performed on the product information of the product to be promoted to obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension; Based on the user feature data in the user profile data of the sales influencer that matches the product feature data in each preset dimension, the matching degree between the sales influencer and the product to be promoted in each preset dimension is determined; Based on the matching degree between the influencer and the product to be promoted in each preset dimension, at least one target influencer is determined from at least one influencer.
2. The method for recommending influencers based on live-streaming sales according to claim 1, characterized in that, The preset dimensions include the product content attribute dimensions, and determining the matching degree between the influencer and the product to be promoted in each preset dimension includes: Determine the first set of tags under the product content attribute dimension of the product to be promoted; the first set of tags includes a first content category tag, a first occupation tag, and a first content persona tag; Obtain the second set of tags from the user profile data of the influencer; the second set of tags includes a second content category tag, a second occupation tag, and a second content persona tag. Based on the first set of tags and the second set of tags, determine the matching degree between the influencer and the product to be promoted in the product content attribute dimension.
3. The method for recommending influencers based on live-streaming sales data according to claim 1, characterized in that, The preset dimensions include the product's commercial attribute dimensions, and determining the match between the influencer and the product to be promoted on each preset dimension includes: Determine the product business attribute sub-dimensions under the product business attribute dimension of the product to be promoted; the product business attribute sub-dimensions include at least one of product name, product brand, product category and product competitors; Based on the influencer's sales data matching each of the product's commercial attribute sub-dimensions in the influencer's user profile data, the first matching degree between the influencer and the product to be promoted on each of the product's commercial attribute sub-dimensions is determined; the influencer's sales data includes the number of times the influencer has promoted the product and the media data on which the influencer promoted the product. Based on all the media data of the influencers promoting products, and the total media data in the user profile data of the influencers, a second matching degree between the influencers and the products to be promoted is determined; Based on the second matching degree and the first matching degree between the influencer and the product to be promoted in each of the product's commercial attribute sub-dimensions, the matching degree between the influencer and the product to be promoted in the product's commercial attribute dimension is determined.
4. The method for recommending live-streaming sales influencers according to claim 1, characterized in that, The preset dimensions include the target audience dimension of the product, and determining the matching degree between the influencer and the product to be promoted in each preset dimension includes: Determine a third set of tags under the target audience dimension of the product to be promoted; the third set of tags includes a first gender audience tag, a first age audience tag, a first geographic audience tag, and a first language audience tag; Obtain the fourth set of tags from the user profile data of the influencer; the fourth set of tags includes a second gender audience tag, a second age audience tag, a second regional audience tag, and a second language audience tag. Based on the third set of tags and the fourth set of tags, determine the matching degree between the influencer and the product to be promoted in the dimension of the product's target audience.
5. The method for recommending influencers based on claims 1, characterized in that, The product information of the product to be promoted was obtained through the following methods: Obtain the product link of the product to be promoted; Access the product page content corresponding to the product to be promoted based on the product link; Extract the product information of the product to be promoted from the product page content.
6. The method for recommending influencers based on claims 1, characterized in that, The at least one influencer was obtained through the following methods: Obtain the advertising requirements for the product to be promoted; the advertising requirements include at least one of the following dimensions: advertising platform, advertising region, advertising language, advertising budget, advertising time, and target audience of the product; Based on the aforementioned placement requirements, at least one influencer will be selected from the influencer database to drive sales.
7. The method for recommending influencers based on live-streaming sales according to claim 1, characterized in that, After determining at least one target influencer from at least one influencer based on the matching degree between the influencer and the product to be promoted in each preset dimension, the method further includes: Based on the matching degree of each target influencer in each preset dimension, at least one target influencer is sorted to obtain a recommended list of influencers; Based on the matching degree of each target influencer in each preset dimension, a recommendation reason for each target influencer is generated using a large language model; The list of recommended influencers and the reasons for recommending each of the target influencers are sent to the merchant's terminal for the product to be promoted.
8. A product recommendation device by livestreaming influencers, characterized in that, include: The first recommendation module is used to obtain product information of the products to be promoted and user profile data of at least one influencer. The second recommendation module is used to extract features from the product information of the product to be promoted, and obtain product feature data for each preset dimension of the product to be promoted; the preset dimension includes at least one of the product content attribute dimension, product commercial attribute dimension and product target audience dimension; The third recommendation module is used to determine the matching degree between the influencer and the product to be promoted in each preset dimension based on the user feature data that matches the product feature data in each preset dimension of the influencer's user profile data. The fourth recommendation module is used to determine at least one target influencer from at least one influencer based on the matching degree between the influencer and the product to be promoted in each preset dimension.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the influencer recommendation method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the influencer recommendation method as described in any one of claims 1 to 7.