Product recommendation method, device, non-volatile storage medium and electronic device

Abstract

The present invention discloses a product recommendation method, device, non-volatile storage medium and electronic device. The method includes: obtaining keywords for browsing products corresponding to a target account, and the categories to which the browsed products belong; inputting the keywords and the categories to which the browsed products belong into a target large language model, and the target large language model outputting a first product set and a second product set, wherein the first product set includes product identifiers that match the keywords, and the second product set includes product identifiers within the categories that match the categories to which they belong, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample; obtaining a union of the first product set and the second product set to obtain a third product set; sorting the product identifiers included in the third product set, and selecting multiple products recommended to the target account in order. The present invention solves the technical problem in the prior art that products cannot be accurately and efficiently recommended to users.

Description

Technical Field

[0001] The present invention relates to the field of product recommendation, and in particular to a product recommendation method, device, non-volatile storage medium and electronic device. Background Art

[0002] Existing product recommendation technologies are often limited to platform user behavior data and product information, lacking the ability to leverage external knowledge such as publicly available user reviews and social media discussions. This results in a limited knowledge boundary for recommendation systems, making it difficult to uncover more potential pairing relationships. Furthermore, manually configured pairing rules rely heavily on the experience of business personnel, requiring significant time and labor costs. Manually configured rules also struggle to capture the complex and subtle logic of product pairings, especially on e-commerce platforms with a vast number of products and a wide variety of categories. Furthermore, in the face of rapidly changing market trends and user preferences, the updating and adjustment of product pairing rules can be delayed, making it difficult to recommend appropriate products to users.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] The embodiments of the present invention provide a product recommendation method, device, non-volatile storage medium and electronic device to at least solve the technical problem in the prior art that products cannot be recommended to users accurately and efficiently.

[0005] According to one aspect of an embodiment of the present invention, a product recommendation method is provided, comprising: obtaining keywords for browsing products corresponding to a target account, and categories to which the browsed products belong; inputting the keywords and categories to which the browsed products belong into a target large language model, and outputting a first product set and a second product set from the target large language model, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within categories matching the categories to which they belong, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords of the sample product and the category to which the sample belongs; obtaining a union of the first product set and the second product set to obtain a third product set; sorting the product identifiers included in the third product set, and selecting, in order, a plurality of products recommended to the target account.

[0006] Optionally, the target large language model is obtained by: obtaining the product name of the sample product; determining the sample keyword corresponding to the sample product and the sample category corresponding to the sample product based on the product name of the sample product; generating a first training sample based on the sample keyword and the sample category; and fine-tuning the pre-trained large language model based on the first training sample to obtain the target large language model.

[0007] Optionally, based on the product name of the sample product, determine the sample keywords corresponding to the sample product and the category to which the sample corresponding to the sample product belongs, including: segmenting the product name of the sample product to obtain the sample keywords corresponding to the sample product; matching the sample keywords with a predetermined category database to determine the category to which the sample corresponding to the sample product belongs.

[0008] Optionally, the product identifiers included in the third product set are sorted, and multiple products recommended to the target account are selected in sequence, including: inputting the third product set into the target deep learning model, and the target deep learning model outputs multiple products recommended to the target account, wherein the target deep learning model is obtained by training the original deep learning model through the second training sample, and the second training sample includes historically recommended product identifiers, and historically browsed product identifiers that have been browsed in the historically recommended product identifiers.

[0009] Optionally, the third product set is input into the target deep learning model, and the target deep learning model outputs multiple products recommended to the target account, including: inputting the third product set into the target deep learning model, and predicting the browsing probability of each product included in the third product set by the target deep learning model; sorting the products included in the third product set in descending order of browsing probability, and using the products with arrangement numbers before a predetermined threshold as the multiple products recommended to the target account.

[0010] Optionally, the third product set is input into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set, including: obtaining the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, wherein the product feature parameters include at least one of the following: price, sales volume, number of evaluations, and evaluation results; the account feature parameters include at least one of the following: historical purchase records, historical search records, and historical browsing records; the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, are input into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set.

[0011] According to another aspect of an embodiment of the present invention, a product recommendation device is provided, including: an acquisition module, configured to acquire keywords for browsing products corresponding to a target account, and categories to which the browsed products belong; an output module, configured to input the keywords and categories to which the browsed products belong into a target large language model, and the target large language model outputs a first product set and a second product set, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within categories matching the categories to which they belong, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords of the sample product and the categories to which the sample belongs; a calculation module, configured to obtain a union of the first product set and the second product set to obtain a third product set; and a recommendation module, configured to sort the product identifiers included in the third product set, and sequentially select multiple products recommended to the target account.

[0012] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided. The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executed by any one of the product recommendation methods.

[0013] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the product recommendation methods.

[0014] According to yet another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, which implements any one of the above-mentioned product recommendation methods when executed by a processor.

[0015] In an embodiment of the present invention, keywords for browsing products corresponding to a target account and categories to which the browsed products belong are obtained; the keywords and categories are input into a target large language model, and the target large language model outputs a first product set and a second product set, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within categories matching the categories, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, and the first training sample includes sample keywords of the sample products and categories to which the samples belong; the first product set and the second product set are unioned to obtain a third product set; the product identifiers included in the third product set are sorted, and multiple products recommended to the target account are selected in sequence, thereby solving the technical problem in the prior art of being unable to accurately and efficiently recommend products to users, achieving the purpose of using a large model for recommendation, being able to utilize knowledge in the public domain, and mining the relationship between products, thereby achieving the technical effect of improving the accuracy and efficiency of product recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 This is a flowchart of a product recommendation method provided by an embodiment of the present invention;

[0018] Figure 2 is a schematic diagram of a product recommendation device provided according to an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] According to an embodiment of the present invention, an embodiment of a method for recommending products is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0023] Figure 1 is a flow chart of a product recommendation method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0024] Step S102: Obtain the keywords for browsing products corresponding to the target account, as well as the categories to which the browsed products belong;

[0025] In this step, we first retrieve information about all viewed products from the user's browsing history. This information typically includes the product ID, browsing time, and product title. Keyword extraction is then performed for each viewed product, identifying core product terms or attribute terms that accurately reflect key information such as the product's type, purpose, and style. Furthermore, we determine the category of the viewed product based on the product title or other information, as well as the platform's pre-defined product category system. For example, if the user browsed a down jacket, the category of the viewed product would be apparel.

[0026] Step S104: Input the keywords and categories to which they belong into the target large language model, and the target large language model outputs a first set of products and a second set of products, wherein the first set of products includes product identifiers that match the keywords, and the second set of products includes product identifiers within categories that match the categories to which they belong. The target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords and categories of sample products.

[0027] In this step, the keywords and categories of products in the user's browsing history are input into the large language model to obtain a first set of products that match the keywords of the products in the browsing history and a second set of products that match the categories of the products in the browsing history. The products included in the first set of products are similar or complementary to the products browsed by the user at the keyword level and may be products that the user may be interested in. The products included in the second set of products are in the same or related categories and have potential to match the products browsed by the user.

[0028] In an optional embodiment, the target large language model is obtained by: obtaining the product name of the sample product; determining the sample keyword corresponding to the sample product and the sample category corresponding to the sample product based on the product name of the sample product; generating a first training sample based on the sample keyword and the sample category; and fine-tuning the pre-trained large language model based on the first training sample to obtain the target large language model.

[0029] Optionally, a series of sample products are first collected. These can be combinations frequently purchased together in historical sales data, products in expert-configured pairing rules, or pairing products mentioned in user reviews. Key attribute words are extracted for each sample product, and the category to which each sample product belongs is determined based on the platform's preset product classification system. The keywords and categories of each sample product are used as a set of training samples, and the pre-trained large language model is fine-tuned. After fine-tuning, the pre-trained large language model becomes a "target large language model" that is more focused on the task of recommending product pairings. A pre-trained large language model is a model that has been trained on massive amounts of text data and possesses strong language understanding and generation capabilities. However, these models may lack professionalism and specificity in specific areas, such as e-commerce product pairing recommendations. Therefore, by fine-tuning the pre-trained large language model using the first training sample, the model can better understand the pairing rules and user preferences within the product domain, improving its performance on the product pairing recommendation task.

[0030] In an optional embodiment, based on the product name of the sample product, the sample keywords corresponding to the sample product and the category to which the sample corresponding to the sample product belongs are determined, including: segmenting the product name of the sample product to obtain the sample keywords corresponding to the sample product; matching the sample keywords with a predetermined category database to determine the category to which the sample corresponding to the sample product belongs.

[0031] Optionally, word segmentation is a fundamental task in natural language processing, which aims to segment continuous text into independent lexical units. Based on word segmentation, the resulting lexical units are analyzed to determine which words best represent the core attributes or characteristics of the product. The extracted "sample keywords" are matched against a predetermined "category database" to determine the category to which the sample product belongs. The category database is a structured data set containing a large number of product categories and their descriptions.

[0032] Step S106: finding the union of the first product set and the second product set to obtain a third product set;

[0033] In this step, the first product set and the second product set are unioned to integrate all products in the two sets to create a comprehensive product set. This can capture more potential products that may be paired with the products browsed by the user, avoid missing important combinations, and avoid duplicate combinations.

[0034] Step S108 : sorting the commodity identifiers included in the third commodity set, and selecting a plurality of commodities recommended to the target account in order.

[0035] Optionally, the items in the third product set are sorted. The sorting process may involve consideration of multiple features, including item similarity scores, users' past behavior with similar items (e.g., clicks, purchases, and favorites), product ratings, prices, promotional information, etc. This sorting process allows users to select items from the third product set that are most likely to pique their interest. This personalized sorting takes into account users' historical behavior and preferences, as well as the comprehensive characteristics of the items, thereby generating a highly relevant and attractive recommendation list.

[0036] In an optional embodiment, a third product set is input into a target deep learning model, and the target deep learning model outputs a plurality of products recommended for a target account, wherein the target deep learning model is obtained by training the original deep learning model using a second training sample, and the second training sample includes historically recommended product identifiers and historically browsed product identifiers that have been browsed in the historically recommended product identifiers.

[0037] Optionally, historical recommended product identifiers are sets of products recommended to users in the past, which may not have been viewed or purchased by users. Historically browsed product identifiers are identifiers of products that users have actually viewed or purchased, representing the user's interest and preference in products. By comparing historically recommended product identifiers with historically browsed product identifiers, the system can construct a dataset for training deep learning models. The model will learn which features or attributes are associated with user interests, and how to predict the user's potential interest in products based on these features. After the third product set is input into the target deep learning model, the model will recommend products to the user based on the user preferences and product features learned during the training process.

[0038] In an optional embodiment, the third product set is input into the target deep learning model, and the target deep learning model outputs multiple products recommended to the target account, including: inputting the third product set into the target deep learning model, and predicting the browsing probability of each product included in the third product set by the target deep learning model; sorting the products included in the third product set in descending order of browsing probability, and using the products with arrangement numbers before a predetermined threshold as the multiple products recommended to the target account.

[0039] Optionally, after the third product set is fed into the target deep learning model, the model outputs a predicted recommendation score or probability for each product identifier based on user preferences and product features learned during training. The product identifiers are then sorted in descending order based on these scores or probabilities, forming a list of decreasing relevance to the user's interests. The system then sequentially selects a certain number of product identifiers, starting from the top of the list, as the final set of products recommended to the target account, for example, selecting the top 10 products and recommending them to the user.

[0040] In an optional embodiment, the third product set is input into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set, including: obtaining the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, wherein the product feature parameters include at least one of the following: price, sales volume, number of evaluations, and evaluation results; the account feature parameters include at least one of the following: historical purchase records, historical search records, and historical browsing records; the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, are input into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set.

[0041] Optionally, for each item in the third product set, the system collects a series of product feature parameters, including price, sales volume, number of reviews, and review results. These feature parameters provide objective information about the item, helping the model make recommendations based on its market performance and feedback. The system also collects feature parameters for the target account, which reflect the user's past shopping behavior and preferences. These parameters can include historical purchases, searches, and browsing history, helping the model make recommendations based on user preferences. The product and account feature parameters are combined and fed into the target deep learning model as input data. The target deep learning model is a trained model that processes and integrates this multi-dimensional feature data to learn the associations between users and items and their preferences. After receiving the input feature parameters, the target deep learning model applies its learned patterns to predict the probability of each item being viewed by the target account. This prediction process comprehensively considers the item's attributes and the user's historical behavior, outputting a probability value that reflects the likelihood of the item being viewed by the user in the current context.

[0042] Through the above, keywords for browsing products corresponding to the target account and the categories to which the browsed products belong are obtained; the keywords and the categories to which they belong are input into the target large language model, and the target large language model outputs a first product set and a second product set, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within the category matching the category to which they belong, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, and the first training sample includes sample keywords of the sample product and the category to which the sample belongs; the first product set and the second product set are unioned to obtain a third product set; the product identifiers included in the third product set are sorted, and multiple products recommended to the target account are selected in sequence, thereby solving the technical problem in the prior art that products cannot be accurately and efficiently recommended to users, achieving the purpose of using a large model for recommendation, being able to utilize knowledge in the public domain, and mining the relationship between products, thereby achieving the technical effect of improving the accuracy and efficiency of product recommendations.

[0043] Based on the above embodiments and optional embodiments, the present invention proposes the following optional implementation manner.

[0044] Step S1, obtaining the keywords of the products browsed by the target account and the categories to which the products browsed belong;

[0045] Optionally, first obtain information about all viewed products from the user's browsing history. This information typically includes the product ID, browsing time, and product title. Keyword extraction is then performed for each viewed product, identifying core product terms or attribute terms that accurately reflect key information such as the product's type, purpose, and style. Furthermore, the product's category is determined based on the product title or other information, as well as the platform's pre-defined product category system. For example, if the user browsed a down jacket, the category described for the viewed product is apparel.

[0046] Step S2: Input the keywords and categories to which they belong into the target large language model, and the target large language model outputs a first set of products and a second set of products, wherein the first set of products includes product identifiers that match the keywords, and the second set of products includes product identifiers within categories that match the categories to which they belong. The target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords and categories of sample products.

[0047] In this step, the keywords and categories of products in the user's browsing history are input into the large language model to obtain a first set of products that match the keywords of the products in the browsing history and a second set of products that match the categories of the products in the browsing history. The products included in the first set of products are similar or complementary to the products browsed by the user at the keyword level and may be products that the user may be interested in. The products included in the second set of products are in the same or related categories and have potential to match the products browsed by the user.

[0048] In step S21, the target large language model is obtained by: obtaining the product name of the sample product; determining the sample keyword corresponding to the sample product and the sample category corresponding to the sample product based on the product name of the sample product; generating a first training sample based on the sample keyword and the sample category; and fine-tuning the pre-trained large language model based on the first training sample to obtain the target large language model.

[0049] Optionally, a series of sample products are first collected. These can be combinations frequently purchased together in historical sales data, products in expert-configured pairing rules, or pairing products mentioned in user reviews. Key attribute words are extracted for each sample product, and the category to which each sample product belongs is determined based on the platform's preset product classification system. The keywords and categories of each sample product are used as a set of training samples, and the pre-trained large language model is fine-tuned. After fine-tuning, the pre-trained large language model becomes a "target large language model" that is more focused on the task of recommending product pairings. A pre-trained large language model is a model that has been trained on massive amounts of text data and possesses strong language understanding and generation capabilities. However, these models may lack professionalism and specificity in specific areas, such as e-commerce product pairing recommendations. Therefore, by fine-tuning the pre-trained large language model using the first training sample, the model can better understand the pairing rules and user preferences within the product domain, improving its performance on the product pairing recommendation task.

[0050] Step S22, based on the product name of the sample product, determines the sample keyword corresponding to the sample product and the category to which the sample corresponding to the sample product belongs, including: segmenting the product name of the sample product to obtain the sample keyword corresponding to the sample product; matching the sample keyword with a predetermined category database to determine the category to which the sample corresponding to the sample product belongs.

[0051] Optionally, word segmentation is a fundamental task in natural language processing, which aims to segment continuous text into independent lexical units. Based on word segmentation, the resulting lexical units are analyzed to determine which words best represent the core attributes or characteristics of the product. The extracted "sample keywords" are matched against a predetermined "category database" to determine the category to which the sample product belongs. The category database is a structured data set containing a large number of product categories and their descriptions.

[0052] Step S3, finding the union of the first product set and the second product set to obtain a third product set;

[0053] Optionally, in this step, the first product set and the second product set are unioned to integrate all products in the two sets to create a comprehensive product set, which can capture more potential products that may be paired with the products browsed by the user, avoid missing important combinations, and avoid repeated combinations.

[0054] Step S4: sort the commodity identifiers included in the third commodity set, and select multiple commodities recommended to the target account in order.

[0055] Optionally, the items in the third product set are sorted. The sorting process may involve consideration of multiple features, including item similarity scores, users' past behavior with similar items (e.g., clicks, purchases, and favorites), product ratings, prices, promotional information, etc. This sorting process allows users to select items from the third product set that are most likely to pique their interest. This personalized sorting takes into account users' historical behavior and preferences, as well as the comprehensive characteristics of the items, thereby generating a highly relevant and attractive recommendation list.

[0056] Step S41, sorting the product identifiers included in the third product set, and selecting multiple products recommended for the target account in order, including: inputting the third product set into the target deep learning model, and outputting multiple products recommended for the target account by the target deep learning model, wherein the target deep learning model is obtained by training the original deep learning model through the second training sample, and the second training sample includes historically recommended product identifiers, and historically browsed product identifiers that have been browsed in the historically recommended product identifiers.

[0057] Optionally, historical recommended product identifiers are sets of products recommended to users in the past, which may not have been viewed or purchased by users. Historically browsed product identifiers are identifiers of products that users have actually viewed or purchased, representing the user's interest and preference in products. By comparing historically recommended product identifiers with historically browsed product identifiers, the system can construct a dataset for training deep learning models. The model will learn which features or attributes are associated with user interests, and how to predict the user's potential interest in products based on these features. After the third product set is input into the target deep learning model, the model will recommend products to the user based on the user preferences and product features learned during the training process.

[0058] Step S42: Input the third product set into the target deep learning model, and the target deep learning model outputs multiple products recommended to the target account, including: inputting the third product set into the target deep learning model, and predicting the browsing probability of each product included in the third product set by the target deep learning model; sorting the products included in the third product set in descending order of browsing probability, and using the products with arrangement numbers before a predetermined threshold as the multiple products recommended to the target account.

[0059] Optionally, after the third product set is fed into the target deep learning model, the model outputs a predicted recommendation score or probability for each product identifier based on user preferences and product features learned during training. The product identifiers are then sorted in descending order based on these scores or probabilities, forming a list of decreasing relevance to the user's interests. The system then sequentially selects a certain number of product identifiers, starting from the top of the list, as the final set of products recommended to the target account, for example, selecting the top 10 products and recommending them to the user.

[0060] Step S43: Input the third product set into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set, including: obtaining the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, wherein the product feature parameters include at least one of the following: price, sales volume, number of evaluations, and evaluation results; the account feature parameters include at least one of the following: historical purchase records, historical search records, and historical browsing records; input the product feature parameters of each product included in the third product set, and the account feature parameters corresponding to the target account, into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set.

[0061] Optionally, for each item in the third product set, the system collects a series of product feature parameters, including price, sales volume, number of reviews, and review results. These feature parameters provide objective information about the item, helping the model make recommendations based on its market performance and feedback. The system also collects feature parameters for the target account, which reflect the user's past shopping behavior and preferences. These parameters can include historical purchases, searches, and browsing history, helping the model make recommendations based on user preferences. The product and account feature parameters are combined and fed into the target deep learning model as input data. The target deep learning model is a trained model that processes and integrates this multi-dimensional feature data to learn the associations between users and items and their preferences. After receiving the input feature parameters, the target deep learning model applies its learned patterns to predict the probability of each item being viewed by the target account. This prediction process comprehensively considers the item's attributes and the user's historical behavior, outputting a probability value that reflects the likelihood of the item being viewed by the user in the current context.

[0062] The above optional implementation method achieves at least the following effects: it solves the technical problem in the prior art of being unable to accurately and efficiently recommend products to users, achieves the purpose of accurately and efficiently recommending products to users, and further achieves the technical effect of improving the accuracy and efficiency of product recommendations.

[0063] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0064] In this embodiment, a product recommendation device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation methods, and will not be repeated here. As used below, the terms "module" and "device" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation by hardware, or a combination of software and hardware, is also possible and conceivable.

[0065] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing a product recommendation method. Figure 2 is a schematic diagram of a product recommendation device according to an embodiment of the present invention. Figure 2 As shown, the above-mentioned product recommendation device includes an acquisition module 21, an output module 22, a calculation module 23, and a recommendation module 24. The device is described below.

[0066] The acquisition module 21 is used to obtain the keywords of the browsed products corresponding to the target account and the categories to which the browsed products belong;

[0067] Output module 22, connected to acquisition module 21, is configured to input keywords and categories into a target large language model, and the target large language model outputs a first product set and a second product set, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within categories matching the categories. The target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords and the categories of the sample products.

[0068] A calculation module 23, connected to the output module 22, is configured to obtain a union of the first product set and the second product set to obtain a third product set;

[0069] The recommendation module 24 is connected to the calculation module 23 and is configured to sort the commodity identifiers included in the third commodity set and sequentially select a plurality of commodities to be recommended to the target account.

[0070] In a product recommendation device provided by an embodiment of the present invention, an acquisition module is set to obtain keywords for browsing products corresponding to a target account and the categories to which the browsed products belong; an output module is used to input the keywords and the categories to which the browsed products belong into a target large language model, and the target large language model outputs a first product set and a second product set, wherein the first product set includes product identifiers matching the keywords, and the second product set includes product identifiers within the categories that match the categories to which they belong, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, and the first training sample includes sample keywords of the sample products and the categories to which the samples belong; a calculation module is used to calculate the union of the first product set and the second product set to obtain a third product set; and a recommendation module is used to sort the product identifiers included in the third product set and sequentially select multiple products to recommend to the target account. This solves the technical problem in the prior art of being unable to accurately and efficiently recommend products to users, achieves the purpose of using a large model for recommendation, and can utilize knowledge in the public domain to mine the relationship between products, thereby achieving the technical effect of improving the accuracy and efficiency of product recommendations.

[0071] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0072] It should be noted that the acquisition module 21, output module 22, calculation module 23, and recommendation module 24 correspond to steps S102 to S108 in the embodiment. The examples and application scenarios implemented by these modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that these modules, as part of the device, can be run on a computer terminal.

[0073] It should be noted that the optional or preferred implementation of this embodiment can be found in the relevant description in the embodiment, which will not be repeated here.

[0074] The above-mentioned product recommendations may also include a processor and a memory. The acquisition module 21, the output module 22, the calculation module 23, the recommendation module 24, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0075] The processor includes a kernel, which retrieves the corresponding program unit from memory. There can be one or more kernels. Memory may include non-permanent memory in a computer-readable medium, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0076] An embodiment of the present invention provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a product recommendation method is implemented.

[0077] like Figure 3 As shown, an embodiment of the present invention provides an electronic device, the electronic device 10 includes a processor, a memory, and a program stored in the memory and runnable on the processor. When the processor executes the program, the following steps are implemented: the memory is used to store a computer program, wherein, when the computer program is executed by the processor, the processor implements the above-mentioned product recommendation method. The device in this article can be a server, PC, etc.

[0078] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: computer instructions are executed by a processor to perform the above-mentioned product recommendation method.

[0079] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0083] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0084] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0085] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0086] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0087] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] The above are merely embodiments of the present invention and are not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A product recommendation method, characterized in that: include: Obtain the keywords of the browsed products corresponding to the target account, as well as the categories to which the browsed products belong; Inputting the keywords and the categories into a target large language model, the target large language model outputting a first set of products and a second set of products, wherein the first set of products includes product identifiers matching the keywords, and the second set of products includes product identifiers within categories matching the categories, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords and the categories of sample products; Calculating the union of the first product set and the second product set to obtain a third product set; The commodity identifiers included in the third commodity set are sorted, and a plurality of commodities recommended to the target account are selected in sequence.

2. The method according to claim 1, characterized in that The target large language model is obtained by: Obtain the product name of the sample product; Determine, based on the product name of the sample product, the sample keyword corresponding to the sample product and the category to which the sample product belongs; generating the first training sample according to the sample keyword and the category to which the sample belongs; Fine-tune the pre-trained large language model according to the first training sample to obtain the target large language model.

3. The method according to claim 2, characterized in that The determining, based on the product name of the sample product, the sample keyword corresponding to the sample product and the category to which the sample product belongs includes: Segmenting the product name of the sample product to obtain sample keywords corresponding to the sample product; The sample keywords are matched with a predetermined category database to determine the category to which the sample product belongs.

4. The method according to any one of claims 1 to 3, characterized in that The step of sorting the product identifiers included in the third product set and sequentially selecting a plurality of products to be recommended to the target account includes: The third product set is input into a target deep learning model, and the target deep learning model outputs the multiple products recommended for the target account, wherein the target deep learning model is obtained by training the original deep learning model through a second training sample, and the second training sample includes historically recommended product identifiers and historically browsed product identifiers that have been browsed among the historically recommended product identifiers.

5. The method according to claim 4, characterized in that Inputting the third product set into a target deep learning model, and having the target deep learning model output the plurality of products recommended for the target account, includes: Inputting the third product set into the target deep learning model, and having the target deep learning model predict the browsing probability of each product included in the third product set; The commodities included in the third commodity set are sorted in descending order of browsing probability, and the commodities whose arrangement numbers are before a predetermined threshold are used as the plurality of commodities recommended to the target account.

6. The method according to claim 5, characterized in that Inputting the third product set into the target deep learning model, and predicting, by the target deep learning model, the browsing probability of each product included in the third product set, includes: Obtaining product characteristic parameters of each product included in the third product set and account characteristic parameters corresponding to the target account, wherein the product characteristic parameters include at least one of the following: price, sales volume, number of reviews, and review results; and the account characteristic parameters include at least one of the following: historical purchase records, historical search records, and historical browsing records; The product feature parameters of each product included in the third product set and the account feature parameters corresponding to the target account are input into the target deep learning model, and the target deep learning model predicts the browsing probability of each product included in the third product set.

7. A product recommendation device, characterized in that: include: An acquisition module is used to obtain keywords of browsed products corresponding to a target account and the categories to which the browsed products belong; an output module, configured to input the keywords and the categories into a target large language model, and output from the target large language model a first set of products and a second set of products, wherein the first set of products includes product identifiers matching the keywords, and the second set of products includes product identifiers within categories matching the categories, wherein the target large language model is obtained by fine-tuning a pre-trained large language model using a first training sample, wherein the first training sample includes sample keywords and the categories of sample products; a calculation module, configured to obtain a union of the first product set and the second product set to obtain a third product set; The recommendation module is configured to sort the commodity identifiers included in the third commodity set and sequentially select a plurality of commodities recommended to the target account.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the product recommendation method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: One or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the product recommendation method described in any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that The computer instructions are executed by a processor to execute the product recommendation method according to any one of claims 1 to 6.

Images