Product recommendation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a product recommendation method and device, electronic equipment and a storage medium, and belongs to the technical field of text processing and artificial intelligence. The method comprises the steps that a user information text of a target user is acquired, and product information of alternative products is acquired; text embedding is carried out on the user information text to obtain a text embedding vector, and the text embedding vector comprises a phrase embedding vector; performing semantic understanding on the text embedding vector to obtain context representation of the word group embedding vector; based on context representation, performing feature extraction on the text embedding vector to obtain target user information features; performing feature extraction on the product information to obtain product features; performing feature matching on the product features based on the target user information features to obtain target features, and screening out target products from the alternative products based on the target features; and recommending products to the target user based on the target product. According to the embodiment of the invention, the product recommendation accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of text processing and artificial intelligence technology, and in particular to a product recommendation method and device, electronic device and storage medium. Background Art

[0002] Product recommendation refers to recommending related products to users based on their relevant information. For example, in the field of product sales in the financial sector, it is necessary to assess the user's loan repayment ability based on the user's credit information, spending power, job stability and other information, and then recommend related products to the user, such as insurance products, based on the results of the loan repayment ability assessment.

[0003] At present, the common product recommendation method is usually to manually interpret user information to recommend products to users. Since the manual interpretation method has strong human subjectivity, it is easy to cause product recommendation errors, resulting in a decrease in the accuracy of product recommendations. Therefore, how to improve the accuracy of product recommendations has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a product recommendation method and device, an electronic device and a storage medium, aiming to improve the accuracy of product recommendations.

[0005] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a product recommendation method, the method comprising:

[0006] Obtain basic information and self-description information of target users, and obtain product information of candidate products;

[0007] Merging the user basic information and the user self-description information to obtain a user information text;

[0008] Performing text embedding on the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector;

[0009] Performing semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector;

[0010] Based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features;

[0011] Extracting features from the product information based on a preset product feature extraction model to obtain product features;

[0012] Based on the target user information feature, feature matching is performed on the product feature to obtain a target feature, and based on the target feature, a target product is screened out from the candidate products;

[0013] A product recommendation is made to the target user based on the target product.

[0014] In some embodiments, the performing semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector includes:

[0015] Performing a linear transformation on the phrase embedding vector to obtain a phrase query vector, a phrase key vector, and a phrase value vector;

[0016] Based on the phrase query vector and the phrase key vector, calculating the similarity between the phrase embedding vectors to obtain an attention score;

[0017] Performing function activation on the attention score to obtain the attention weight;

[0018] Based on the attention weights, the phrase value vectors are weighted summed to obtain a contextual representation of the phrase embedding vector.

[0019] In some embodiments, the feature extraction is performed on the text embedding vector based on the context representation to obtain target user information features;

[0020] Based on the context representation and preset user key features, feature matching is performed on the text embedding vector to obtain a user feature vector;

[0021] Based on the user feature vector, generating an information feature description text of the target user;

[0022] Feature recognition is performed on the information feature description text to obtain the target user information feature.

[0023] In some embodiments, the merging of the user basic information and the user self-description information to obtain the user information text includes:

[0024] Based on preset user information keywords, extract key information from the user basic information to obtain user key information;

[0025] Based on the user key information, the user self-description information is supplemented to obtain the user information text.

[0026] In some embodiments, the performing feature matching on the product features based on the target user information features to obtain target features, and screening out the target product from the candidate products based on the target features, includes:

[0027] Calculating feature similarity between the target user information feature and the product feature to obtain feature similarity;

[0028] Based on the feature similarity, screening out the target feature from the product features;

[0029] Based on the target features, product matching is performed on the candidate products to obtain the target product.

[0030] In some embodiments, extracting features from the product information based on a preset product feature extraction model to obtain product features includes:

[0031] Perform word segmentation processing on the product information based on the product feature extraction model to obtain product keyword groups;

[0032] Performing named entity recognition on the product keyword group to obtain product feature entities;

[0033] Based on the product feature entity, keywords are extracted from the product information to obtain the product features.

[0034] In some embodiments, the text embedding of the user information text based on a preset text feature extraction model to obtain a text embedding vector includes:

[0035] Performing text preprocessing on the user information text to obtain a preprocessed text;

[0036] Based on the preprocessed text, construct a text vocabulary;

[0037] Encoding the text vocabulary based on the text feature extraction model to obtain the phrase embedding vector;

[0038] Perform vector concatenation on the phrase embedding vectors to obtain the text embedding vector.

[0039] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a product recommendation device, the device comprising:

[0040] An information acquisition module is used to obtain basic user information and user self-description information of the target user, and obtain product information of the candidate products;

[0041] An information synthesis module, used for merging the user basic information and the user self-description information to obtain a user information text;

[0042] A text embedding module, used for embedding the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector;

[0043] A semantic understanding module, used to perform semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector;

[0044] A text feature extraction module, used to extract features from the text embedding vector based on the context representation to obtain target user information features;

[0045] A product feature extraction module, used to extract features from the product information based on a preset product feature extraction model to obtain product features;

[0046] A product screening module, configured to perform feature matching on the product features based on the target user information features to obtain target features, and screen the target product from the candidate products based on the target features;

[0047] A product recommendation module is used to recommend products to the target user based on the target product.

[0048] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0049] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0050] The product recommendation method and device, electronic device and storage medium proposed in the present application merge the basic information of the target user and the self-description information of the user, so that the merged user information text describes the target user more accurately. Furthermore, the user information text is embedded through a text feature extraction model, and the text embedding vector is semantically understood to obtain the context representation of the phrase embedding vector. Then, based on the context representation, the text embedding vector is feature extracted to obtain the target user information feature, thereby reducing the manual participation in the information feature extraction link of the user information, thereby improving the extraction accuracy of the information feature of the target user. Secondly, by extracting the product information of the candidate products according to the preset product feature extraction model, the product features are obtained, thereby reducing the manual participation in the information feature extraction link of the candidate products, thereby improving the extraction accuracy of the information feature of the candidate products. Finally, according to the target user information features, the product features are feature matched to obtain the target features. According to the target features, the target product is screened out from the candidate products, and the product is recommended to the target user based on the target product, so that the recommended product is more in line with the target user's own conditions, thereby improving the accuracy of product recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of a product recommendation method provided in an embodiment of the present application;

[0052] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0053] Figure 3 yes Figure 1 Flow chart of step S103 in FIG.

[0054] Figure 4 yes Figure 1 Flow chart of step S104 in FIG.

[0055] Figure 5 yes Figure 1 Flow chart of step S105 in FIG.

[0056] Figure 6 yes Figure 1 Flow chart of step S106 in FIG.

[0057] Figure 7 yes Figure 1 Flow chart of step S107 in FIG.

[0058] Figure 8 is a schematic diagram of the structure of a product recommendation device provided in an embodiment of the present application;

[0059] Fig. 9It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0061] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0063] First, some nouns involved in this application are analyzed:

[0064] Big Language Model: Big Language Model is a natural language processing model based on deep learning technology. It can understand and generate natural language by training massive text data. These models have strong context understanding capabilities and can perform various language tasks such as text generation, translation, question and answer, and dialogue. They are widely used in fields such as intelligent assistants, customer service systems, and content creation.

[0065] Attention mechanism: The attention mechanism is a computational method that mimics human visual attention, which aims to enable the model to dynamically focus on the most relevant information when processing input data, thereby improving performance. Its benefit is that it can effectively capture long-distance dependencies and important features, significantly improving the accuracy and efficiency of tasks such as natural language processing, image recognition, and translation. By adaptively adjusting the focus of attention, the attention mechanism enhances the model's expressiveness and enables it to perform better in complex tasks.

[0066] Product recommendation refers to recommending related products to users based on their relevant information. For example, in the field of product sales in the financial sector, it is necessary to assess the user's loan repayment ability based on the user's credit information, spending power, job stability and other information, and then recommend related products to the user, such as insurance products, based on the results of the loan repayment ability assessment.

[0067] At present, the common product recommendation method is usually to manually interpret user information to recommend products to users. Since the manual interpretation method has strong human subjectivity, it is easy to cause product recommendation errors, resulting in a decrease in the accuracy of product recommendations. Therefore, how to improve the accuracy of product recommendations has become a technical problem that needs to be solved urgently.

[0068] Based on this, the embodiments of the present application provide a product recommendation method and device, an electronic device and a storage medium, aiming to improve the accuracy of product recommendations.

[0069] The product recommendation method and device, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the product recommendation method in the embodiments of the present application is described.

[0070] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0071] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0072] The product recommendation method provided in the embodiment of the present application relates to the field of text processing and artificial intelligence technology. The product recommendation method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the product recommendation method, etc., but is not limited to the above forms.

[0073] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0074] Figure 1 is an optional flowchart of the product recommendation method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S108.

[0075] Step S101, obtaining basic user information and user self-description information of the target user, and obtaining product information of the candidate products;

[0076] Step S102, merging the user basic information and the user self-description information to obtain a user information text;

[0077] Step S103, embedding the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector;

[0078] Step S104, performing semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector;

[0079] Step S105, based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features;

[0080] Step S106, extracting features from the product information based on a preset product feature extraction model to obtain product features;

[0081] Step S107, based on the target user information feature, feature matching is performed on the product feature to obtain the target feature, and based on the target feature, the target product is screened out from the candidate products;

[0082] Step S108: recommending products to target users based on the target products.

[0083] Steps S101 to S107 shown in the embodiment of the present application, by acquiring the basic user information and the user self-description information of the target user, realize information merging of the basic user information and the user self-description information, thereby obtaining a user information text. At the same time, it is also necessary to obtain the product information of the candidate products. Further, the user information text is embedded according to a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector. Secondly, the text embedding vector is semantically understood according to the feature extraction model to obtain a context representation of the phrase embedding vector, and based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features. Finally, according to a preset product feature extraction model, feature extraction is performed on the above product information to obtain product features, and then feature matching is performed on the product features according to the target user information features to obtain target features. According to the target features, the target product is screened out from the candidate products, and product recommendations are made to the target user based on the target product. Therefore, the present application merges the basic user information and the self-description information of the target user, so that the merged user information text describes the target user more accurately. Furthermore, the user information text is embedded through a text feature extraction model, and the text embedding vector is semantically understood to obtain the context representation of the phrase embedding vector. Then, based on the context representation, feature extraction is performed on the text embedding vector to obtain the target user information feature, thereby reducing the manual participation in the information feature extraction link of the user information, thereby improving the extraction accuracy of the information feature of the target user. Secondly, by extracting features from the product information of the candidate products according to a preset product feature extraction model to obtain product features, the manual participation in the information feature extraction link of the candidate products is reduced, thereby improving the extraction accuracy of the information feature of the candidate products. Finally, based on the target user information features, feature matching is performed on the product features to obtain the target features. Based on the target features, the target product is screened out from the candidate products, and the product is recommended to the target user based on the target product, so that the recommended product is more in line with the target user's own conditions, thereby improving the accuracy of product recommendation.

[0084] In step S101 of some embodiments, basic user information refers to the product-related attributes of the target user. For example, in the loan product recommendation business, basic user information may include the user's name, age, average salary, and other information. User self-description information refers to the product-related attributes of the target user described in a self-descriptive tone. For example, in the loan product recommendation business, the user self-description information may be "In the past year, my job was stable and my salary did not change much." Product information refers to the basic attributes contained in the candidate products. For example, the product information of a loan product may include information such as the loan amount, loan interest rate, and repayment method.

[0085] When the target user is purchasing a product, the present application can collect the target user's information by displaying an information filling page to the target user, thereby obtaining the target user's basic information and user self-description information. It should be noted that the information that needs to be collected in the information filling page can be freely filled in by the target user or selected by the target user from fixed options.

[0086] In addition, the present application can obtain product information of the candidate products by performing feature extraction on the data carried by the candidate products.

[0087] In step S102 of some embodiments, the user information text is obtained by supplementing the valid information in the user basic information to the user self-description information.

[0088] For details, see Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S202:

[0089] Step S201, extracting key information from basic user information based on preset user information keywords to obtain key user information;

[0090] Step S201: supplement the user's self-description information based on the user's key information to obtain a user information text.

[0091] In step S201 of some embodiments, the user information keyword may be a key information positioning phrase pre-set according to the product type. For example, in a loan product recommendation service, if the user's basic information is "average monthly salary 5,000", the user information keyword may be average monthly salary.

[0092] The present application can locate the position of key information that needs to be extracted in the user's basic information through pre-set user information keywords, and then extract information based on the position to obtain the user's key information. For example, when the user's basic information is "age 28, average monthly salary 5000", and the user information keyword is the average monthly salary, the key information 5000 in the user's basic information can be located, and then the data can be extracted to obtain the user's key information average monthly salary 5000.

[0093] In step S202 of some embodiments, the information filling position in the user self-description information is determined by searching for information related to the user key information in the user self-description information, and then the above user key information is filled into the information filling position to obtain the user information text. For example, when the user self-description information is "In the past year, my job was stable and the salary did not change much", and the user key information is an average monthly salary of 5,000, it can be determined that the information filling position of the user self-description information is after "The salary did not change much". Therefore, after supplementing the user self-description information, the user information text can be obtained as "In the past year, my job was stable and the salary did not change much, with an average monthly salary of 5,000".

[0094] In steps S201 to S202 shown in the present embodiment, key information is extracted from the user's basic information based on preset user information keywords to obtain user key information, and then the user's self-description information is supplemented based on the user key information to obtain user information text, so that more accurate and detailed user information can be obtained, thereby improving the accuracy of product recommendations to users.

[0095] In step S103 of some embodiments, the text feature extraction model can be a large language model such as Bert (Bidirectional Encoder Representations from Transformers) and GPT (Generative Pre-trained Transformer), and the text feature extraction model includes a text embedding module, a semantic understanding module and a feature extraction module, wherein the text embedding module can be a continuous bag of words model or a skip-word model, and the semantic understanding module and the feature extraction module can be implemented using a transformer architecture.

[0096] The present application can obtain a preprocessed text by preprocessing the user information text, then encode the phrases in the preprocessed text one by one to obtain a phrase embedding vector, and then concatenate the phrase embedding vectors in the order of the user information text to obtain a text embedding vector.

[0097] For details, see Figure 3 In some embodiments, step S103 may include but is not limited to steps S301 to S304:

[0098] Step S301, performing text preprocessing on the user information text to obtain a preprocessed text;

[0099] Step S302, constructing a text vocabulary based on the preprocessed text;

[0100] Step S303, encoding the text vocabulary based on the text feature extraction model to obtain a phrase embedding vector;

[0101] Step S304, concatenating the phrase embedding vectors to obtain a text embedding vector.

[0102] In step S301 of some embodiments, the preprocessed text can be obtained by segmenting the user information text and then removing stop words from the segmented user information text. It should be noted that if the user information text contains characters with inconsistent capitalization, the capitalization needs to be unified to improve the recognition accuracy of the text feature extraction model.

[0103] In step S302 of some embodiments, based on the above segmented user information text, phrases are extracted one by one to obtain text phrases, and then table building processing is performed on the text phrases to obtain a text vocabulary.

[0104] In steps S303 to S304 of some embodiments, encoding processing of the text vocabulary can be achieved by mapping the phrases in the above-mentioned text vocabulary to the vector space of the text feature extraction model, thereby obtaining the phrase embedding vector, and further, the phrase embedding vectors are spliced ​​according to the phrase order of the user information text, so as to obtain the text embedding vector of the user information text.

[0105] In steps S301 to S304 shown in the present embodiment, by performing text preprocessing on the user information text to obtain the preprocessed text, the accuracy of the text feature extraction model's understanding of the user information text can be improved. Furthermore, the text feature extraction model is used to encode the preprocessed text word by word to obtain a phrase embedding vector, and the phrase embedding vector is vector concatenated to obtain a text embedding vector, which facilitates the subsequent understanding of the user information text by the text feature extraction model.

[0106] In step S104 of some embodiments, the contextual representation of each phrase embedding vector can be obtained by performing attention weight allocation, vector weighted summation and other processing on the embedding vector through the attention mechanism of the semantic understanding module in the text feature extraction model.

[0107] For details, see Figure 4 In some embodiments, step S104 may include but is not limited to steps S401 to S404:

[0108] Step S401, performing a linear transformation on the phrase embedding vector to obtain a phrase query vector, a phrase key vector and a phrase value vector;

[0109] Step S402, based on the phrase query vector and the phrase key vector, calculate the similarity between the phrase embedding vectors to obtain an attention score;

[0110] Step S403, performing function activation on the attention score to obtain the attention weight;

[0111] Step S404: Based on the attention weights, weighted summation is performed on the phrase value vectors to obtain a contextual representation of the phrase embedding vector.

[0112] In step S401 of some embodiments, the present application can obtain the phrase query vector, phrase key vector and phrase value vector of the phrase embedding vector by performing vector product with the pre-set model parameters. It should be noted that the model parameters corresponding to the phrase query vector, phrase key vector and phrase value vector are different.

[0113] In step S402 of some embodiments, the phrase query vector of the target phrase embedding vector in the phrase embedding vector is dot-producted with the phrase key vectors of the remaining phrase embedding vectors, so that the similarity between the target phrase embedding vector and the remaining phrase embedding vectors, i.e., the attention score of the target phrase embedding vector, can be calculated.

[0114] In step S403 of some embodiments, a softmax function may be used to perform function activation on the attention score of each of the above-mentioned phrase embedding vectors, thereby obtaining the attention weight of each phrase embedding vector.

[0115] In step S404 of some embodiments, based on the above attention weights, the phrase value vectors of each phrase embedding vector are weighted summed to obtain a contextual representation of the target phrase embedding vector.

[0116] Furthermore, by repeating the method in step S401 to step S404, the context representation of each phrase embedding vector can be obtained.

[0117] In steps S401 to S404 shown in the present embodiment, a phrase query vector, a phrase key vector and a phrase value vector are obtained by performing a linear transformation on the phrase embedding vector, and then the similarity between the phrase embedding vectors is calculated based on the phrase query vector and the phrase key vector to obtain an attention score. Furthermore, a function activation is performed on the attention score to obtain an attention weight, and based on the attention weight, a weighted sum is performed on the phrase value vector to obtain a contextual representation of the phrase embedding vector, which can improve the accuracy of the text feature extraction model in understanding each phrase in the user information text, thereby improving the feature extraction accuracy of the text feature extraction model.

[0118] In step S105 of some embodiments, based on the context representation of each phrase embedding vector, the text feature extraction model can filter out target user information features that meet the product recommendation business from multiple phrase embedding vectors.

[0119] For details, see Figure 5 In some embodiments, step S105 may include but is not limited to steps S501 to S503:

[0120] Step S501, based on the context representation and preset user key features, feature matching is performed on the text embedding vector to obtain a user feature vector;

[0121] Step S502, generating a target user's information feature description text based on the user feature vector;

[0122] Step S503: perform feature recognition on the information feature description text to obtain target user information features.

[0123] In step S501 of some embodiments, the key features of the user may be feature parameters obtained when the text feature extraction model is trained, and are used to extract feature information from a certain type of user information. For example, in the loan product recommendation business, the key features of the user may be average monthly salary, average consumption, etc.

[0124] In one embodiment of the present application, the text feature extraction model can use the contextual representation of the phrase embedding vector to deepen the text understanding of the user information text, and then the text feature extraction model extracts a user feature vector that matches the pre-set user key features from the text embedding vector based on the deepened text understanding.

[0125] In step S502 of some embodiments, a pre-trained text generation model can be used to respond to the above-mentioned user feature vector to generate an information feature description text. For example, in the loan product recommendation business, based on the user feature vector of "average monthly income of 10,000, average monthly expenditure of 4,000", an information feature description text of "having the ability to repay within 6,000 per month" can be generated.

[0126] In step S503 of some embodiments, the feature phrases in the information feature description text are identified by the trained text feature extraction model to obtain the target user information features of the target user. For example, the trained loan text feature extraction model can be used to identify the target user information feature "having the ability to repay less than 6,000 yuan per month" in the loan information feature description text "having the ability to repay less than 6,000 yuan per month".

[0127] In steps S501 to S503 shown in this embodiment, feature matching is performed on the text embedding vector according to the context representation and the preset user key features to obtain the user feature vector, and then based on the user feature vector, the information feature description text of the target user is generated, which can improve the accuracy of the text feature extraction model in understanding the user information. Furthermore, feature recognition is performed on the information feature description text to obtain the target user information features, so that the extracted user information features are more accurate.

[0128] In step S106 of some embodiments, the product feature extraction model may be a NLP (Natural Language Processing) model.

[0129] This application can extract product features of candidate products from product information by using a product feature extraction model to perform steps such as word segmentation, named entity recognition and keyword extraction on product information.

[0130] For details, see Figure 6 In some embodiments, step S106 may include but is not limited to steps S601 to S603:

[0131] Step S601, performing word segmentation processing on product information based on a product feature extraction model to obtain a product keyword group;

[0132] Step S602, performing named entity recognition on the product keyword group to obtain product feature entities;

[0133] Step S603: extract keywords from product information based on product feature entities to obtain product features.

[0134] In step S601 of some embodiments, the product information of the above-mentioned alternative products is segmented through the product feature extraction model to obtain product keyword groups of the alternative products. For example, in a loan product, the product information may be "This product is a credit loan". Through word segmentation processing, the product keyword groups "this", "product", "for", and "credit loan" can be obtained.

[0135] In step S602 of some embodiments, by performing named entity recognition on the above-mentioned product keyword groups, the phrase part of speech and phrase meaning of each product keyword can be clarified, and then the product feature entity in the product keyword group is determined based on the phrase part of speech and phrase meaning of the product keyword. For example, in the product keyword groups "this", "product", "for", and "credit loan", through named entity recognition, it can be determined that the product feature entity is "credit loan".

[0136] In step S603 of some embodiments, the above-mentioned product feature entity is used as a text keyword, and then the product feature is extracted from the product information based on the text keyword.

[0137] In steps S601 to S603 shown in this embodiment, product information is segmented according to the product feature extraction model to obtain product keyword groups, and named entity recognition is then performed on the product keyword groups to obtain product feature entities. Furthermore, based on the product feature entities, keywords are extracted from the product information to obtain product features, which can improve the accuracy of the model's feature extraction of candidate products and make the product features of the candidate products extracted by the model more accurate.

[0138] In step S107 of some embodiments, after obtaining the target user information features of the target user and the product features of the alternative products, the target features that match the target user information features can be screened out from the product features by calculating the similarity between the target user information features and the product features, and then the alternative products corresponding to the target features are queried to obtain the target product.

[0139] For details, see Figure 7 In some embodiments, step S107 may include but is not limited to steps S701 to S703:

[0140] Step S701, calculating feature similarity of target user information features and product features to obtain feature similarity;

[0141] Step S702, based on feature similarity, select target features from product features;

[0142] Step S703: Based on the target features, product matching is performed on the candidate products to obtain the target product.

[0143] In step S701 and step S702 of some embodiments, the degree of match between the target user information features and the product features can be determined by calculating the feature similarity between the target user information features and the product features, and then the product features with high feature similarity to the target user information features are selected as the target features. For example, in the loan product recommendation business, the target user information feature is "having a monthly repayment capacity of 6,000 yuan", the product feature of product A is "a minimum monthly repayment of 10,000 yuan", and the product feature of product B is "a minimum monthly repayment of 5,000 yuan". Through the feature similarity calculation, the feature similarity between the product feature of product A and the target user information feature is 30%, and the feature similarity between the product feature of product B and the target user information feature is 90%, so that it can be determined that the target feature of the target user information feature is the product feature of product B.

[0144] In step S703 of some embodiments, by tracing the target feature, the candidate product corresponding to the target feature, that is, the target product, can be obtained.

[0145] In steps S701 to S703 shown in this embodiment, feature similarity is calculated for target user information features and product features to obtain feature similarity, and then target features are screened out from product features based on the feature similarity, and product matching is performed on candidate products based on the target features to obtain target products, thereby improving the accuracy of product recommendations.

[0146] In step S108 of some embodiments, after obtaining the target product, product information of the target product is sent to the target user to recommend the product to the target user. It should be noted that when the number of target products exceeds one, product recommendations can be made according to pre-set product recommendation rules. For example, multiple target products are sorted by product name and recommended to the target user in the sorted order, or multiple target products are sorted by product price and recommended to the target user in the sorted order.

[0147] The present application merges the basic information of the target user and the self-description information of the user, so that the merged user information text describes the target user more accurately. Furthermore, the user information text is embedded through a text feature extraction model, and the text embedding vector is semantically understood to obtain the context representation of the phrase embedding vector. Then, based on the context representation, feature extraction is performed on the text embedding vector to obtain the target user information feature, thereby reducing the manual participation in the information feature extraction link of the user information, thereby improving the extraction accuracy of the information feature of the target user. Secondly, by extracting features from the product information of the candidate products according to a preset product feature extraction model to obtain product features, the manual participation in the information feature extraction link of the candidate products is reduced, thereby improving the extraction accuracy of the information feature of the candidate products. Finally, based on the target user information features, feature matching is performed on the product features to obtain the target features. Based on the target features, the target product is screened out from the candidate products, and the product is recommended to the target user based on the target product, so that the recommended product is more in line with the target user's own conditions, thereby improving the accuracy of product recommendation.

[0148] See also Figure 8 The embodiment of the present application also provides a product recommendation device, which can implement the above-mentioned product recommendation method, and the device includes:

[0149] The information acquisition module 801 is used to acquire basic user information and user self-description information of the target user, and acquire product information of the candidate products;

[0150] The information synthesis module 802 is used to merge the user's basic information and the user's self-description information to obtain the user information text;

[0151] A text embedding module 803 is used to embed the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector;

[0152] A semantic understanding module 804 is used to perform semantic understanding on the text embedding vector based on a text feature extraction model to obtain a contextual representation of the phrase embedding vector;

[0153] The text feature extraction module 805 is used to extract features from the text embedding vector based on the context representation to obtain target user information features;

[0154] The product feature extraction module 806 is used to extract features from product information based on a preset product feature extraction model to obtain product features;

[0155] The product screening module 807 is used to perform feature matching on product features based on target user information features to obtain target features, and screen the target product from candidate products based on the target features;

[0156] The product recommendation module 808 is used to make product recommendations to target users based on target products.

[0157] The specific implementation of the product recommendation device is basically the same as the specific implementation of the above-mentioned product recommendation method, and will not be repeated here.

[0158] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above product recommendation method when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.

[0159] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0160] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0161] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the product recommendation method of the embodiment of this application;

[0162] Input / output interface 903, used to implement information input and output;

[0163] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0164] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0165] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0166] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned product recommendation method is implemented.

[0167] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0168] The product recommendation method, product recommendation device, electronic device and storage medium provided in the embodiments of the present application obtain the basic user information and user self-description information of the target user to realize the information merging of the basic user information and the user self-description information, thereby obtaining the user information text. At the same time, it is also necessary to obtain the product information of the candidate products. Furthermore, the user information text is embedded according to a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector. Secondly, the text embedding vector is semantically understood according to the feature extraction model to obtain a context representation of the phrase embedding vector, and based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features. Finally, according to a preset product feature extraction model, feature extraction is performed on the above-mentioned product information to obtain product features, and then feature matching is performed on the product features according to the target user information features to obtain target features. According to the target features, the target product is screened out from the candidate products, and the product is recommended to the target user based on the target product.

[0169] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0170] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0171] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0172] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0173] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application 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 data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0174] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0175] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0176] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0178] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0179] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A product recommendation method, characterized in that: The method comprises: Obtain basic information and self-description information of target users, and obtain product information of candidate products; Merging the user basic information and the user self-description information to obtain a user information text; Performing text embedding on the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector; Performing semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector; Based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features; Extracting features from the product information based on a preset product feature extraction model to obtain product features; Based on the target user information feature, feature matching is performed on the product feature to obtain a target feature, and based on the target feature, a target product is screened out from the candidate products; A product recommendation is made to the target user based on the target product.

2. The method according to claim 1, characterized in that The performing semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector includes: Performing a linear transformation on the phrase embedding vector to obtain a phrase query vector, a phrase key vector, and a phrase value vector; Based on the phrase query vector and the phrase key vector, calculating the similarity between the phrase embedding vectors to obtain an attention score; Performing function activation on the attention score to obtain the attention weight; Based on the attention weights, the phrase value vectors are weighted summed to obtain a contextual representation of the phrase embedding vector.

3. The method according to claim 1, characterized in that Based on the context representation, feature extraction is performed on the text embedding vector to obtain target user information features; Based on the context representation and preset user key features, feature matching is performed on the text embedding vector to obtain a user feature vector; Based on the user feature vector, generating an information feature description text of the target user; Feature recognition is performed on the information feature description text to obtain the target user information feature.

4. The method according to claim 1, characterized in that The user basic information and the user self-description information are merged to obtain a user information text, including: Based on preset user information keywords, extract key information from the user basic information to obtain user key information; Based on the user key information, the user self-description information is supplemented to obtain the user information text.

5. The method according to claim 1, characterized in that: The performing feature matching on the product features based on the target user information features to obtain target features, and screening out the target product from the candidate products based on the target features, includes: Calculating feature similarity between the target user information feature and the product feature to obtain feature similarity; Based on the feature similarity, screening out the target feature from the product features; Based on the target features, product matching is performed on the candidate products to obtain the target product.

6. The method according to any one of claims 1 to 5, characterized in that: The extracting features of the product information based on a preset product feature extraction model to obtain product features includes: Perform word segmentation processing on the product information based on the product feature extraction model to obtain product keyword groups; Performing named entity recognition on the product keyword group to obtain product feature entities; Based on the product feature entity, keywords are extracted from the product information to obtain the product features.

7. The method according to any one of claims 1 to 5, characterized in that: The text embedding of the user information text based on a preset text feature extraction model to obtain a text embedding vector includes: Performing text preprocessing on the user information text to obtain a preprocessed text; Based on the preprocessed text, construct a text vocabulary; Encoding the text vocabulary based on the text feature extraction model to obtain the phrase embedding vector; Perform vector concatenation on the phrase embedding vectors to obtain the text embedding vector.

8. A product recommendation device, characterized in that: The device comprises: An information acquisition module is used to obtain basic user information and user self-description information of the target user, and obtain product information of the candidate products; An information synthesis module, used for merging the user basic information and the user self-description information to obtain a user information text; A text embedding module, used for embedding the user information text based on a preset text feature extraction model to obtain a text embedding vector, wherein the text embedding vector includes a phrase embedding vector; A semantic understanding module, used to perform semantic understanding on the text embedding vector based on the text feature extraction model to obtain a contextual representation of the phrase embedding vector; A text feature extraction module, used to extract features from the text embedding vector based on the context representation to obtain target user information features; A product feature extraction module, used to extract features from the product information based on a preset product feature extraction model to obtain product features; A product screening module, configured to perform feature matching on the product features based on the target user information features to obtain target features, and screen the target product from the candidate products based on the target features; A product recommendation module is used to recommend products to the target user based on the target product.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the product recommendation method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the product recommendation method according to any one of claims 1 to 7 is implemented.

Images