Product recommendation method, product recommendation device, electronic device, and storage medium

Abstract

The embodiment of the application provides a product recommendation method, a product recommendation device, an electronic device and a storage medium, and belongs to the field of financial technology. The method comprises the following steps: acquiring audio data and virtual behavior data generated by a target object in a virtual conversation scene; performing content analysis on the audio data to obtain semantic representation data; performing decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data; performing portrait construction based on the semantic representation data and the decision preference data to obtain an object portrait of the target object; performing preference theme identification on the target object based on the semantic representation data to obtain an object preference theme; and performing product recommendation on the target object based on a preset knowledge graph, the object preference theme and the object portrait. The embodiment of the application can improve the accuracy of product recommendation.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more particularly to a product recommendation method, a product recommendation device, an electronic device, and a storage medium. Background Technology

[0002] With the development of smart finance, customers' demand for financial products such as investment, wealth management, and insurance is increasing. However, in the information explosion of the internet age, there is a vast amount of information and data, while customers' time is often fragmented, making it difficult for them to quickly find relevant investment information of interest from the large amount of information.

[0003] Currently, most product recommendation methods rely on factors such as trending topics and novelty when recommending financial products. These methods often lack in-depth analysis and modeling specific to the financial investment and wealth management field, frequently overlooking the actual needs of customers and resulting in low accuracy in product recommendations. Summary of the Invention

[0004] The main objective of this application is to provide a product recommendation method, a product recommendation device, an electronic device, and a storage medium, with the aim of improving the accuracy of product recommendations.

[0005] To achieve the above objectives, a first aspect of this application proposes a product recommendation method, the method comprising:

[0006] Acquire audio data and virtual behavior data generated by the target object in a virtual session scenario;

[0007] Content analysis is performed on the audio data to obtain semantic representation data;

[0008] Based on the virtual behavior data, a decision preference analysis is performed on the target object to obtain decision preference data;

[0009] Based on the semantic representation data and the decision preference data, a profile is constructed to obtain the object profile of the target object;

[0010] Based on the semantic representation data, the target object's preference topics are identified to obtain the object's preference topics;

[0011] Based on a pre-defined knowledge graph, the object's preferred topics, and the object's profile, product recommendations are made to the target object.

[0012] In some embodiments, the content analysis of the audio data to obtain semantic representation data includes:

[0013] The audio data is subjected to speech recognition to obtain text data;

[0014] Semantic analysis is performed on the text data to obtain the semantic representation data.

[0015] In some embodiments, the virtual behavior data includes decision data for each virtual behavior, and the step of performing decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data includes:

[0016] For each virtual behavior, the baseline features and behavior weights of the virtual behavior are obtained, and the temporal features of the virtual behavior are extracted from the decision data;

[0017] For each virtual behavior, a deviation analysis is performed on the temporal features and the baseline features based on the behavior weights to obtain decision behavior deviation data;

[0018] The decision-making behavior deviation data of all virtual behaviors are integrated and processed to obtain the total behavior deviation data;

[0019] The decision preference data is obtained by integrating the total behavioral deviation data.

[0020] In some embodiments, the step of identifying preference topics for the target object based on the semantic representation data to obtain object preference topics includes:

[0021] Keyword extraction is performed on the semantic representation data to obtain target keywords;

[0022] Calculate the similarity between the target keywords and the preset topics to obtain the topic matching degree;

[0023] The preferred topics of the object are filtered from the preset topics based on the topic matching degree.

[0024] In some embodiments, the knowledge graph is used to reflect the degree of correlation between object profiles, object preference topics, and candidate products; the step of recommending products to the target object based on the preset knowledge graph, the object preference topics, and the object profile includes:

[0025] Traverse the knowledge graph to obtain the first association score between each candidate product and the object's preference topic;

[0026] Traverse the knowledge graph to obtain a second association score between each candidate product and the object profile;

[0027] Based on the first association score and the second association score, the target product is selected from the multiple candidate products;

[0028] The target product is recommended to the target audience.

[0029] In some embodiments, selecting the target product from a plurality of candidate products based on the first association score and the second association score includes:

[0030] The first association score and the second association score are weighted and summed to obtain the total association score for each candidate product.

[0031] Based on the total correlation score, a predetermined number of target products are selected from the multiple candidate products.

[0032] In some embodiments, after recommending the target product to the target object, the method further includes:

[0033] Obtain feedback data from the target object regarding the target product;

[0034] Based on the feedback data, at least one of the object profile and the object preference topics is updated.

[0035] To achieve the above objectives, a second aspect of this application provides a product recommendation device, the device comprising:

[0036] The data acquisition module is used to acquire audio data and virtual behavior data generated by the target object in the virtual session scenario;

[0037] The content analysis module is used to perform content analysis on the audio data to obtain semantic representation data;

[0038] The decision preference analysis module is used to perform decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data.

[0039] The profile building module is used to build a profile based on the semantic representation data and the decision preference data to obtain an object profile of the target object.

[0040] The topic recognition module is used to identify the preferred topics of the target object based on the semantic representation data, so as to obtain the object's preferred topics.

[0041] The recommendation module is used to recommend products to the target object based on a preset knowledge graph, the object's preferred topics, and the object's profile.

[0042] To achieve the above objectives, a third aspect of the present application provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method described in the first aspect.

[0043] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0044] The product recommendation method, device, electronic equipment, and storage medium proposed in this application acquire audio data and virtual behavior data generated by a target object in a virtual session scenario. Content analysis of the audio data yields semantic representation data, enabling relatively accurate determination of the semantic content of the audio data. Furthermore, decision preference analysis of the target object is performed based on the virtual behavior data to obtain decision preference data, facilitating the determination of the target object's decision preferences. Further, a profile is constructed based on the semantic representation data and decision preference data to obtain an object profile of the target object; preference topic identification is performed on the target object based on the semantic representation data to obtain object preference topics; and product recommendations are made to the target object based on a pre-set knowledge graph, object preference topics, and object profile. This approach utilizes the knowledge graph and combines the target object's object profile and preference topics to recommend products, making the recommended products more aligned with the target object's habits, preferences, and actual needs, thus improving the rationality and accuracy of product recommendations. Attached Figure Description

[0045] Figure 1 This is a flowchart of the product recommendation method provided in the embodiments of this application;

[0046] Figure 2 yes Figure 1 The flowchart of step S102 in the document;

[0047] Figure 3 yes Figure 1 The flowchart of step S103 in the process;

[0048] Figure 4 yes Figure 1 The flowchart of step S105 in the process;

[0049] Figure 5 yes Figure 1 The flowchart of step S106 in the process;

[0050] Figure 6 yes Figure 5 The flowchart of step S503 in the process;

[0051] Figure 7 This is another flowchart of the product recommendation method provided in the embodiments of this application;

[0052] Figure 8 This is a schematic diagram of the product recommendation device provided in the embodiments of this application;

[0053] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] It should be noted that although functional modules are divided in the device schematic diagram and a 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 aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0057] First, let's analyze some of the terms used in this application:

[0058] Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0059] Natural Language Processing (NLP): NLP uses computers to process, understand, and utilize human language (such as Chinese and English). NLP is a branch of artificial intelligence and an interdisciplinary field of computer science and linguistics, often referred to as computational linguistics. NLP includes syntactic analysis, semantic analysis, and discourse understanding. It is commonly used in machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, intent recognition, information extraction and filtering, text classification and clustering, sentiment analysis, and opinion mining. It involves data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computation.

[0060] Information Extraction (NER) is a text processing technique that extracts factual information such as entities, relationships, and events from natural language text and outputs it as structured data. Information extraction is a technique for extracting specific information from text data. Text data is composed of specific units, such as sentences, paragraphs, and chapters. Text information is composed of smaller, specific units, such as characters, words, phrases, sentences, paragraphs, or combinations of these units. Extracting noun phrases, names of people, and place names from text data is an example of text information extraction. Of course, text information extraction techniques can extract information of various types.

[0061] With the development of smart finance, customers' demand for financial products such as investment, wealth management, and insurance is increasing. However, in the information explosion of the internet age, there is a vast amount of information and data, while customers' time is often fragmented, making it difficult for them to quickly find relevant investment information of interest from the large amount of information.

[0062] Currently, most product recommendation methods rely on factors such as trending topics and novelty when recommending financial products. These methods often lack in-depth analysis and modeling specific to the financial investment and wealth management field, frequently overlooking the actual needs of customers and resulting in low accuracy in product recommendations.

[0063] Based on this, embodiments of this application provide a product recommendation method, a product recommendation device, an electronic device, and a storage medium, aiming to improve the accuracy of product recommendations.

[0064] The product recommendation method, apparatus, electronic device, and storage medium provided in this application are specifically described through the following embodiments. First, the product recommendation method in this application is described.

[0065] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0066] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0067] The product recommendation method provided in this application relates to the field of fintech. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the product recommendation method, but is not limited to the above forms.

[0068] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: 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, and distributed computing environments including any of the above systems or devices. This 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. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0069] It should be noted that in various specific embodiments of this application, when processing data related to the identity or characteristics of an object, such as object information, object behavior data, object historical data, and object location information, the object's permission or consent will be obtained first. Furthermore, the collection, use, and processing of this data will comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require obtaining the object's personal information, separate permission or consent from the object will be obtained through pop-ups or redirection to a confirmation page. Only after obtaining the object's separate permission or consent will the necessary object-related data for the proper functioning of the embodiments of this application be obtained.

[0070] Figure 1 This is an optional flowchart of the product recommendation method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S106.

[0071] Step S101: Obtain audio data and virtual behavior data generated by the target object in the virtual session scenario;

[0072] Step S102: Perform content analysis on the audio data to obtain semantic representation data;

[0073] Step S103: Analyze the decision preferences of the target object based on the virtual behavior data to obtain decision preference data;

[0074] Step S104: Based on semantic representation data and decision preference data, a profile is constructed to obtain an object profile of the target object;

[0075] Step S105: Based on semantic representation data, perform preference topic identification on the target object to obtain the object's preference topic;

[0076] Step S106: Recommend products to the target object based on the preset knowledge graph, object preference topics, and object profile.

[0077] Steps S101 to S106 of this embodiment involve acquiring audio data and virtual behavior data generated by the target object in a virtual session scenario; performing content analysis on the audio data to obtain semantic representation data, which can accurately determine the semantic content of the audio data. Further, based on the virtual behavior data, decision preference analysis is performed on the target object to obtain decision preference data, which can conveniently determine the target object's decision preferences. Further, based on the semantic representation data and decision preference data, a profile is constructed to obtain the target object's profile; based on the semantic representation data, preference topics are identified for the target object to obtain the object's preference topics; based on a preset knowledge graph, object preference topics, and object profile, product recommendations are made to the target object. This approach utilizes the knowledge graph and combines the target object's profile and preference topics to recommend products to the target object, making the recommended products more aligned with the target object's habits, preferences, and actual needs, thus improving the rationality and accuracy of product recommendations.

[0078] In step S101 of some embodiments, the virtual session scenario refers to the simulated trading experience scenario provided by various financial trading institutions. During the simulated trading process, the target object can interact with the trading agent in the virtual session scenario through various interactive methods such as voice and text, thereby generating audio data or text data. At the same time, the target object can also make various types of trading decisions during the simulated trading process to achieve a more realistic trading experience. The trading agent can be relevant personnel from various financial trading institutions in the real world, or it can be a virtual character set up by various financial trading institutions in the simulated trading scenario; there are no restrictions.

[0079] Since all relevant data in various virtual session scenarios is stored on the server in the form of logs, audio data and virtual behavior data generated by the target object in the virtual session scenario can be obtained from the corresponding background logs. Audio data refers to the data generated by the target object communicating with the transaction agent in the virtual session scenario, while virtual behavior data refers to the decision data generated by the target object performing various virtual behaviors in the virtual session scenario.

[0080] It should be noted that in practical applications, the decision-making time and the number of decision-related parameters for the target object depend on the number of virtual session scenarios provided by the financial trading institution. The more virtual session scenarios there are, the longer the decision-making time and the more decision-related content the target object will have.

[0081] Simulating transactions through virtual session scenarios can reduce the consumption of real resources and intelligently obtain the target audience's trading habits and preferences, enabling more accurate product recommendations and improving the target audience's trading experience.

[0082] Please see Figure 2 In some embodiments, step S102 may include, but is not limited to, steps S201 to S202:

[0083] Step S201: Perform speech recognition on the audio data to obtain text data;

[0084] Step S202: Perform semantic analysis on the text data to obtain semantic representation data.

[0085] Steps S201 to S201 are described in detail below.

[0086] In step S201 of some embodiments, speech recognition can be performed on audio data using ASR technology. Specifically, the audio data is first input into a speech recognition model based on ASR technology. The speech recognition model includes a fully connected module, a bidirectional RNN layer, and a fully connected layer connected in sequence, wherein the fully connected module includes three fully connected layers connected in sequence. After the audio data is input into the speech recognition model, the fully connected module extracts the audio frame features of the audio data in the time dimension. Next, the bidirectional RNN layer extracts the content of the audio frame features to obtain the text features of each frame; finally, the softmax function of the fully connected layer calculates the probability of the text feature belonging to each character, and selects the character with the highest probability as the character corresponding to the text feature. All strings are concatenated in chronological order to obtain the text data corresponding to the audio data.

[0087] In step S202 of some embodiments, the text data is first encoded to obtain text encoding features. Then, a transformer-based pre-trained model is used to extract word-level semantic information from the text encoding features. Subsequently, a transformer-based self-attention mechanism is used to fuse the word-level semantic information to obtain fused semantic features. Finally, a decoder is used to decode the fused semantic features to obtain semantic representation data, where the semantic representation data is the result of semantic understanding of the audio data.

[0088] Through the aforementioned steps S201 to S202, the character data corresponding to each audio frame can be identified relatively accurately. The character data is then concatenated in chronological order to obtain text data, thus improving the accuracy and comprehensiveness of the text data. Furthermore, semantic analysis of the text data allows for the convenient determination of the semantic information to be represented by the audio data, enabling a more accurate semantic understanding of the audio data and obtaining semantic representation data.

[0089] Please see Figure 3In some embodiments, step S103 may include, but is not limited to, steps S301 to S304:

[0090] Step S301: For each virtual behavior, obtain the baseline features and behavior weights of the virtual behavior, and extract the temporal features of the virtual behavior from the decision data;

[0091] Step S302: For each virtual behavior, perform deviation analysis on the temporal features and baseline features based on the behavior weight to obtain decision behavior deviation data;

[0092] Step S303: Integrate and process the decision behavior deviation data of all virtual behaviors to obtain the total behavior deviation data;

[0093] Step S304: Integrate the total behavioral deviation data to obtain decision preference data.

[0094] Steps S301 to S304 are described in detail below.

[0095] Virtual behavior data includes decision data for each virtual behavior. In a virtual session scenario, various types of choices are provided for the target object to select from. The target object's behavior in choosing different choices is called virtual behavior, and the decision data for virtual behavior refers to the judgments made by the target object on different choices.

[0096] For example, in some virtual conversation scenarios, various opportunities such as stocks, funds, securities, wealth management, insurance, and real estate are offered to the target audience for selection. When the target audience chooses real estate, this choice is considered a virtual action, and the amount invested in real estate by the target audience constitutes the decision data for this virtual action.

[0097] In step S301 of some embodiments, for each virtual behavior, some calibration data is set in the virtual session scenario. This calibration data is used to measure the deviation of the decision data of the virtual behavior of the target object. The calibration data includes baseline features and behavior weights of the virtual behavior. The baseline features refer to the decision standard data of each virtual behavior, and the behavior weights are the importance of this virtual behavior to the prediction of the target object's decision preference. Since the calibration data in each virtual session scenario is pre-set, the baseline features and behavior weights of the virtual behavior can be directly read.

[0098] Furthermore, since the target object will generate various decision-making operations in the virtual session scenario, decision data is obtained based on the data corresponding to the target object's decision-making operations on the interface. Then, the behavioral content in the decision data is extracted in the time dimension to obtain the temporal characteristics of the virtual behavior.

[0099] In step S302 of some embodiments, when performing deviation analysis on temporal features and baseline features based on behavior weights for each virtual behavior, the temporal features and baseline features are first divided to obtain a division result, which represents the magnitude of the deviation between the temporal features and the baseline features. Next, the division result is multiplied by the behavior weights to obtain a product, which represents the impact of the decision deviation of this virtual behavior on the overall decision preference. Finally, the product is squared to obtain the decision behavior deviation data.

[0100] In step S303 of some embodiments, the decision behavior deviation data of all virtual behaviors are summed using a summation function in statistics to integrate the decision deviations of all virtual behaviors of the target object in the virtual session scene and obtain the total behavior deviation data.

[0101] In step S304 of some embodiments, statistical methods are used to perform integral calculations on the total behavioral deviation data, thereby calculating the integral value of the total behavioral deviation data over time, and using the obtained integral value as decision preference data.

[0102] For example, in one specific embodiment, the calculation process of decision preference data can be represented as shown in formula (1):

[0103]

[0104] Where f(x) refers to decision preference data, n refers to the nth virtual behavior, m refers to the total number of virtual behaviors, m is a positive integer greater than 0, and n is an integer greater than 0 and less than or equal to m. n This refers to the behavior weight of the nth virtual behavior, b n b refers to the temporal characteristics in the decision data of the nth virtual behavior, and b refers to the baseline characteristics of the nth virtual behavior.

[0105] Through the above steps S301 to S304, all virtual behavior decision data can be used to predict the decision preferences of the target object. At the same time, for the various virtual behaviors of the target object in the entire virtual session scenario, various statistical mathematical operations are used to quantify the decision preferences, which can more clearly reflect the decision preferences of the target object and improve the accuracy of decision preference judgment.

[0106] In step S104 of some embodiments, when constructing a profile based on semantic representation data and decision preference data to obtain an object profile of the target object, the semantic representation data is first extracted using a named entity recognition algorithm to obtain target keywords; then, the target keywords and decision preference data are integrated to obtain the object profile.

[0107] For example, named entity recognition algorithms can be used to extract keywords from semantic representation data, yielding keywords related to investment risk and investment returns for the target object. Based on decision-making preference data, the risk tolerance of the target object can be assessed, along with its preferences for various financial instruments and resource allocation habits. Keywords related to investment risk include health and accidents, while keywords related to investment returns include profit and loss.

[0108] Please see Figure 4 In some embodiments, step S105 may include, but is not limited to, steps S401 to S403:

[0109] Step S401: Extract keywords from the semantic representation data to obtain target keywords;

[0110] Step S402: Calculate the similarity between the target keywords and the preset topics to obtain the topic matching degree;

[0111] Step S403: Filter the object's preferred topics from the preset topics based on topic matching degree.

[0112] Steps S401 to S403 are described in detail below.

[0113] In step S401 of some embodiments, when extracting keywords from semantic representation data, the semantic representation data is first mapped from the semantic space to the vector space to obtain a semantic representation vector. Then, a named entity recognition algorithm is used to identify entity features in the semantic representation vector, extracting entity features from the semantic representation data. Further, based on the matching relationship between entity features and words in a preset dictionary, the word with the highest matching degree with the entity features in the preset dictionary is selected as the target keyword. The matching degree can be calculated based on Euclidean distance, Manhattan distance, etc., and is not limited thereto.

[0114] In step S402 of some embodiments, a cosine similarity algorithm can be used to calculate the similarity between the target keyword and the preset topic. Specifically, the preset topic and the target keyword are first vectorized, and then the cosine similarity algorithm is used to calculate the similarity between the vectorized preset topic and the target keyword. The calculated similarity is used as the topic matching degree. The higher the topic matching degree, the higher the probability that the target keyword belongs to the preset topic, and the more the specific content of the semantic representation data matches the preset topic.

[0115] In step S403 of some embodiments, a preset topic with the highest topic matching degree with the target keyword can be directly selected as the object's preferred topic. Further, to improve topic diversity, preset topics with a topic matching degree greater than a preset matching threshold with the target keyword can also be selected as the object's preferred topics. Alternatively, preset topics with a predetermined ranking of topic matching degree with the target keyword can be selected as the object's preferred topics, without restriction. The preset topics can be categorized according to product type, including insurance, securities, wealth management, or stocks. Preset topics can also be categorized according to business type, including risk, return, etc.

[0116] Through the above steps S401 to S403, the named entity recognition algorithm can be used to obtain the target keywords in the semantic representation data. Furthermore, the similarity of the target keywords with the preset topics is compared, thereby selecting the preset topics with higher similarity as the object preference topics of the target object, which can improve the prediction accuracy of the object preference topics.

[0117] Please see Figure 5 In some embodiments, step S106 may include, but is not limited to, steps S501 to S504:

[0118] Step S501: Traverse the knowledge graph and obtain the first association score between each candidate product and the object preference topic;

[0119] Step S502: Traverse the knowledge graph and obtain the second association score between each candidate product and the object profile;

[0120] Step S503: Based on the first correlation score and the second correlation score, select the target product from multiple candidate products;

[0121] Step S504: Recommend the target product to the target audience.

[0122] Steps S501 to S504 are described in detail below.

[0123] Knowledge graphs are used to reflect the degree of association between object profiles, object preference topics, and candidate products. Essentially, a knowledge graph is an entity-relationship model, where different object profiles, object preference topics, and candidate products exhibit varying probabilistic relationships of strong and weak associations.

[0124] Specifically, in a knowledge graph, object profiles, object preference topics, and candidate products are nodes. When an object profile and a candidate product are related, an edge exists between the node corresponding to the object profile and the node corresponding to the candidate product; the association score between the candidate product and the object profile is the edge weight. Similarly, when an object preference topic and a candidate product are related, an edge exists between the node corresponding to the preference topic and the node corresponding to the candidate product; the association score between the candidate product and the preference topic is the edge weight. Candidate products are common products in financial transactions, including but not limited to insurance products, wealth management products, stock products, and securities products.

[0125] In step S501 of some embodiments, the knowledge graph is traversed to find the node corresponding to the object preference topic, and each edge determined based on the node corresponding to the object preference topic is queried. Based on each edge, candidate products that are associated with the node corresponding to the object preference topic are determined, and the edge weight of each edge is used as the first association score between each candidate product and the object preference topic.

[0126] In step S502 of some embodiments, the knowledge graph is traversed to find the node corresponding to the object profile, and each edge determined based on the node corresponding to the object profile is queried. Based on each edge, candidate products that are associated with the node corresponding to the object profile are determined, and the edge weight of each edge is used as the second association score between each candidate product and the object profile.

[0127] In some specific embodiments, a target product is selected from multiple candidate products based on a first association score. Specifically, candidate products can be sorted in descending order according to their first association scores, and the candidate products ranked at the top of the predetermined order are selected as the target products. This approach can select candidate products with higher relevance to the target audience based on the degree of association between the candidate products and the audience's preferred topics, thereby improving the rationality of product recommendations.

[0128] In some specific embodiments, a target product is selected from multiple candidate products based on a second association score. Specifically, candidate products can be sorted in descending order according to their second association scores, and the candidate products ranked at the top of the predetermined order are selected as the target products. This approach can select candidate products with higher relevance to the target user based on the degree of association between the candidate products and the user profile, thereby improving the rationality of product recommendations.

[0129] In step S503 of some embodiments, firstly, the first association score and the second association score are weighted and summed to obtain the total association score for each candidate product. Then, based on the total association score, a predetermined number of target products are selected from the multiple candidate products.

[0130] In step S504 of some embodiments, the target product is recommended to the target object in the form of video, audio or images, so that the target object can easily understand the relevant information of the target product and select the target product of interest for investment and transaction.

[0131] Through steps S501 to S504, the correlation between candidate products and object profiles, as well as the correlation between candidate products and object preference topics in the knowledge graph, can be used to accurately determine the degree of correlation between each candidate product and the target object. Furthermore, by using scores to quantify the degree of correlation between each candidate product and the object profile and object preference topics of the target object, it is possible to more accurately select target products that are more suitable for the target object from multiple candidate products, thereby recommending the target products to the target object and improving the accuracy and rationality of product recommendations.

[0132] Please see Figure 6 In some embodiments, step S503 includes, but is not limited to, steps S601 to S602:

[0133] Step S601: The first association score and the second association score are weighted and summed to obtain the total association score for each candidate product;

[0134] Step S602: Based on the total correlation score, select a predetermined number of target products from multiple candidate products.

[0135] Steps S601 to S602 are described in detail below.

[0136] In step S601 of some embodiments, firstly, a first weight and a second weight are set. The first weight is used to characterize the importance of the association score between the candidate product and the object's preference topic in product screening, and the second weight is used to characterize the importance of the association score between the candidate product and the object's profile in product screening. Next, for each candidate product, the first weight and the first association score are multiplied to obtain a first weighted association score; the second weight and the second association score are multiplied to obtain a second weighted association score. Finally, the first weighted association score and the second weighted association score of the candidate product are added together to obtain the total association score of the candidate product.

[0137] In step S602 of some embodiments, the predetermined number can be determined according to actual circumstances and is not limited. Specifically, the methods for selecting a predetermined number of target products from multiple candidate products based on the total association score include, but are not limited to, the following:

[0138] Method 1: Select products with a total correlation score higher than a preset score threshold from multiple candidate products as preliminary products. Then, select a predetermined number of products from the preliminary products as target products.

[0139] Method 2: Based on the total score, sort all candidate products in descending order and select the top-ranked candidate products as preliminary products. Then, select a predetermined number of target products from the preliminary products.

[0140] Method 3: From multiple candidate products, select those with a total correlation score higher than a preset score threshold as preliminary products. Next, based on the total correlation score, sort all preliminary products in descending order and select the top-ranked preliminary products as intermediate products. Finally, select a predetermined number of intermediate products as target products. Target products include, but are not limited to, common products in the financial trading field such as insurance products, wealth management products, stock products, and securities products.

[0141] Through steps S601 to S602, the first association score between candidate products and the target audience's preference topics, and the second association score between candidate products and the target audience's profile, can be used to comprehensively evaluate the degree of association between each candidate product and the target audience. Furthermore, based on the total association score of each candidate product, target products are selected from the candidate products using methods such as threshold comparison and score ranking. This allows for better control of the number of target products and also enables the recommendation of candidate products with a higher degree of association with the target audience, improving the accuracy and rationality of product recommendations.

[0142] Please see Figure 7 In some embodiments, after step S504, the product recommendation method may also include, but is not limited to, steps S701 to S702:

[0143] Step S701: Obtain feedback data from the target object regarding the target product;

[0144] Step S702: Update at least one of the object profile and object preference topics based on the feedback data.

[0145] Steps S701 to S702 are described in detail below.

[0146] In step S701 of some embodiments, since the target object's level of interest in the target product may vary—that is, some target products are what the target object truly needs and is interested in, while others are not needed—there is a recommendation bias. Therefore, after recommending the target product to the target object, it is also necessary to collect various behavioral operations of the target object towards the target product, and use the data corresponding to these behavioral operations as feedback data from the target object. These behavioral operations include deleting, closing, adding to favorites, purchasing, sharing, etc.

[0147] For example, when a target audience adds a product to their favorites, purchases, or shares it, it indicates that the product is of interest to them. Conversely, when a target audience deletes or closes a product, it indicates that the product is of no interest to them.

[0148] In step S702 of some embodiments, based on the different behavioral operations of the target object on different target products in the feedback data, at least one of the object profile and the object preference topic is updated. If the target product is a product that the target object is interested in, more relevant content about the target product is added to the object preference topic and the object profile; if the target product is a product that the target object is not interested in, relevant content about the target product is deleted from the object preference topic and the object profile.

[0149] For example, if the target product is not of interest to the target audience, then the topic to which the target product belongs is removed from the audience's preference topics.

[0150] Through the above steps S701 to S702, the object profile and object preference topics can be dynamically updated based on the feedback data of the target object on the target product, making the object profile and object preference topics more accurate, which helps to improve the accuracy of subsequent product recommendations.

[0151] The product recommendation method of this application embodiment acquires audio data and virtual behavior data generated by a target object in a virtual session scenario; performs content analysis on the audio data to obtain semantic representation data, which can accurately determine the semantic content of the audio data. Further, it analyzes the target object's decision preferences based on the virtual behavior data to obtain decision preference data, which can conveniently determine the target object's decision preferences. Further, it constructs a profile based on the semantic representation data and decision preference data to obtain an object profile of the target object; it identifies the target object's preference topics based on the semantic representation data to obtain object preference topics; and it recommends products to the target object based on a preset knowledge graph, object preference topics, and object profile. This method utilizes the knowledge graph and combines the target object's object profile and object preference topics to recommend products to the target object, making the recommended products more aligned with the target object's habits, preferences, and actual needs, thus improving the rationality and accuracy of product recommendations.

[0152] Please see Figure 8 This application also provides a product recommendation device that can implement the above-described product recommendation method. The device includes:

[0153] Data acquisition module 801 is used to acquire audio data and virtual behavior data generated by the target object in the virtual session scenario;

[0154] Content analysis module 802 is used to perform content analysis on audio data to obtain semantic representation data;

[0155] The decision preference analysis module 803 is used to perform decision preference analysis on the target object based on virtual behavior data to obtain decision preference data.

[0156] The profile building module 804 is used to build profiles based on semantic representation data and decision preference data to obtain object profiles of target objects.

[0157] The topic recognition module 805 is used to identify the preferred topics of the target object based on semantic representation data, and obtain the object's preferred topics.

[0158] The recommendation module 806 is used to recommend products to target objects based on preset knowledge graphs, object preference topics, and object profiles.

[0159] The specific implementation of the product recommendation device is basically the same as the specific implementation of the product recommendation method described above, and will not be repeated here.

[0160] This application also provides an electronic device, which includes: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for communication between the processor and the memory. When the program is executed by the processor, it implements the product recommendation method described above. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0161] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0162] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, 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 this application.

[0163] The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 using the product recommendation method of the embodiments of this application.

[0164] The input / output interface 903 is used to implement information input and output;

[0165] The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0166] Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904);

[0167] The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0168] This application also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the product recommendation method described above.

[0169] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0170] The product recommendation method, device, electronic device, and computer-readable storage medium provided in this application acquire audio data and virtual behavior data generated by a target object in a virtual session scenario; perform content analysis on the audio data to obtain semantic representation data, which can accurately determine the semantic content of the audio data. Further, perform decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data, which can conveniently determine the target object's decision preferences. Further, construct a profile based on the semantic representation data and decision preference data to obtain an object profile of the target object; identify preference topics of the target object based on the semantic representation data to obtain object preference topics; and recommend products to the target object based on a preset knowledge graph, object preference topics, and object profile. This approach utilizes the knowledge graph and combines the target object's object profile and object preference topics to recommend products to the target object, making the recommended products more aligned with the target object's habits, preferences, and actual needs, thus improving the rationality and accuracy of product recommendations.

[0171] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0172] It will be understood by those skilled in the art that Figure 1-7 The technical solutions shown do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0173] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0174] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0175] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0176] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0177] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0178] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0180] If the integrated unit is implemented as 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 this application, in essence, 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. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0181] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A product recommendation method, characterized in that, The method includes: Acquire audio data and virtual behavior data generated by the target object in a virtual session scenario; Content analysis is performed on the audio data to obtain semantic representation data; Based on the virtual behavior data, a decision preference analysis is performed on the target object to obtain decision preference data; Based on the semantic representation data and the decision preference data, a profile is constructed to obtain the object profile of the target object; Based on the semantic representation data, the target object's preference topics are identified to obtain the object's preference topics; Based on a preset knowledge graph, the object's preferred topics, and the object's profile, product recommendations are made to the target object. The virtual behavior data includes decision data for each virtual behavior. The step of performing decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data includes: For each virtual behavior, the baseline features and behavior weights of the virtual behavior are obtained, and the temporal features of the virtual behavior are extracted from the decision data; For each virtual behavior, the temporal features and the baseline features are divided to obtain the division result. The division result is then multiplied by the behavior weight to obtain the product result. The product result is then squared to obtain the decision behavior bias data. The decision-making behavior deviation data of all virtual behaviors are integrated and processed to obtain the total behavior deviation data; The decision preference data is obtained by integrating the total behavioral deviation data.

2. The product recommendation method according to claim 1, characterized in that, The content analysis of the audio data to obtain semantic representation data includes: The audio data is subjected to speech recognition to obtain text data; Semantic analysis is performed on the text data to obtain the semantic representation data.

3. The product recommendation method according to claim 1, characterized in that, The step of identifying the target object's preference topics based on the semantic representation data to obtain the object's preference topics includes: Keyword extraction is performed on the semantic representation data to obtain target keywords; Calculate the similarity between the target keywords and the preset topics to obtain the topic matching degree; The preferred topics of the object are filtered from the preset topics based on the topic matching degree.

4. The product recommendation method according to any one of claims 1 to 3, characterized in that, The knowledge graph is used to reflect the degree of correlation between object profiles, object preference topics, and candidate products; the step of recommending products to the target object based on the preset knowledge graph, the object preference topics, and the object profile includes: Traverse the knowledge graph to obtain the first association score between each candidate product and the object's preference topic; Traverse the knowledge graph to obtain a second association score between each candidate product and the object profile; Based on the first association score and the second association score, a target product is selected from the multiple candidate products; The target product is recommended to the target audience.

5. The product recommendation method according to claim 4, characterized in that, The step of selecting the target product from multiple candidate products based on the first association score and the second association score includes: The first association score and the second association score are weighted and summed to obtain the total association score for each candidate product. Based on the total correlation score, a predetermined number of target products are selected from the multiple candidate products.

6. The product recommendation method according to claim 4, characterized in that, After recommending the target product to the target object, the method further includes: Obtain feedback data from the target object regarding the target product; Based on the feedback data, at least one of the object profile and the object preference topics is updated.

7. A product recommendation device, characterized in that, The device includes: The data acquisition module is used to acquire audio data and virtual behavior data generated by the target object in the virtual session scenario; The content analysis module is used to perform content analysis on the audio data to obtain semantic representation data; The decision preference analysis module is used to perform decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data. The profile building module is used to build a profile based on the semantic representation data and the decision preference data to obtain an object profile of the target object. The topic recognition module is used to identify the preferred topics of the target object based on the semantic representation data, so as to obtain the object's preferred topics. The recommendation module is used to recommend products to the target object based on a preset knowledge graph, the object's preferred topics, and the object's profile. The virtual behavior data includes decision data for each virtual behavior. The step of performing decision preference analysis on the target object based on the virtual behavior data to obtain decision preference data includes: For each virtual behavior, the baseline features and behavior weights of the virtual behavior are obtained, and the temporal features of the virtual behavior are extracted from the decision data; For each virtual behavior, the temporal features and the baseline features are divided to obtain the division result. The division result is then multiplied by the behavior weight to obtain the product result. The product result is then squared to obtain the decision behavior bias data. The decision-making behavior deviation data of all virtual behaviors are integrated and processed to obtain the total behavior deviation data; The decision preference data is obtained by integrating the total behavioral deviation data.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the product recommendation method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the product recommendation method according to any one of claims 1 to 6.

Images