Information interaction method and electronic equipment

Abstract

The embodiment of the invention discloses an information interaction method and electronic equipment, and the method comprises the steps: calling a first artificial intelligence AI model according to the first behavior data of a user in a process of displaying commodity information in a target page, so as to carry out the reasoning of the current purchase intention type of the user; if the purchase intention type is that the purchase intention type has a category intention for a target commodity category and does not have a deterministic single commodity intention, calling a second AI model according to second behavior data, related to the target commodity category, of the user so as to deduce cognitive information of the user for the target commodity category; and if the deterministic single item intention is still not reasoned, calling a third AI model according to third behavior data related to the target commodity category, so that the third AI model performs reasoning on the purchase demand of the user under the target commodity category. According to the embodiment of the invention, the effectiveness of the AI shopping guide information can be improved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to an information interaction method and an electronic device. Background Art

[0002] In product information service systems, product recommendations are a common shopping guide. However, traditional product recommendation information is typically presented passively. Even when integrated with algorithmic recommendation capabilities, consumers still need to continuously explore the products they need, and their access often involves permutations and combinations of products. Furthermore, recommendations are typically based on user preferences, but user shopping intentions change rapidly, resulting in a low browse-to-purchase conversion rate in product recommendation scenarios.

[0003] With the emergence and rapid development of AI (Artificial Intelligence) models, some product information service systems have also incorporated AI functionality. Users can now interact with AI to locate suitable products during the shopping process. For example, in one existing solution, users can enter a dedicated AI interactive interface and engage in multiple rounds of text conversations with the AI ​​to express their needs. The AI ​​then helps the user recommend relevant products. For example, after entering the AI ​​interactive interface, the user can express their budget, family composition, and car usage frequency through text, and then ask the AI ​​to recommend a car. The AI ​​may also ask the user follow-up questions based on the user's specific situation, and the user can answer to provide more information. After multiple rounds of interaction, the AI ​​can complete the recommendation of related products, and so on. This method can help users find the products they need, but it is generally suitable for scenarios with high decision-making costs, such as buying a car or a mobile phone, which requires a lot of upfront knowledge and input. It is not suitable for scenarios with lower decision-making costs, such as purchasing clothing or daily necessities. Summary of the Invention

[0004] This application provides an information interaction method and electronic device that can improve the effectiveness of AI shopping guide information.

[0005] This application provides the following solutions:

[0006] An information interaction method, comprising:

[0007] In the process of displaying product information on the target page, the first artificial intelligence (AI) model is called based on the user's first behavior data to infer the user's current purchase intention type;

[0008] If the purchase intention type is category intention for the target product category but no definite single product intention, then calling a second AI model based on the second behavioral data of the user related to the target product category to infer the user's cognitive information about the target product category, and determining whether definite single product intention can be inferred based on the cognitive information;

[0009] If the deterministic single product intention cannot be inferred, the third AI model is called based on the third behavioral data of the user related to the target product category to infer the user's purchasing needs under the target product category, and if the deterministic single product intention cannot be inferred based on the purchase demand inference result, the corresponding demand option is output so as to continue the purchase demand inference in combination with the user's selection result of the demand option.

[0010] Among them, the process of inferring the user's purchasing needs under the target product category is executed in a loop multiple times, and each time the purchasing needs are refined from one dimension. After receiving the user's selection result for the demand option, the third AI model is re-called according to the selection result and the third behavior data, and the step of refining the purchasing needs from another dimension is executed until a deterministic single product intention is inferred or the number of cycles reaches a threshold.

[0011] Among them, when the third AI model performs detailed reasoning on purchase needs each time it is called, it performs detailed reasoning on purchase needs according to the personalized link corresponding to the target product category; wherein, the personalized link includes the dimensional order corresponding to the target product category and the demand options corresponding to each dimension; the dimensional order is determined with the goal of inferring the user's intention for a single product under the target product category with fewer cycles.

[0012] Among them, the third AI model is composed of a basic model and multiple category sub-models. The multiple category sub-models are used to generate personalized link information for corresponding categories. The basic model is used to determine the dimensions selected each time the purchase demand refinement reasoning is performed and the corresponding demand options based on the personalized link information generated by the category sub-models.

[0013] Among them, also include:

[0014] If the deterministic single product intention can be inferred, the corresponding AI model is also used to generate product recommendation information based on the deterministic single product intention, or single product price comparison information, or to determine whether the single product has multiple minimum inventory unit SKU decision attributes. If so, the same product with different SKUs is generated for selection strategy information for display to users.

[0015] Among them, also include:

[0016] If after jumping from the target page to the details page of a certain product, the user returns to the target page, and it is inferred that the user has a definite single product intention, the information displayed in the resource position where the product was originally located in the target page will be replaced with the information generated by the corresponding AI model.

[0017] Among them, also include:

[0018] If the purchase intention type inferred by the first AI model is no clear purchase intention, candidate category options are generated according to regional differentiated operation strategies, weather strategies or season / season strategies, so that after receiving the user's selection result for one of the categories, it is determined that the user has category intention for the category, and the cognitive information judgment after identifying the category intention and the steps of reasoning according to the purchase needs are executed.

[0019] An information interaction device, comprising:

[0020] An intention type inference unit is used to call a first artificial intelligence (AI) model based on the user's first behavior data to infer the user's current purchase intention type during the process of displaying product information on the target page;

[0021] a category cognition inference unit, configured to, if the purchase intention type is category intention for a target product category but no definite individual product intention, invoke a second AI model based on the second behavioral data of the user related to the target product category to infer the user's cognition information about the target product category, and determine whether definite individual product intention can be inferred based on the cognition information;

[0022] The purchase demand inference unit is used to call the third AI model based on the third behavioral data of the user related to the target product category to infer the user's purchase demand under the target product category if the deterministic single product intention cannot be inferred, and output the corresponding demand option if the deterministic single product intention cannot be inferred based on the purchase demand inference result, so as to continue the purchase demand inference in combination with the user's selection result of the demand option.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the aforementioned methods.

[0024] An electronic device, comprising:

[0025] one or more processors; and

[0026] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of any of the aforementioned methods.

[0027] A computer program product comprises a computer program / computer executable instructions, wherein the computer program / computer executable instructions are capable of implementing the steps of any of the aforementioned methods when executed by a processor in an electronic device.

[0028] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0029] Through the embodiment of the present application, in the process of displaying product information on the target page, the first AI model can be first called according to the user's first behavior data, so that the first AI model can infer the user's current purchase intention type. If the inferred purchase intention type is a category intention for the target product category but no deterministic single product intention, the second AI model can be called according to the user's second behavior data related to the target product category, so that the second AI model can infer the user's cognitive information about the target product category, and determine whether it can infer a deterministic single product intention based on the cognitive information; if the deterministic single product intention still cannot be inferred, the third AI model can be further called according to the third behavior data related to the target product category, so that the third AI model can infer the user's purchase demand under the target product category, and if the deterministic single product intention cannot be inferred based on the purchase demand inference result, the corresponding demand option can be output so as to continue to infer the purchase demand inference in combination with the user's selection result of the demand option. In this way, a horizontal shopping guide capability that is independent of the traditional recommendation algorithm is provided, and by judging the user's current purchase intention, incremental value is brought to the system on the basis of traditional passive shopping guide. Moreover, since multiple rounds of interactions can be achieved between the main program and the AI ​​model and the user, when the AI ​​model cannot directly infer the definite intention of a single product through one round of interaction, it is not necessary to force the AI ​​model to give an inference result on the single product dimension. Instead, it can first determine whether there is a category intention. Then, through multiple rounds of interactions such as identifying the user's cognitive information about the category and inferring purchase needs, the AI ​​model can gradually obtain more valuable information. In addition, each round of interaction may have the possibility of inferring a clear intention of a single product. If not, the needs will be further refined, and combined with the demand option selection results fed back by the user, a more accurate judgment can be made. Therefore, the quality of the interaction results can be improved, and the effectiveness of proactive shopping guide information can be improved.

[0030] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 It is a schematic diagram of the system architecture provided by the embodiment of the present application;

[0033] Figure 2 is a flow chart of the method provided in an embodiment of the present application;

[0034] Figure 3 This is the first interface intention provided by the embodiment of the present application;

[0035] Figure 4 This is a schematic diagram of the second interface provided in an embodiment of the present application;

[0036] Figure 5 Schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

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

[0038] In an embodiment of the present application, in order to help users select products in a low-decision-making cost scenario, AI functions can be introduced into the product recommendation scenario. That is, in the process of recommending products to users through the client homepage of the product information service system, the channel page of certain channels, etc., the user's relevant behavioral data can be inferred and analyzed through the AI ​​model, and by actively interacting with the user (rather than passively waiting for the user to input questions), the user can be helped to locate the required products more quickly.

[0039] It's important to note that when AI models proactively help users make product recommendations, one approach involves feeding the AI ​​model with product information related to various user behavioral data (e.g., historical purchases, recent searches, recent browsing of product details, and recent cart additions). The AI ​​model then directly predicts the user's likely purchase intent (i.e., the user's purchase intent for a particular product or specific products). This information is then presented to the user via a pop-up window, from which the user can select a specific product and perform actions such as adding it to the cart. While this "single-round conversation" approach appears to be relatively efficient, it suffers from the problem that, in many cases, it may be impossible to accurately infer a user's product intent based solely on their behavioral data. However, since the AI ​​model is required to directly output the product, it can only complete the task as required. Even if it cannot accurately infer the user's likely purchase intent, it is still forced to output product information. Clearly, this interactive approach is difficult to guarantee and may even cause user dissatisfaction.

[0040] Of course, one way to improve this is to require the AI ​​model to self-check the accuracy of its inference results. If the accuracy reaches a certain threshold, it will output product information, otherwise it will not output it, so as not to cause user disgust. However, in reality, it is relatively rare to directly infer the products that users want to buy based on their behavioral data, and the accuracy must meet the requirements. This means that although the AI ​​model may be constantly performing inference and analysis in the background, the opportunity to actually interact with the user is very rare, resulting in a large amount of computing resources being wasted.

[0041] Based on the above situation, in the process of applying the AI ​​model to help users select goods, the embodiment of the present application does not directly require the AI ​​model to output clear inference results of the product granularity. Instead, when it is determined that the user does not have a clear intention for a single product, it can be determined whether the user has a category intention, that is, whether the user has a purchase intention for a certain product category (also referred to as "category intention"). For example, assuming that a user has been browsing various brands of paper towels in the past half hour, at this time, it may be determined that the user has a category intention for the paper towel category, and so on. If it is determined that the user has a category intention, then the AI ​​model can continue to be used to determine the user's cognitive information about the product category. Among them, the so-called cognitive information, that is, the user's degree of cognition of the product in the category reflected in the user's historical behavior data, this degree of cognition can also be a kind of preference information of the user when purchasing or adding the product category. Therefore, the "cognitive information" here can also be called the user's "shopping preference information" for the product category, and so on. For example, by collecting statistics on orders related to the category or add-to-cart purchases (add to cart) for the category in the past year (or other time period), it can be determined whether the user has a preference for a certain brand of products in the category (for example, assuming that for the paper towel category, most orders or add-to-cart products related to paper towels may be for the same brand), or whether the user always chooses large-sized paper towels, etc. If so, it can be inferred that the user may have a preference for the brand or large-sized paper towel products, that is, in the user's perception of the product category, the brand or the size is better or more suitable for their needs.

[0042] Based on the user's cognitive information about a specific product category, the AI ​​model can determine whether it can deduce a definite single product intention. If so, it can make recommendations related to the single product. Otherwise, the AI ​​model can continue to reason from the perspective of the user's purchasing needs in the product category. For example, also for the paper towel category, the purchasing needs can be refined according to multiple dimensions such as usage scenarios, specifications and sizes. Among them, the process of refined reasoning can be carried out in a step-by-step cycle multiple times, and each time it can be refined from one dimension. After receiving user feedback, if the single product intention can be deduced, the single product recommendation can be made. Otherwise, it can be further refined from another dimension, and so on. Of course, the maximum number of refinement rounds can also be set as an end condition. That is to say, assuming the maximum number of refinement rounds is three, then after three refinements, if the single product intention still cannot be deduced, it is not necessary to continue to refine and there is no need to make single product recommendations, so as not to affect the user's normal browsing.

[0043] Among them, when refining the reasoning of purchasing needs under a specific product category, since it may involve multiple interactions, each interaction is refined from only one dimension. If the intention of a single product cannot be inferred, the refinement is continued from another dimension. Therefore, this multi-round interaction process can form a refinement chain, which can include multiple nodes, and each node corresponds to the refinement dimension used in each interaction. Among them, different product categories can correspond to different refinement chains, that is, when refining needs for different product categories, the order of refinement and the specific demand options can be different. For example, for the paper towel category, the specific refinement can be carried out in the order of scene, specification, price, etc. Specific scene options can include kitchen, living room, bathroom, etc., and specification options can include small packages, stockpiling packages, etc. If it is a category of relatively high-priced electronic products such as mobile phones, the demand can be refined from the price dimension first, and then from the brand dimension, and so on. In order to achieve the goal of refining different product categories according to different links, the AI ​​model for specific demand refinement can be implemented through a structure such as a basic model + multiple category sub-models. Each category sub-model can be used to learn the dimension order corresponding to the specific category and the demand options corresponding to each dimension.

[0044] From the perspective of system architecture, see Figure 1 , the embodiment of the present application can provide AI interactive functions for relevant pages in the product information service system, and this function can be divided into a client and a server. Among them, the server can run the main program to read the user's behavior data, assemble it into prompt information of the AI ​​model, and call the AI ​​model to provide the client with option information inferred by the AI ​​model, etc. The client is used to display front-end information, including the display of demand options, the display of product information recommended during the reasoning process, and so on. Among them, the specific relevant page can be the client homepage, or a channel page, an event venue page, and so on. In the process of displaying product recommendation information to users through the above-mentioned relevant pages, the server main program can first analyze and infer the user's behavior data by calling the first AI model; if the definite single product intention cannot be inferred, but the category intention can be inferred, the user's cognitive information for the product category can be continued to be judged by calling the second AI model. On this basis, it is judged again whether the deterministic single product intention can be inferred. If it still cannot be inferred, the third AI model can be called to infer the user's purchasing needs based on the product category. If the deterministic single product intention still cannot be inferred after reasoning based on shopping needs, the demand options can be displayed to the user. After the user makes a selection, it is judged again whether the deterministic single product intention can be inferred. If not, the purchase demand reasoning will continue, and this cycle will be executed until the deterministic single product intention can be inferred, or the number of cycles reaches the threshold.

[0045] Specifically, AI models can refer to deep learning models containing massive parameters. Due to their large scale, these AI models can store and process vast amounts of information, enabling higher performance across various tasks. Therefore, leveraging the powerful content understanding and logical reasoning capabilities of pre-trained AI models, we can perform tasks such as determining user purchase intent, inferring purchase awareness, and refining demand.

[0046] The specific implementation scheme provided in the embodiments of this application is introduced in detail below.

[0047] First, the present invention provides a method for exchanging commodity information. Figure 2 , the method may specifically include:

[0048] S201: In the process of displaying product information on the target page, the first AI model is called according to the user's first behavior data to infer the user's current purchase intention type.

[0049] Among them, as mentioned above, the target page can be a client page, channel page, event venue page, store page, etc. in the product information service system. As long as the page needs to display aggregated information of multiple products, the AI ​​interactive function provided by the embodiment of this application can be introduced to realize AI-based product shopping guide.

[0050] In a specific implementation, the AI ​​inference process can be initiated after the user enters the target page, or it can be triggered by certain events, etc. After the AI ​​inference process is initiated, the first AI model can be called based on the user's first behavior data. This first AI model can be used to infer the user's current purchase intent type. Specifically, the first behavior data can refer to the user's behavior data generated within the current product information service system on the same day or in the last few hours. Of course, the specific input to the AI ​​model can be the specific operation type and the corresponding product information, such as information about products recently added to the shopping cart, information about products whose detail pages were recently viewed, information about recently purchased products, information about recently searched products, etc. In a specific implementation, since the system stores logs of various user actions, identification information such as the product ID associated with the specific action can be extracted from them. The product title, specifications, description information, etc. corresponding to the specific product ID can then be obtained from the product database, etc., and this information can then be input into the first AI model. Among them, the AI ​​model used in the embodiment of the present application can be a multimodal model, that is, it can process information in multiple modalities such as text and images. Therefore, the product information specifically input into the AI ​​model can also include product pictures, etc., so that the AI ​​model can make a comprehensive judgment through multimodal information.

[0051] Among them, in the embodiment of the present application, the purchase intention types can be divided into three categories, which can specifically include deterministic single product intention, category intention, and no clear intention. In specific implementation, it can be determined first whether the user has a clear single product intention. If not, it can be determined whether there is a category intention. If neither, it can be determined that there is no clear intention, and so on. For example, assuming that the user clicks on the details page of a certain product from the current page and performs an add-to-cart operation, it can usually be determined that the user has a clear single product intention for the product; or, assuming that the user repeatedly clicks on multiple products from the current page to view the details pages, and they are all products of the same category, then the user may have a category intention for the category; or, assuming that the user is just browsing the current page, but has not clicked to view a certain details page, or has viewed the details pages of certain products, but these products are scattered in multiple categories, then it can be considered that the user has no clear intention, and so on. The specific judgment process can be performed by the AI ​​model based on the knowledge learned during the pre-training process. Of course, the AI ​​model can also achieve autonomous learning during the judgment process.

[0052] S202: If the purchase intention type is a category intention for the target product category but no definite single product intention, then the second AI model is called based on the second behavioral data of the user related to the target product category to infer the user's cognitive information about the target product category, and determine whether a definite single product intention can be inferred based on the cognitive information.

[0053] After completing the intention type judgment, if the user has a definite intention for a single product, the AI ​​model can further generate recommendation information related to the specific single product (for example, assuming that the user adds a single product to the shopping cart but has not placed an order, it can analyze whether the specific single product is worth buying and provide analysis results), or it can also be price comparison information between single products (for example, price comparison information between single products of the same category, same specifications, and different brands, etc.), or it can also determine whether a specific single product has multiple SKU (minimum stock keeping unit) decision attributes. If so, it can also generate the same product selection strategy information for different SKUs, etc. For example, a certain product may have multiple different specifications, and decision-making auxiliary information can be provided about what scenarios each specification is suitable for, etc.

[0054] If it is determined that the user does not have a clear single product intention, it can be determined whether the user has category intention for a certain product category. If so, a second AI model can be further invoked based on the user's second behavioral data related to the specific target product category, so that the second AI model can infer the user's purchase awareness of the target product category and determine whether it can infer a definitive single product intention based on this purchase awareness information. Since it is necessary to analyze the user's purchase awareness of a certain product category, and this information usually requires the analysis of more data to obtain, the so-called second behavioral data here can generally be behavioral data related to a specific product category over a longer period of time compared to the aforementioned first behavioral data. For example, it can be data related to purchases and add-on purchases of products in the current target product category within the past year. By analyzing this data, the second AI model can infer information about the user's awareness of the target product category. For example, whether there is a clear brand preference or a clear specification preference for the product category. If so, there is a greater chance of inferring a clear single product intention based on this.

[0055] S203: If the deterministic single product intention cannot be inferred, the third AI model is called according to the third behavioral data of the user related to the target product category to infer the user's purchase demand under the target product category, and if the deterministic single product intention cannot be inferred based on the purchase demand inference result, the corresponding demand option is output so as to continue the purchase demand inference in combination with the user's selection result of the demand option.

[0056] If a clear intention for a specific product cannot be inferred based on the user's cognitive information about the target product category, a third AI model can be invoked based on third behavioral data related to the target product category, allowing the third AI model to infer the user's purchase needs within the target product category. The third behavioral data can be the same as or different from the second behavioral data. For example, the third behavioral data can include both the first and second behavioral data, i.e., it can include both current behavioral data and historical behavioral data over a longer period of time. Reasoning from the perspective of purchase needs specifically involves reasoning from perspectives such as specific purchase scenarios and price points. After reasoning from the perspective of purchase needs, it can be further determined whether a definitive intention for a specific product can be inferred based on the purchase needs reasoning results. If not, corresponding demand options can be output to continue reasoning from the perspective of shopping needs based on the user's selection of the demand options. That is, if the user's category intention is initially inferred, the user's cognitive information regarding the category can be first determined. If a clear intention for a specific product cannot be inferred, the AI ​​model can continue reasoning from the perspective of the user's purchase needs, hoping to determine whether a definitive intention for a specific product can be inferred.

[0057] Specifically, the process of inferring the user's purchase demand for the target product category can be repeated multiple times, each time refining the purchase demand from one dimension. If a single product intent cannot be inferred, multiple optional demand options within that dimension can be presented to the user. After receiving the user's selection of the desired option, the third AI model can be re-invoked based on the user's selection and the aforementioned third behavior data, and the purchase demand refinement step from another dimension can be performed until a definitive single product intent is derived or the number of cycles reaches a threshold. In this way, multiple interactions between the AI ​​and the user can be achieved, and more information can be gradually acquired during these interactions, gradually helping the AI ​​model output more accurate inference results. Furthermore, after each interaction, there is an opportunity to infer a clear single product intent. Therefore, compared to requiring the user to select from multiple dimensions at once, the inference process can be completed in a more lightweight interactive manner with minimal user disruption. So-called lightweight interaction means that the user is presented with multiple demand options from only one dimension at a time, and the user only needs to select one demand option from among them, rather than having to select multiple demand options from multiple dimensions at once. For example, in an embodiment of the present application, multiple options are first provided from the scene dimension, and the user selects one of the options, that is, the interaction is completed by performing a single click operation. If demand options on multiple dimensions are provided at one time, the demand options may be provided from the scene, brand, price and other dimensions respectively, and the user needs to perform at least three click operations to complete the selection of the demand options on the three dimensions. However, in actual applications, it may be possible to infer the intention of a single product only from the selection results on the scene dimension. At this time, the user does not need to make choices from other dimensions such as brand and price, making it possible to infer the intention of a single product through fewer user operations. Although the method of providing demand options on multiple (for example, three) dimensions at one time may collect more information at one time, for the user, each interaction requires at least three click operations to complete the interaction.

[0058] In specific implementations, when the third AI model performs purchase demand refinement reasoning each time it is called, the specific dimensions from which demand refinement is required for different categories, as well as the specific demand options corresponding to each demand, may all be different. In addition, as mentioned above, in the embodiments of the present application, there is also a goal to infer the user's single product intention with as few cycles as possible. Therefore, the specific dimension from which interaction is performed first is also relatively critical, and the priorities of products in different categories are obviously different in terms of specific dimensions. For example, if it is a paper towel product, when the offline service staff guides the user, they may first ask the user in what scenario the product needs to be used, then ask what specifications are required, and finally ask what price range is required. Following this order, it is easiest to locate the product that meets the user's needs. However, if it is a mobile phone product, when the offline service staff guides the user, they may first ask the user what brand they need, then ask what price range they need. After asking questions in these two dimensions, it may be possible to locate the specific single product, and so on. Therefore, purchase demand can be refined by following the personalized links corresponding to the target product category. This personalized link can include the order of dimensions corresponding to the target product category and the demand options corresponding to each dimension. The order of dimensions can be determined with the goal of inferring the user's individual product intentions within a specific product category with fewer iterations.

[0059] The above-mentioned personalized link information can be implemented by pre-configuring it into the third AI model, or inputting it into the third AI model as part of the input information. Alternatively, in a more preferred manner, the specific third AI model can be composed of a basic model and multiple category sub-models. The specific category sub-models are used to generate personalized link information for the corresponding categories, while the basic model can be used to determine the dimensions and corresponding demand options selected each time the purchase demand is refined based on the personalized link information generated by the category sub-models. In this way, specific personalized link information can also be automatically generated by the AI ​​model, eliminating the need for manual configuration or input.

[0060] It should be noted that in the solution provided in the embodiment of the present application, although the reasoning process is performed step by step, in fact, in each step of the reasoning, it is possible to infer a clear single product intention. In the case of introducing a deterministic single product intention, the corresponding AI model can also be used to generate product recommendation information based on the deterministic single product intention, or single product price comparison information, or determine whether the single product has multiple SKU decision attributes. If so, it can also generate the same product selection strategy information with different SKUs for display to the user.

[0061] In addition, if the first AI model finds that the user has no clear purchasing intention, it can display some general information, such as regional differentiated operation strategies, weather strategies, season / solar term strategies, etc. Specifically, assuming that it is summer, product information related to hot-selling sunscreen products, etc. can be provided. Among them, when displaying various strategies, category options under various strategies can also be generated for users to choose. After the user makes a selection for one of the categories, it means that the user has a category intention for the category. After that, it can enter step S202 to execute the processing logic after determining the category intention, that is, it can also determine the user's purchase awareness of the category, and can continue to refine the user's purchase needs for the category. Each time it is refined, it can be determined whether a clear single product intention can be inferred. If not, it can continue to be refined, and so on.

[0062] Specifically, when displaying AI interactive information, if it is necessary to display demand options or the above-mentioned regional differentiated operation strategies, weather strategies, solar term strategies, etc., the specific demand options can be displayed through the overlay layer of the current target page (for example, floating layer, pop-up window, drawer layer, drop-down box, etc.) or auxiliary interface (for example, sidebar, or vertical stretching, etc.), etc. The user can select the specific demand option and submit the selection result. After the user selects a demand option, if the definite single product intention cannot be inferred, some related products can be recommended based on the selection result, and demand options in other dimensions can also be provided for the user to continue to select, etc. After the user makes a selection, the specific recommended product information and demand options in other dimensions can be inserted into the current target page for display, or the product information currently in the display focus on the current target page can be replaced, etc.

[0063] For example, when displaying the weather strategy, in the initial state, you can Figure 3 As shown at 31 in (A), a floating interface can be displayed on the target page, which can display interactive information generated according to the weather strategy and related demand options. For example, it can include category options such as "light sunscreen", "ultra-light parasol", and "folding small fan". Assuming that the user selects "ultra-light parasol", at this time, Figure 3 As shown at 32 in (B), product recommendation information related to "ultralight parasol" can be displayed at the center of the current display focus area on the target page, and brand dimension demand options can also be displayed in this area. For example, Figure 3 (B) shows the brand names of AA, BB, CC, etc. If the user continues to select one of the brands "AA", then Figure 3As shown at 33 in (C), product recommendation results related to the "AA" brand "ultra-light parasol" can be displayed at the center of the current display focus area.

[0064] In addition, another specific display method in the interaction process may be that after jumping from the target page to the details page of a certain product and then returning to the target page, it can usually be inferred that the user has a definite single product intention for the product. At this time, the information displayed in the resource position where the product was originally located in the target page can be replaced with the information generated by the corresponding AI model.

[0065] For example, suppose Figure 4 (A) shows a target page described in an embodiment of the present application. The page is a channel page of a "flash sale" channel, which displays information of multiple products. Assume that the user clicks on the product link shown in 41 and browses the following Figure 4 (B) shows the product details page, and then returns to the original target page. Figure 4 As shown at 42 in (C), the product link originally displayed at 41 in the target page is changed to relevant AI interactive information about "more convenient fast food recommendations".

[0066] It should be noted that, in specific implementations, the first, second, and third AI models can be pre-trained or fine-tuned using the same basic AI model to enable them to identify purchase intent type, purchase awareness level, and demand refinement. Alternatively, the first, second, and third AI models can correspond to different basic AI models, which is not limited here.

[0067] In summary, through the embodiment of the present application, in the process of displaying product information on the target page, the first AI model can be first called according to the user's first behavior data, so that the first AI model can infer the user's current purchase intention type. If the inferred purchase intention type is a category intention for the target product category but no deterministic single product intention, the second AI model can be called according to the user's second behavior data related to the target product category, so that the second AI model can infer the user's cognitive information about the target product category, and determine whether it can infer a deterministic single product intention based on the cognitive information; if the deterministic single product intention still cannot be inferred, the third AI model can be further called according to the third behavior data related to the target product category, so that the third AI model can infer the user's purchase demand under the target product category, and if the deterministic single product intention cannot be inferred based on the purchase demand inference result, the corresponding demand option can be output so as to continue to infer the purchase demand inference in combination with the user's selection result of the demand option. In this way, a horizontal shopping guide capability that is independent of the traditional recommendation algorithm is provided, and by judging the user's current purchase intention, incremental value is brought to the system on the basis of traditional passive shopping guide. Moreover, since multiple rounds of interactions can be achieved between the main program and the AI ​​model and the user, when the AI ​​model cannot directly infer the definite intention of a single product through one round of interaction, it is not necessary to force the AI ​​model to give an inference result on the single product dimension. Instead, it can first determine whether there is a category intention. Then, through multiple rounds of interactions such as identifying the user's cognitive information about the category and inferring purchase needs, the AI ​​model can gradually obtain more valuable information. In addition, each round of interaction may have the possibility of inferring a clear intention of a single product. If not, the needs will be further refined, and combined with the demand option selection results fed back by the user, a more accurate judgment can be made. Therefore, the quality of the interaction results can be improved, and the effectiveness of proactive shopping guide information can be improved.

[0068] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0069] Corresponding to the aforementioned method, an embodiment of the present application further provides an information interaction device, which may include:

[0070] An intention type inference unit is used to call a first artificial intelligence (AI) model based on the user's first behavior data to infer the user's current purchase intention type during the process of displaying product information on the target page;

[0071] a category cognition inference unit, configured to, if the purchase intention type is category intention for a target product category but no definite individual product intention, invoke a second AI model based on the second behavioral data of the user related to the target product category to infer the user's cognition information about the target product category, and determine whether definite individual product intention can be inferred based on the cognition information;

[0072] The purchase demand inference unit is used to call the third AI model based on the third behavioral data of the user related to the target product category to infer the user's purchase demand under the target product category if the deterministic single product intention cannot be inferred, and output the corresponding demand option if the deterministic single product intention cannot be inferred based on the purchase demand inference result, so as to continue the purchase demand inference in combination with the user's selection result of the demand option.

[0073] Among them, the process of inferring the user's purchasing needs under the target product category is executed in a loop multiple times, and each time the purchasing needs are refined from one dimension. After receiving the user's selection result for the demand option, the third AI model is re-called according to the selection result and the third behavior data, and the step of refining the purchasing needs from another dimension is executed until a deterministic single product intention is inferred or the number of cycles reaches a threshold.

[0074] Among them, when the third AI model performs detailed reasoning on purchase needs each time it is called, it performs detailed reasoning on purchase needs according to the personalized link corresponding to the target product category; wherein, the personalized link includes the dimensional order corresponding to the target product category and the demand options corresponding to each dimension; the dimensional order is determined with the goal of inferring the user's intention for a single product under the target product category with fewer cycles.

[0075] Specifically, the third AI model consists of a basic model and multiple category sub-models. The multiple category sub-models are used to generate personalized link information for corresponding categories. The basic model is used to determine the dimensions and corresponding demand options selected each time the purchase demand refinement reasoning is performed based on the personalized link information generated by the category sub-models.

[0076] In addition, the device may further include:

[0077] The unit for providing single product recommendation information is used to generate product recommendation information based on the deterministic single product intention if the deterministic single product intention can be inferred, or single product price comparison information, or to determine whether the single product has multiple minimum inventory unit SKU decision attributes. If so, it generates product selection strategy information for the same product but different SKUs for display to users.

[0078] The page information replacement unit is used to replace the information displayed in the resource position of the product in the target page with the information generated by the corresponding AI model if the user jumps from the target page to the details page of a certain product and then returns to the target page, and it is inferred that the user has a deterministic single product intention.

[0079] Furthermore, the device may further include:

[0080] The no-clear-intention interaction unit is used to generate candidate category options according to regional differentiated operation strategies, weather strategies or season / season strategies if the purchase intention type inferred by the first AI model is no clear purchase intention, so as to determine that the user has category intention for the category after receiving the user's selection result for one of the categories, and execute the cognitive information judgment after identifying the category intention and the steps of reasoning according to the purchase needs.

[0081] In addition, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of any one of the methods in the aforementioned method embodiments are implemented.

[0082] And an electronic device comprising:

[0083] one or more processors; and

[0084] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.

[0085] A computer program product includes a computer program / computer executable instructions, which implement the steps of the method described in the above method embodiment when executed by a processor in an electronic device.

[0086] in, Figure 5 The electronic device architecture is shown as an example, and may include a processor 510, a video display adapter 511, a disk drive 512, an input / output interface 513, a network interface 514, and a memory 520. The processor 510, the video display adapter 511, the disk drive 512, the input / output interface 513, the network interface 514, and the memory 520 may be communicatively connected via a communication bus 530.

[0087] Among them, the processor 510 can be implemented by a general-purpose CPU (Central Processing Unit, processor), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solutions provided in this application.

[0088] The memory 520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 520 can store an operating system 521 for controlling the operation of the electronic device 500, and a basic input and output system (BIOS) for controlling the low-level operations of the electronic device 500. In addition, a web browser 523, a data storage management system 524, and an information interaction processing system 525, etc. can also be stored. The above-mentioned information interaction processing system 525 can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present application. In short, when the technical solution provided in this application is implemented by software or firmware, the relevant program code is stored in the memory 520 and is called and executed by the processor 510.

[0089] The input / output interface 513 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0090] The network interface 514 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WIFI, Bluetooth, etc.).

[0091] The bus 530 comprises a pathway for transmitting information between the various components of the device (eg, the processor 510 , the video display adapter 511 , the disk drive 512 , the input / output interface 513 , the network interface 514 , and the memory 520 ).

[0092] It should be noted that although the above device only shows a processor 510, a video display adapter 511, a disk drive 512, an input / output interface 513, a network interface 514, a memory 520, a bus 530, etc., in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may also include only the components necessary to implement the solution of the present application, and does not necessarily include all the components shown in the figure.

[0093] Through the description of the above embodiments, it can be seen that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present application or certain parts of the embodiments.

[0094] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0095] The above describes in detail the information interaction method and electronic device provided by this application. Specific examples are used herein to illustrate the principles and implementation methods of this application. The description of the above embodiments is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the contents of this specification should not be understood as limiting this application.

Claims

1. An information interaction method, characterized in that: include: In the process of displaying product information on the target page, the first artificial intelligence (AI) model is called based on the user's first behavior data to infer the user's current purchase intention type; If the purchase intention type is category intention for the target product category but no definite single product intention, then calling a second AI model based on the second behavioral data of the user related to the target product category to infer the user's cognitive information about the target product category, and determining whether definite single product intention can be inferred based on the cognitive information; If the deterministic single product intention cannot be inferred, the third AI model is called based on the third behavioral data of the user related to the target product category to infer the user's purchasing needs under the target product category, and if the deterministic single product intention cannot be inferred based on the purchase demand inference result, the corresponding demand option is output so as to continue the purchase demand inference in combination with the user's selection result of the demand option.

2. The method according to claim 1, characterized in that The process of inferring the user's purchasing needs under the target product category is executed in a loop multiple times, and each time the purchasing needs are refined from one dimension. After receiving the user's selection result for the demand option, the third AI model is re-called according to the selection result and the third behavior data, and the step of refining the purchasing needs from another dimension is executed until a deterministic single product intention is inferred or the number of cycles reaches a threshold.

3. The method according to claim 2, characterized in that Each time the third AI model is called to perform detailed reasoning on purchase needs, it performs detailed reasoning on purchase needs according to the personalized link corresponding to the target product category; wherein, the personalized link includes the dimensional order corresponding to the target product category and the demand options corresponding to each dimension; the dimensional order is determined with the goal of inferring the user's intention for a single product under the target product category with fewer cycles.

4. The method according to claim 3, characterized in that The third AI model consists of a basic model and multiple category sub-models. The multiple category sub-models are used to generate personalized link information for corresponding categories. The basic model is used to determine the dimensions selected and the corresponding demand options for each purchase demand refinement reasoning based on the personalized link information generated by the category sub-models.

5. The method according to claim 1, wherein Also includes: If the deterministic single product intention can be inferred, the corresponding AI model is also used to generate product recommendation information based on the deterministic single product intention, or single product price comparison information, or to determine whether the single product has multiple minimum inventory unit SKU decision attributes. If so, the same product with different SKUs is generated for selection strategy information for display to users.

6. The method according to claim 5, characterized in that Also includes: If after jumping from the target page to the details page of a certain product, the user returns to the target page, and it is inferred that the user has a definite single product intention, the information displayed in the resource position where the product was originally located in the target page will be replaced with the information generated by the corresponding AI model.

7. The method according to claim 1, characterized in that Also includes: If the purchase intention type inferred by the first AI model is no clear purchase intention, candidate category options are generated according to regional differentiated operation strategies, weather strategies or season / season strategies, so that after receiving the user's selection result for one of the categories, it is determined that the user has category intention for the category, and the cognitive information judgment after identifying the category intention and the steps of reasoning according to the purchase needs are executed.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

9. An electronic device, characterized in that: include: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / computer executable instructions, characterized in that When the computer program / computer executable instructions are executed by a processor in an electronic device, the steps of the method according to any one of claims 1 to 7 are implemented.

Images