Method, device and electronic device for generating recommendation information based on anthropomorphic intelligent agent

Abstract

The embodiments of the present disclosure disclose a method, device, and electronic device for generating recommendation information based on an anthropomorphic intelligent agent. A specific implementation of the method includes: collecting a set of behavior information corresponding to a target user; extracting features of a multi-dimensional personality dimension from the behavior information set to obtain a set of behavior features; determining the personality attribute label corresponding to the target user based on the behavior feature set; matching the target anthropomorphic intelligent agent corresponding to the personality attribute label; determining the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic intelligent agent and the behavior feature set; filling the template of the information to be recommended based on the recommendation template corresponding to the recommendation type to obtain the target information to be recommended. This implementation achieves highly scalable and accurate information recommendation.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology, in particular to the fields of artificial intelligence, computer behavior modeling, and information recommendation, and specifically to a method, device, and electronic device for generating recommendation information based on an anthropomorphic intelligent agent. Background Art

[0002] An intelligent agent is an agent that can perceive its environment and perform actions to achieve specific goals. In the field of recommendation information generation, recommendations are typically made using methods such as collaborative filtering or static rule matching. However, these methods suffer from poor scalability and inaccurate recommendations in specific areas.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0004] The content of this disclosure is used to briefly introduce concepts that will be described in detail in the detailed description section below. The content of this disclosure is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] Some embodiments of the present disclosure propose a method, device, and electronic device for generating recommendation information based on an anthropomorphic intelligent agent to solve the technical problems mentioned in the above background technology section.

[0006] In a first aspect, some embodiments of the present disclosure provide a method for generating recommendation information based on an anthropomorphic agent, the method comprising: collecting a behavior information set corresponding to a target user, wherein the target user is a user who has authorized the corresponding user behavior collection authority and is to be recommended information; performing feature extraction of a multi-dimensional personality dimension on the behavior information set to obtain a behavior feature set, wherein each behavior feature in the behavior feature set corresponds to a personality dimension; determining the personality attribute label corresponding to the target user based on the behavior feature set; matching a target anthropomorphic agent corresponding to the personality attribute label, wherein the target anthropomorphic agent is an anthropomorphic agent that recommends information according to the recommendation style corresponding to the personality attribute label; determining the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic agent and the behavior feature set; and filling in the template of the information to be recommended based on the recommendation template corresponding to the recommendation type to obtain the target information to be recommended.

[0007] In a second aspect, some embodiments of the present disclosure provide a recommendation information generation device based on an anthropomorphic intelligent agent, the device comprising: a collection unit, configured to collect a behavior information set corresponding to a target user, wherein the target user is a user who has authorized the corresponding user behavior collection authority and is to be recommended information; a feature extraction unit, configured to perform feature extraction of a multi-dimensional personality dimension on the behavior information set to obtain a behavior feature set, wherein each behavior feature in the behavior feature set corresponds to a personality dimension; a first determination unit, configured to determine the personality attribute label corresponding to the target user based on the behavior feature set; a matching unit, configured to match a target anthropomorphic intelligent agent corresponding to the personality attribute label, wherein the target anthropomorphic intelligent agent is an anthropomorphic intelligent agent that recommends information according to the recommendation style corresponding to the personality attribute label; a second determination unit, configured to determine the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic intelligent agent and the behavior feature set; a template filling unit, configured to perform template filling on the information to be recommended based on the recommendation template corresponding to the recommendation type to obtain the target information to be recommended.

[0008] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0009] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation of the first aspect is implemented.

[0010] The aforementioned embodiments of the present disclosure have the following beneficial effects: Through the anthropomorphic agent-based recommendation information generation method of some embodiments of the present disclosure, highly scalable and accurate information recommendation is achieved. Specifically, first, a set of behavior information corresponding to a target user is collected, where the target user is a user who has authorized permission to collect corresponding user behavior and for whom information recommendation is to be made. Second, multi-dimensional personality feature extraction is performed on the behavior information set to obtain a behavior feature set, where each behavior feature in the behavior feature set corresponds to a personality dimension. In practice, conventional behavior information extraction methods typically directly extract corresponding features based on the behavior type of the behavior information. However, this extraction method ignores the feature description from the perspective of personality dimensions, resulting in poor expressiveness of the extracted features. Second, based on the behavior feature set, a personality attribute label corresponding to the target user is determined. Next, a target anthropomorphic agent corresponding to the personality attribute label is matched, where the target anthropomorphic agent is an anthropomorphic agent that recommends information according to the recommendation style corresponding to the personality attribute label. Furthermore, based on the target anthropomorphic agent and the behavior feature set, the information to be recommended and the recommendation type corresponding to the information to be recommended are determined. In practice, conventional methods often directly recall recommended information after extracting features. However, due to the poor expressiveness of the extracted features and the lack of corresponding classification of different features, the accuracy of the recalled recommended information is poor. Based on this, the present disclosure determines the personality attribute label by combining behavioral characteristics, matches the corresponding target anthropomorphic intelligent agent, and recalls the corresponding recommended information based on the target anthropomorphic intelligent agent. This method realizes the recall of recommended information for users with the same or similar personality attributes. Compared with the "thousands of faces" recall method, the recall efficiency is significantly improved. In addition, due to the combination of personality attribute labels, anthropomorphic recommended information recall is realized, thereby improving the accuracy of the recalled recommended information. Finally, according to the recommendation template corresponding to the above recommendation type, the above information to be recommended is template-filled to obtain the target information to be recommended. In this way, the display flexibility of the recommended information is improved, thereby promoting the user's residence probability. In summary, the recommendation information generation method based on anthropomorphic intelligent agent realizes highly scalable and accurate information recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0012] Figure 1 is a flowchart of some embodiments of the method for generating recommendation information based on an anthropomorphic agent according to the present disclosure;

[0013] Figure 2 It is a schematic diagram of the time window division process;

[0014] Figure 3 This is a schematic diagram of the recommendation process corresponding to the conventional method;

[0015] Figure 4 is a schematic diagram of the recommendation process incorporating an anthropomorphic intelligent agent disclosed herein;

[0016] Figure 5 is a schematic structural diagram of some embodiments of the apparatus for generating recommendation information based on an anthropomorphic intelligent agent according to the present disclosure;

[0017] Figure 6 is a schematic structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0019] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other.

[0020] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0023] Before performing the collection, storage, and use of behavioral information involved in this disclosure, relevant organizations or individuals must fulfill their obligations, including conducting personal information security impact assessments, fulfilling their obligation to inform personal information subjects, and obtaining prior authorization and consent from personal information subjects (e.g., target users).

[0024] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0025] refer to Figure 1 , shows a process 100 of some embodiments of the method for generating recommendation information based on an anthropomorphic agent according to the present disclosure. The method for generating recommendation information based on anthropomorphic agent includes the following steps:

[0026] Step 101: Collect a set of behavior information corresponding to the target user.

[0027] In some embodiments, the execution entity (eg, a computing device) of the recommendation information generation method based on anthropomorphic agents may collect a set of behavior information corresponding to the target user.

[0028] The target user is the user who has authorized the corresponding user behavior collection permission and is to be recommended information. The behavior information represents the user behavior of the target user.

[0029] In practice, before collecting the aforementioned behavioral information set, the aforementioned execution entity may initiate a permission request prompt from the target user for permission to collect user behavior information, as well as other permission prompts for analyzing, storing, and processing behavioral information. Upon obtaining authorization from the target user, the execution entity may then collect the corresponding behavioral information set from the target user. Specifically, for example, when behavioral information corresponding to the target user is generated, the execution entity may monitor the generated behavioral information through a trigger, thereby passively collecting the generated behavioral information. Alternatively, the execution entity may actively collect the generated behavioral information corresponding to the target user at regular intervals.

[0030] It should be noted that the computing device described above can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules, for example, to provide distributed services, or as a single software or software module. No specific limitations are given here.

[0031] Optionally, the behavior information in the behavior information set includes: behavior content, content type, behavior initiation time, content access rights, content exposure information, and content comments. The behavior content represents the specific content of the behavior corresponding to the behavior information. The content type represents the data type of the behavior content. For example, content types may include text, audio, video, and mixed types consisting of at least one of these. Content access rights represent the access rights set by the target user for the user behavior. For example, content access rights may include first, second, and third content access rights. The first content access right represents that any user can access the behavior content. The second content access right represents that a limited number of users, as set by the target user, can access the behavior content. The third content access right represents that only the target user can access the behavior content. Content exposure information represents the level of attention paid to the behavior content by other users. For example, content exposure information may include other interactivity metrics such as dwell time, number of likes, and number of favorites. Dwell time represents the length of time that other users spend on the behavior content. Number of likes represents the number of likes received by other users for the behavior content. The number of collections represents the number of collections of the action content by other users. Content comments represent the comments of other users on the user's action.

[0032] Step 102: extracting features of multi-dimensional personality dimensions from the behavior information set to obtain a behavior feature set.

[0033] In some embodiments, the execution subject may perform feature extraction of multi-dimensional personality dimensions on the behavior information set to obtain a behavior feature set.

[0034] Each behavioral characteristic in the above set of behavioral characteristics corresponds to a personality dimension. The personality dimensions can correspond to the five personality dimensions (The Five-Factor Model of Personality). Specifically, the personality dimensions may include: the first personality dimension, the second personality dimension, the third personality dimension, the fourth personality dimension, and the fifth personality dimension. The first personality dimension may correspond to "extraversion." The second personality dimension may correspond to "openness to experience." The third personality dimension may correspond to "conscientiousness." The fourth personality dimension may correspond to "neuroticism." The fifth personality dimension may correspond to "agreeableness."

[0035] In practice, the above-mentioned execution subject can combine the first personality dimension, the second personality dimension, the third personality dimension, the fourth personality dimension and the fifth personality dimension to perform multi-dimensional personality dimension feature extraction on the behavior information set to obtain a behavior feature set.

[0036] In some optional implementations of some embodiments, the execution subject extracts features of a multi-dimensional personality dimension from the behavior information set to obtain a behavior feature set, including:

[0037] Step S1: Divide the time distribution corresponding to the above-mentioned behavior information set into time windows according to a preset time granularity to obtain at least one time window.

[0038] The window length of the time window corresponds to the preset time granularity.

[0039] In practice, the execution entity may divide the time distribution corresponding to the behavior information set into time windows with a preset time granularity as the window length to obtain at least one time window.

[0040] For example, see Figure 2 The diagram shows a time window division process, where the behavior information set may include: behavior information A1, behavior information A2, behavior information A3, ..., behavior information An-1, behavior information An. The time distribution corresponding to the behavior information set may be the time period from the "T-start" time point to the "T-end" time point. The preset time granularity may be hourly or daily, so the window length of the time window may be hourly or daily. At least one time window may include: time window W1, ..., time window Wm.

[0041] Step S2: grouping the behavior information in the behavior information set according to the at least one time window to obtain a behavior information group set.

[0042] Each piece of behavior information in the behavior information group corresponds to the same time window.

[0043] For example, see further Figure 2 , wherein the behavior information group corresponding to the time window W1 may include: behavior information A1, behavior information A2, behavior information A3, .... The behavior information group corresponding to the time window Wn may include: ..., behavior information An-1, behavior information An.

[0044] Step S3: For each behavior information group in the above behavior information group set, perform the following processing steps:

[0045] Step S31: generating a local heat map matrix corresponding to the behavior information group according to the content type included in the behavior information in the behavior information group.

[0046] The matrix dimension of the local heat map matrix is ​​L×1, where L represents the number of content types.

[0047] In practice, the execution entity may count the number of content types included in the behavior information in the behavior information group, and map them into thermal values ​​to obtain the local heat map matrix.

[0048] Step S32: Determine the content domain label, expression factor, and emotion factor corresponding to each behavior information in the behavior information group according to the behavior content and content type included in the behavior information.

[0049] The content domain label represents the content domain to which the behavioral content included in the behavioral information belongs. The expression factor represents the coherence of the behavioral content included in the behavioral information. The emotion factor represents the degree of emotional expression of the behavioral content included in the behavioral information. Specifically, the value range of the expression factor and the emotion factor is [0, 1]. In particular, the higher the value of the expression factor, the more coherent the content of the corresponding behavioral content. The higher the value of the emotion factor, the less emotional expression of the corresponding behavioral content.

[0050] In practice, the aforementioned execution entities can first convert the behavioral content into plain text based on the content type. For example, when the content type is video, optical character recognition (OCR) can be used to extract the text from the behavioral content, resulting in the text to be extracted. Alternatively, when the content type is audio, a speech-to-text model can be used to extract the text from the behavioral content, resulting in the text to be extracted. In particular, when the content type is text, the behavioral content can be used directly as the text to be extracted. Secondly, feature extraction is performed on the text to be extracted using the Word2Vec model to obtain text features. Finally, the text features are input into the content domain label classifier, expression factor mapper, and sentiment factor mapper, respectively, to obtain the content domain label, expression factor, and sentiment factor. The content domain label classifier is a multi-classifier. The expression factor mapper and sentiment factor mapper are both value mappers.

[0051] Step S33: Determine the interactivity factor corresponding to each piece of behavior information in the behavior information group based on the content access rights, content exposure information, and content comments included in the behavior information.

[0052] The interactivity factor can be used to characterize the interactive collaboration of the behavior information. Specifically, the value range of the interactivity factor can be [0, 1]. In particular, the higher the value of the interactivity factor, the higher the interactive collaboration of the corresponding behavior information.

[0053] In practice, first, since content exposure information includes other interactivity metrics such as dwell time, number of likes, and number of favorites, a feature vector corresponding to the content exposure information can be obtained through indicator mapping. For example, one-hot encoding can be used for indicator mapping. Second, text feature extraction is performed on the content comments to obtain a feature vector corresponding to the content comments. For example, a Word2Vec model can be used to extract text features from the content comments. Next, the feature vectors corresponding to the content exposure information and the feature vectors corresponding to the content comments are concatenated and applied to an interactivity factor mapper to obtain an initial interactivity factor. The interactivity factor mapper can be a value mapper. Finally, the initial interactivity factor is weighted using the weights corresponding to the content access rights to obtain the interactivity factor. In particular, since content access rights include first, second, and third content access rights, different weights can be assigned to the first, second, and third content access rights. For example, the weight associated with the first content access right can be 1, the weight associated with the second content access right can be 0.7, and the weight associated with the third content access right can be 0. In practice, since different content access rights will affect interactive collaboration, corresponding weights are set according to different content access rights to adjust the interactivity factor.

[0054] Step S4: performing heat map normalization splicing on at least one local heat map matrix obtained in time sequence to obtain a heat map matrix as the behavioral feature corresponding to the first personality dimension.

[0055] In practice, since at least one time window is obtained by dividing the time distribution corresponding to the behavioral information set into time windows, and each time window corresponds to a local heat map matrix, the heat map matrices in the at least one local heat map matrix can be spliced ​​in time sequence to obtain a heat map matrix as the behavioral feature corresponding to the first personality dimension. The matrix dimension of the heat map matrix can be L×m, where m represents the number of time windows in the at least one time window.

[0056] Step S5: Project the content domain labels, expression factors, emotional factors, and interactivity factors corresponding to the behavioral information into the matrix space corresponding to the heat map matrix, respectively, to obtain the behavioral characteristics corresponding to the second personality dimension, the behavioral characteristics corresponding to the third personality dimension, the behavioral characteristics corresponding to the fourth personality dimension, and the behavioral characteristics corresponding to the fifth personality dimension.

[0057] In practice, since the heat map matrix is ​​constructed based on the local heat map matrix corresponding to the behavior information group, projection can also be performed based on the behavior information group. In particular, since direct projection may cause feature dimension mismatch, when projecting the content domain labels, expression factors, emotional factors, and interactivity factors corresponding to the behavior information into the matrix space corresponding to the heat map matrix, it is necessary to use a downsampling network (FPN (Feature Pyramid Networks)) to compress the feature dimensions to ensure that the feature dimensions of the projected behavior features are consistent with the matrix dimensions of the heat map matrix.

[0058] Step 103: Determine the personality attribute label corresponding to the target user based on the behavioral feature set.

[0059] In some embodiments, the execution entity may determine a personality attribute tag corresponding to the target user based on a set of behavioral characteristics.

[0060] Among them, the personality attribute label can represent the personality attributes of the target user and the personality attributes of the selected anthropomorphic agent.

[0061] In practice, since the behavioral features in a behavioral feature set share the same dimension, the individual behavioral features in the set can first be superimposed to obtain a superimposed behavioral feature. This superimposed behavioral feature can then be used to classify the personality attribute labels using a classification model (e.g., a CNN (Convolutional Neural Network) model) to obtain a personality attribute label.

[0062] Optionally, the personality attribute label includes: a first personality attribute label and a second personality attribute label. The first personality attribute label represents the personality attribute self-assessed by the target user, and the second personality attribute label represents the predicted personality attribute for the target user. In practice, the first personality attribute label can be obtained through a personality attribute self-assessment questionnaire (e.g., the NEO-PI-R (revised Eysenck Personality Questionnaire) or other personality inventory). The second personality attribute label can be predicted using the aforementioned set of behavioral characteristics. Specifically, both the first and second personality attribute labels include weight values ​​for five personality dimensions (first, second, third, fourth, and fifth). For example, the first personality attribute label can be [0.3, 0.3, 0.2, 0.7, 0.8].

[0063] In some optional implementations of some embodiments, the execution entity determines the personality attribute label corresponding to the target user based on the behavioral feature set, including:

[0064] Step S1: Determine whether a target personality attribute label exists.

[0065] The target personality attribute label is a personality attribute self-evaluated by the target user before the current moment.

[0066] In practice, when a (target) user generates a personality attribute label through self-evaluation, it will be stored in the server. Therefore, it is possible to determine whether the target personality attribute label exists by querying whether there is a self-evaluated personality attribute label associated with the target user.

[0067] Step S2: In response to the existence of the target personality attribute label, determine whether the target personality attribute label is valid.

[0068] Among them, the above-mentioned target personality attribute label controls the corresponding label validity through the label validity period.

[0069] In practice, the personality attributes of target users may change at different ages, in different living environments, and under the influence of other factors. Therefore, it is necessary to set a corresponding label validity period for the target user's self-evaluated (target) personality attribute label so that the target user can regularly determine his or her own personality attributes through self-evaluation.

[0070] Step S3: In response to the target personality attribute label being valid, the target personality attribute label is determined as the first personality attribute label.

[0071] Step S4: In response to the target personality attribute tag being invalid, initiating entry of the first personality attribute tag for the target user to determine the first personality attribute tag and update the corresponding tag validity period.

[0072] In practice, a personality attribute self-assessment form may be sent to the target user to retrieve the first personality attribute label.

[0073] Step S5: generating the second personality attribute label according to the pre-trained personality attribute label prediction model, the first personality attribute label and the behavioral feature set.

[0074] In practice, the personality attribute label prediction model can use a convolutional neural network (CNN) as the backbone network. Specifically, the personality attribute label model consists of five parallel sub-CNNs, a feature stacking layer, and a multi-classifier. The five parallel sub-CNNs independently perform deep feature extraction on the behavioral features corresponding to the five different personality attribute labels, generating a deep feature set. The feature stacking layer stacks the deep features in the deep feature set to generate stacked features. The classifier outputs a candidate second personality attribute label based on the stacked features. Specifically, the second personality attribute label represents the weights of the five personality dimensions. Finally, the candidate second personality attribute label is weighted and summed with the first personality attribute label to obtain the second personality attribute label. Since a user's actual personality attributes may differ from those corresponding to their online behavior, correction is performed in conjunction with the first personality attribute label to ensure the validity of the predicted second personality attribute label.

[0075] Step 104: Match the target anthropomorphic agent corresponding to the personality attribute label.

[0076] In some embodiments, the above-mentioned execution subject can match the target anthropomorphic intelligent agent corresponding to the personality attribute label.

[0077] Among them, the target anthropomorphic intelligent agent is an anthropomorphic intelligent agent that recommends information according to the recommendation style corresponding to the personality attribute label.

[0078] In practice, conventional content recommendation methods primarily rely on user behavior to make recommendations, without fully considering the user's personality attributes. Therefore, the present disclosure utilizes an anthropomorphic agent to select and recommend recommended information during the recommendation process, thereby improving the accuracy and relevance of recommendations. Furthermore, since the dimensions of personality attributes are fixed (five personality dimensions), differences between users primarily manifest in the weights of personality attribute labels across the five personality dimensions, which are enumerable. Therefore, a limited number of anthropomorphic agents can be set up for all users to select and recommend corresponding recommended information. Compared to conventional "one-size-fits-all" recommendation methods that require training a separate recommender for each user, the present disclosure utilizes an anthropomorphic agent as a relay to select and recommend recommended information, effectively reducing training costs. Furthermore, in extreme cases where user behavior is scarce, the anthropomorphic agent of the present disclosure collects behavioral information from multiple users with similar personality attributes as a data source, enabling effective agent training and ensuring recommendation accuracy.

[0079] For example, see Figure 3The conventional recommendation method, such as the "one-size-fits-all" recommendation method, requires training a separate recommender for each user, such as Figure 3 As shown, a recommender R1 is set for user U1, a recommender R2 is set for user U2, ..., and a recommender Rn is set for user Un. When user behavior is scarce, there is a lack of sufficient training samples for training the recommender, resulting in poor recommendation results.

[0080] As yet another example, see Figure 4 The schematic diagram of the recommendation process combined with anthropomorphic agents is shown, wherein the present disclosure sets anthropomorphic agents as repeaters for users with the same personality attributes to select and recommend recommended information, such as Figure 4 As shown, an anthropomorphic agent G1 is set for user U1, user U2, .... An anthropomorphic agent Gm is set for user ... and user Un. Because personality attribute labels are enumerable, training costs are exponentially reduced in scenarios with a large user base. Furthermore, since each personality attribute label corresponds to an anthropomorphic agent, meaning that multiple users correspond to one anthropomorphic agent, this effectively avoids the lack of user behavior and ensures the accuracy of the anthropomorphic agent's recommendations.

[0081] Optionally, during the training and updating process of the anthropomorphic agent, a set of behavioral characteristics is combined and mapped into input factors for the five personality dimensions using a behavioral attribution model (behavioral attribution theory). A second personality attribute label is then output using a network model constructed using the Takagi-Sugeno fuzzy inference structure (comprising an input layer, a fuzzy layer, a rule layer, a normalization layer, and an output layer). Specifically, the initial fuzzy rules set in the rule layer can be set by experts. Subsequently, new rules can be automatically generated or existing rules optimized based on training to ensure that the anthropomorphic agent's information recommendations are more consistent with the personality attribute labels. During the training phase, parameter adjustment and updates are performed using a combination of backpropagation algorithms and least squares methods. In addition to information recommendation, the anthropomorphic agent can also be used in multiple fields, including human-computer interaction, automated anthropomorphic agent operation, and personalized social account simulation.

[0082] In some optional implementations of some embodiments, the execution subject matches the target anthropomorphic agent corresponding to the personality attribute label, including:

[0083] Step S1: Determine whether there is an anthropomorphic intelligent agent that matches the above-mentioned first personality attribute label.

[0084] In practice, the first personality attribute label can be used to match whether there is an anthropomorphic intelligent agent in a working state, so as to determine whether there is an anthropomorphic intelligent agent matching the above first personality attribute label.

[0085] Step S2: In response to the existence of an anthropomorphic agent matching the first personality attribute label, the anthropomorphic agent matching the first personality attribute label is added to the candidate anthropomorphic agent sequence as a candidate anthropomorphic agent.

[0086] Step S3: In response to the absence of an anthropomorphic agent matching the first personality attribute label, recall the anthropomorphic agent corresponding to the first approximate personality attribute label as a candidate anthropomorphic agent and add it to the candidate anthropomorphic agent sequence.

[0087] The first approximate personality attribute label is an attribute label whose label similarity with the first personality attribute label is greater than a preset threshold.

[0088] In practice, the weight corresponding to a personality dimension can be weakened to serve as a first approximate personality attribute label. Specifically, since there are five personality dimensions, when the weight of any personality dimension is weakened, at least five anthropomorphic agents can be recalled as candidate anthropomorphic agents.

[0089] Step S4: Determine whether there is an anthropomorphic intelligent agent that matches the second personality attribute label;

[0090] In practice, the first personality attribute label can be used to match whether there is an anthropomorphic intelligent agent in a working state, thereby determining whether there is an anthropomorphic intelligent agent matching the above-mentioned second personality attribute label.

[0091] Step S5: In response to the existence of an anthropomorphic agent matching the second personality attribute label, the anthropomorphic agent matching the second personality attribute label is added to the candidate anthropomorphic agent sequence as a candidate anthropomorphic agent.

[0092] Step S6: In response to the absence of an anthropomorphic agent matching the second personality attribute label, recall the anthropomorphic agent corresponding to the second approximate personality attribute label as a candidate anthropomorphic agent and add it to the candidate anthropomorphic agent sequence.

[0093] The second similar personality attribute tag is an attribute tag whose tag similarity with the second personality attribute tag is greater than a preset threshold.

[0094] In practice, the weight corresponding to a personality dimension can be weakened to serve as a second approximate personality attribute label. Specifically, since there are five personality dimensions, when the weight of any personality dimension is weakened, at least five anthropomorphic agents can be recalled as candidate anthropomorphic agents.

[0095] Step S7: Based on the label weight corresponding to the first personality attribute label and the label weight corresponding to the second personality attribute label, the recommendation weights of the candidate anthropomorphic agents in the candidate anthropomorphic agent sequence are updated to obtain an updated candidate anthropomorphic agent sequence.

[0096] Among them, the label weight is used to highlight that the selected anthropomorphic agent tends to be inclined to the first personality attribute label or the second personality attribute label. The label weight of the first personality attribute label and the label weight of the second personality attribute label can be 0.5 and 0.5. In particular, when there is a matching anthropomorphic agent for the first personality attribute label and a matching anthropomorphic agent for the second personality attribute label, the candidate anthropomorphic agent sequence includes 2 anthropomorphic agents. When there is no matching anthropomorphic agent for the first personality attribute label and a matching anthropomorphic agent for the second personality attribute label, the candidate anthropomorphic agent sequence includes 10 anthropomorphic agents. Therefore, the number of anthropomorphic agents in the candidate anthropomorphic agent sequence can be in the range of [2, 10]. The anthropomorphic agents in the candidate anthropomorphic agent sequence are set with similarity weights, and the similarity weights are represented by the similarity between the corresponding first approximate personality attribute label and the first personality attribute label, or the similarity between the second approximate personality attribute label and the second personality attribute label. For the anthropomorphic agent among the candidate anthropomorphic agents corresponding to the first approximate personality attribute label, the corresponding similarity weight is updated using the label weight corresponding to the first personality attribute label. For the anthropomorphic agent among the candidate anthropomorphic agents corresponding to the second approximate personality attribute label, the corresponding similarity weight is updated using the label weight corresponding to the second personality attribute label.

[0097] As an example, the candidate anthropomorphic agent sequence includes: anthropomorphic agent G1, anthropomorphic agent G2, anthropomorphic agent G3, anthropomorphic agent G4, anthropomorphic agent G5, anthropomorphic agent G6, anthropomorphic agent G7, anthropomorphic agent G8, anthropomorphic agent G9, and anthropomorphic agent G10. Among them, anthropomorphic agent G1, anthropomorphic agent G2, anthropomorphic agent G3, anthropomorphic agent G4, and anthropomorphic agent G5 are recalled anthropomorphic agents corresponding to the first approximate personality attribute label. Anthropomorphic agent G6, anthropomorphic agent G7, anthropomorphic agent G8, anthropomorphic agent G9, and anthropomorphic agent G10 are recalled anthropomorphic agents corresponding to the second approximate personality attribute label. Among them, the similarity weight corresponding to anthropomorphic agent G1 may be 0.9, the similarity weight corresponding to anthropomorphic agent G2 may be 0.82, the similarity weight corresponding to anthropomorphic agent G3 may be 0.80, the similarity weight corresponding to anthropomorphic agent G4 may be 0.78, the similarity weight corresponding to anthropomorphic agent G5 may be 0.77, the similarity weight corresponding to anthropomorphic agent G6 may be 0.92, the similarity weight corresponding to anthropomorphic agent G7 may be 0.91, the similarity weight corresponding to anthropomorphic agent G8 may be 0.90, the similarity weight corresponding to anthropomorphic agent G9 may be 0.89, and the similarity weight corresponding to anthropomorphic agent G10 may be 0.88. That is, the similarity weight sequence corresponding to the anthropomorphic agents included in the candidate anthropomorphic agent sequence may be [0.9, 0.82, 0.80, 0.78, 0.77, 0.92, 0.91, 0.90, 0.89, 0.88]. The label weight corresponding to the first personality attribute label can be 0.6, and the label weight corresponding to the second personality attribute label can be 0.4. The updated similarity weight matrix can then be [0.54, 0.492, 0.48, 0.468, 0.462, 0.368, 0.364, 0.36, 0.356, 0.352]. Finally, based on the updated similarity weights corresponding to the anthropomorphic agents, the anthropomorphic agents in the candidate anthropomorphic agent sequence are sorted in descending order to obtain an updated candidate anthropomorphic agent sequence.

[0098] Step S8: Select an anthropomorphic agent that meets the first screening condition from the updated candidate anthropomorphic agent sequence as the target anthropomorphic agent.

[0099] Among them, the first screening condition can be: the updated similarity weight corresponding to the anthropomorphic intelligent agent is the largest.

[0100] Step 105 : Determine the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic intelligent agent and the behavioral feature set.

[0101] In some embodiments, the execution subject may determine the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic intelligent agent and the set of behavioral characteristics.

[0102] The recommendation type may represent the data type of the information to be recommended. In practice, the data type may include text type, audio type, video type, and a mixed type consisting of at least one of the text type, audio type, and video type.

[0103] In practice, the above-mentioned execution subject can obtain the information to be recommended and the recommendation type corresponding to the information to be recommended by taking the behavioral feature set as the input of the target anthropomorphic intelligent agent.

[0104] In some optional implementations of some embodiments, the execution entity determines the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic agent and the behavioral feature set, including:

[0105] Step S1: According to the recommendation style corresponding to the personality attribute label, based on the above-mentioned target anthropomorphic intelligent agent and the pre-built recommendation information pool, recall the candidate recommendation information sequence.

[0106] Recommendation styles can include conservative and expansive. The conservative recommendation style only recommends information matching personality attribute labels (primary and secondary). The expansive recommendation style, in addition to the conservative style, also recommends information matching a certain percentage of other personality attribute labels. In practice, the target anthropomorphic agent can combine a set of behavioral features and, through vector recall, retrieve candidate recommendation sequences from a pre-built recommendation information pool.

[0107] Step S2: Select candidate recommendation information that meets the second screening condition from the candidate recommendation information sequence as the information to be recommended.

[0108] The second screening condition may be: the recommendation degree corresponding to the candidate recommendation information is greater than a preset recommendation threshold.

[0109] Step S3: Determine the recommendation type corresponding to the information to be recommended based on the behavioral characteristics corresponding to the second personality dimension, the behavioral characteristics corresponding to the third personality dimension, the behavioral characteristics corresponding to the fourth personality dimension in the behavioral characteristic set and a pre-trained recommendation type classification model.

[0110] The recommendation type classification model uses the behavioral characteristics corresponding to the second, third, and fourth personality dimensions in the behavioral characteristic set as model inputs and outputs the recommendation type. For example, the recommendation type classification model can be constructed using a convolutional neural network (CNN) as the backbone network, coupled with a recommendation type classifier (or multiple classifiers).

[0111] In practice, the second personality dimension can correspond to "Openness to experience," the third personality dimension can correspond to "Conscientiousness," and the fourth personality dimension can correspond to "Neuroticism." Therefore, this disclosure combines the second, third, and fourth personality dimensions to generate corresponding recommendation types, thereby measuring user content acceptance under different recommendation information display methods.

[0112] Step 106 : Fill in the template of the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain the target information to be recommended.

[0113] In some embodiments, the execution entity may fill in the template of the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain the target information to be recommended.

[0114] Among them, different credit limit recommendation types can be preset with recommendation templates. The recommendation template can include multiple filling positions and preset template content. In particular, each filling position can be pre-set with a filling rule.

[0115] In practice, the information to be recommended can be filled into the recommendation template by matching the filling rules to obtain the target content to be recommended.

[0116] In some optional implementations of some embodiments, the execution entity fills the template of the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain the target information to be recommended, including:

[0117] Step S1: extract key content from the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain a set of key content to be filled.

[0118] In practice, NER (Named Entity Recognition) can be used to extract key content from the recommended information, generating a set of key content to be filled. The extracted content after NER extraction is then filtered using the filling rules corresponding to the fill positions in the recommendation template to obtain the set of key content to be filled. This approach can streamline the fill content and avoid overwhelming the user's reading experience with excessive content.

[0119] Step S2: Fill the key content to be filled in the key content set to be filled into the recommendation template to obtain at least one initial recommendation information.

[0120] The at least one initial recommendation information is recommendation information obtained by filling in different template filling orders.

[0121] In practice, since there are multiple filling positions in the recommendation template and multiple key contents to be filled (key content sets to be filled), there are multiple combinations of filling positions and key contents to be filled, so at least one initial recommendation information can be obtained.

[0122] Step S3: for each piece of initial recommendation information in the at least one initial recommendation information, determine style evaluation information corresponding to the initial recommendation information.

[0123] In practice, the model interface of the LLM (Large Language Model) model can be called to perform style evaluation on the initial recommendation information and generate an evaluation score as the style evaluation information. In particular, to ensure the accuracy of the evaluation, a guide word can be set, such as "You are a designer. Please evaluate [the initial recommendation information] from multiple dimensions such as aesthetics, information effectiveness, and user acceptance, and form a (comprehensive) evaluation score." Specifically, training a dedicated style evaluation model for multimodal data is relatively expensive. Therefore, the present disclosure reduces costs by combining the LLM model to implement style evaluation for initial recommendation information.

[0124] Step S4: Filtering out the initial recommendation information whose corresponding style evaluation information satisfies the third filtering condition from the at least one initial recommendation information as the target information to be recommended.

[0125] Among them, the third screening condition may be: the evaluation score corresponding to the style evaluation information is the highest.

[0126] In some optional implementations of some embodiments, the above method further includes:

[0127] Step S1: Add the target information to be recommended to the information distribution queue.

[0128] The information distribution queue is an information forwarding queue corresponding to the target anthropomorphic intelligent agent. By setting the information distribution queue, the recommended information corresponding to the target anthropomorphic intelligent agent and requiring forwarding can be centrally managed.

[0129] Step S2: Determine the distribution mode.

[0130] The aforementioned distribution modes include active and passive distribution modes. The active distribution mode can represent the target anthropomorphic agent proactively pushing the recommended information. The passive distribution mode represents the target anthropomorphic agent recommending the recommended information after the target user refreshes their account.

[0131] In practice, the distribution mode can be dynamically adjusted according to the network load of the communication network where the anthropomorphic agent is located.

[0132] Step S3: In response to the target information to be recommended being located at the first position of the information distribution queue, according to the distribution mode, the target information to be recommended is sent to the target terminal corresponding to the target user through the target anthropomorphic agent.

[0133] The aforementioned embodiments of the present disclosure have the following beneficial effects: Through the anthropomorphic agent-based recommendation information generation method of some embodiments of the present disclosure, highly scalable and accurate information recommendation is achieved. Specifically, first, a set of behavior information corresponding to a target user is collected, where the target user is a user who has authorized permission to collect corresponding user behavior and for whom information recommendation is to be made. Second, multi-dimensional personality feature extraction is performed on the behavior information set to obtain a behavior feature set, where each behavior feature in the behavior feature set corresponds to a personality dimension. In practice, conventional behavior information extraction methods typically directly extract corresponding features based on the behavior type of the behavior information. However, this extraction method ignores the feature description from the perspective of personality dimensions, resulting in poor expressiveness of the extracted features. Second, based on the behavior feature set, a personality attribute label corresponding to the target user is determined. Next, a target anthropomorphic agent corresponding to the personality attribute label is matched, where the target anthropomorphic agent is an anthropomorphic agent that recommends information according to the recommendation style corresponding to the personality attribute label. Furthermore, based on the target anthropomorphic agent and the behavior feature set, the information to be recommended and the recommendation type corresponding to the information to be recommended are determined. In practice, conventional methods often directly recall recommended information after extracting features. However, due to the poor expressiveness of the extracted features and the lack of corresponding classification of different features, the accuracy of the recalled recommended information is poor. Based on this, the present disclosure determines the personality attribute label by combining behavioral characteristics, matches the corresponding target anthropomorphic intelligent agent, and recalls the corresponding recommended information based on the target anthropomorphic intelligent agent. This method realizes the recall of recommended information for users with the same or similar personality attributes. Compared with the "thousands of faces" recall method, the recall efficiency is significantly improved. In addition, due to the combination of personality attribute labels, anthropomorphic recommended information recall is realized, thereby improving the accuracy of the recalled recommended information. Finally, according to the recommendation template corresponding to the above recommendation type, the above information to be recommended is template-filled to obtain the target information to be recommended. In this way, the display flexibility of the recommended information is improved, thereby promoting the user's residence probability. In summary, the recommendation information generation method based on anthropomorphic intelligent agent realizes highly scalable and accurate information recommendation.

[0134] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a device for generating recommendation information based on an anthropomorphic intelligent agent. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the recommendation information generation device based on anthropomorphic intelligent agent can be specifically applied to various electronic devices.

[0135] like Figure 5As shown, the recommendation information generation device 500 based on anthropomorphic intelligent agent in some embodiments includes: a collection unit 501, a feature extraction unit 502, a first determination unit 503, a matching unit 504, a second determination unit 505 and a template filling unit 506. Among them, the collection unit 501 is configured to collect a behavior information set corresponding to the target user, wherein the above-mentioned target user is a user who has authorized the corresponding user behavior collection authority and is to be recommended information; the feature extraction unit 502 is configured to perform feature extraction of multi-dimensional personality dimensions on the above-mentioned behavior information set to obtain a behavior feature set, wherein each behavior feature in the above-mentioned behavior feature set corresponds to a personality dimension; the first determination unit 503 is configured to determine the personality attribute label corresponding to the above-mentioned target user based on the above-mentioned behavior feature set; the matching unit 504 is configured to match the target anthropomorphic intelligent agent corresponding to the above-mentioned personality attribute label, wherein the target anthropomorphic intelligent agent is an anthropomorphic intelligent agent that recommends information according to the recommendation style corresponding to the personality attribute label; the second determination unit 505 is configured to determine the information to be recommended and the recommendation type corresponding to the above-mentioned information to be recommended based on the above-mentioned target anthropomorphic intelligent agent and the above-mentioned behavior feature set; the template filling unit 506 is configured to perform template filling on the above-mentioned information to be recommended according to the recommendation template corresponding to the above-mentioned recommendation type to obtain the target information to be recommended.

[0136] It is understandable that the units described in the recommendation information generating device 500 based on anthropomorphic intelligent agent are similar to those in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the recommendation information generating device 500 based on anthropomorphic intelligent agent and the units included therein, and will not be described in detail here.

[0137] Reference below Figure 6 , which shows a structural schematic diagram of an electronic device (eg, a computing device) suitable for implementing some embodiments of the present disclosure. Figure 6 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure. Figure 6As shown, the computer device includes a processor, a memory and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory. The non-volatile storage medium may store an operating system and a computer program. The computer program includes program instructions, which, when executed, may enable the processor to execute any of the above methods. The processor is used to provide computing and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, may enable the processor to execute any of the above methods. The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 6 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present disclosure, and does not constitute a limitation on the computer device to which the solution of the present disclosure is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0138] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0139] In one embodiment, the processor is used to run a computer program stored in a memory to implement the following steps: collecting a behavior information set corresponding to a target user, wherein the target user is a user who has authorized the corresponding user behavior collection authority and is to be recommended information; performing multi-dimensional personality dimension feature extraction on the behavior information set to obtain a behavior feature set, wherein each behavior feature in the behavior feature set corresponds to a personality dimension; determining the personality attribute label corresponding to the target user based on the behavior feature set; matching a target anthropomorphic intelligent agent corresponding to the personality attribute label, wherein the target anthropomorphic intelligent agent is an anthropomorphic intelligent agent that recommends information according to the recommendation style corresponding to the personality attribute label; determining the information to be recommended and the recommendation type corresponding to the information to be recommended based on the target anthropomorphic intelligent agent and the behavior feature set; and filling in the template of the information to be recommended based on the recommendation template corresponding to the recommendation type to obtain the target information to be recommended.

[0140] An embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the method described above in the present disclosure.

[0141] The computer-readable storage medium may be an internal storage unit of the computer device described in the aforementioned embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a SmartMedia Card (SMC), a Secure Digital (SD) card, a flash memory card, etc., provided on the computer device.

[0142] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0143] The above descriptions are merely some preferred embodiments of the present disclosure and illustrate the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present disclosure is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for generating recommendation information based on an anthropomorphic agent, characterized in that: include: Collecting a set of behavior information corresponding to a target user, wherein the target user is a user who has authorized the corresponding user behavior collection permission and for whom information recommendation is to be made; Performing feature extraction of a multidimensional personality dimension on the behavior information set to obtain a behavior feature set, wherein each behavior feature in the behavior feature set corresponds to a personality dimension, including: Dividing the time distribution corresponding to the behavior information set into time windows according to a preset time granularity to obtain at least one time window; grouping the behavior information in the behavior information set according to the at least one time window to obtain a behavior information group set, wherein each behavior information in the behavior information group corresponds to the same time window; For each behavior information group in the behavior information group set, the following processing steps are performed: generating a local heat map matrix corresponding to the behavior information group according to the content type of the behavior information in the behavior information group; Determining, based on the behavior content and content type included in the behavior information, a content domain label, an expression factor, and an emotion factor corresponding to each behavior information in the behavior information group; Determining an interactivity factor corresponding to each piece of behavior information in the behavior information group based on content access rights, content exposure information, and content comments included in the behavior information; Performing heat map normalization and splicing on at least one local heat map matrix obtained in time sequence to obtain a heat map matrix as the behavioral feature corresponding to the first personality dimension; Project the content domain labels, expression factors, emotional factors, and interactivity factors corresponding to the behavioral information onto the matrix space corresponding to the heat map matrix, respectively, to obtain the behavioral characteristics corresponding to the second personality dimension, the third personality dimension, the fourth personality dimension, and the fifth personality dimension; Determining a personality attribute label corresponding to the target user based on the set of behavioral features, wherein the personality attribute label includes: a first personality attribute label and a second personality attribute label, wherein the first personality attribute label represents the personality attribute self-assessed by the target user using a personality attribute self-assessment form; and the second personality attribute label is a predicted personality attribute for the target user generated based on a pre-trained personality attribute label prediction model, the first personality attribute label, and the set of behavioral features, wherein the personality attribute label prediction model includes five parallel sub-convolutional neural networks for independently performing deep feature extraction on behavioral features corresponding to the five different personality attribute labels, a feature stacking layer, and a multi-classifier. Matching a target anthropomorphic agent corresponding to the personality attribute tag, wherein the target anthropomorphic agent is an anthropomorphic agent that recommends information according to the recommendation style corresponding to the personality attribute tag, wherein the anthropomorphic agent acts as a repeater to select and recommend recommended information to users with the same personality attribute; Determining, based on the target anthropomorphic intelligent agent and the set of behavioral characteristics, information to be recommended and a recommendation type corresponding to the information to be recommended; The information to be recommended is filled with a template according to a recommendation template corresponding to the recommendation type to obtain target information to be recommended.

2. The method according to claim 1, characterized in that Determining the personality attribute label corresponding to the target user based on the behavioral feature set includes: Determining whether a target personality attribute label exists, wherein the target personality attribute label is a personality attribute self-evaluated by the target user before the current moment; In response to the existence of the target personality attribute tag, determining whether the target personality attribute tag is valid, wherein the target personality attribute tag controls the corresponding tag validity through a tag validity period; In response to the target personality attribute label being valid, determining the target personality attribute label as the first personality attribute label; In response to the target personality attribute tag being invalid, initiating entry of a first personality attribute tag for the target user to determine the first personality attribute tag and update a corresponding tag validity period; The second personality attribute label is generated according to a pre-trained personality attribute label prediction model, the first personality attribute label, and the behavioral feature set.

3. The method according to claim 2, characterized in that The matching of the target anthropomorphic intelligent agent corresponding to the personality attribute label includes: Determining whether there is an anthropomorphic intelligent agent matching the first personality attribute label; In response to the existence of an anthropomorphic intelligent agent matching the first personality attribute label, adding the anthropomorphic intelligent agent matching the first personality attribute label as a candidate anthropomorphic intelligent agent to a candidate anthropomorphic intelligent agent sequence; In response to the absence of an anthropomorphic agent matching the first personality attribute label, recalling an anthropomorphic agent corresponding to a first approximate personality attribute label as a candidate anthropomorphic agent and adding it to a candidate anthropomorphic agent sequence, wherein the first approximate personality attribute label is an attribute label having a label similarity with the first personality attribute label greater than a preset threshold; Determining whether there is an anthropomorphic intelligent agent matching the second personality attribute label; In response to the existence of an anthropomorphic agent matching the second personality attribute label, adding the anthropomorphic agent matching the second personality attribute label as a candidate anthropomorphic agent to a candidate anthropomorphic agent sequence; In response to the absence of an anthropomorphic agent matching the second personality attribute label, recalling an anthropomorphic agent corresponding to a second approximate personality attribute label as a candidate anthropomorphic agent and adding the agent to a candidate anthropomorphic agent sequence, wherein the second approximate personality attribute label is an attribute label having a label similarity with the second personality attribute label greater than a preset threshold; updating the recommendation weights of the candidate anthropomorphic agents in the candidate anthropomorphic agent sequence according to the label weight corresponding to the first personality attribute label and the label weight corresponding to the second personality attribute label, to obtain an updated candidate anthropomorphic agent sequence; An anthropomorphic agent that meets the first screening condition is selected from the updated candidate anthropomorphic agent sequence as the target anthropomorphic agent.

4. The method according to claim 3, characterized in that The determining, based on the target anthropomorphic intelligent agent and the set of behavioral characteristics, the information to be recommended and the recommendation type corresponding to the information to be recommended includes: Recalling candidate recommendation information sequences according to the recommendation style corresponding to the personality attribute label, based on the target anthropomorphic intelligent agent and a pre-built recommendation information pool; Selecting candidate recommendation information that meets a second screening condition from the candidate recommendation information sequence as the information to be recommended; Determine the recommendation type corresponding to the information to be recommended based on the behavioral characteristics corresponding to the second personality dimension, the behavioral characteristics corresponding to the third personality dimension, the behavioral characteristics corresponding to the fourth personality dimension in the behavioral characteristic set and a pre-trained recommendation type classification model.

5. The method according to claim 4, characterized in that Filling the template of the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain target information to be recommended includes: Extracting key content from the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain a set of key content to be filled; Filling the recommendation template with the key content to be filled in the key content set to be filled, to obtain at least one initial recommendation information, wherein the at least one initial recommendation information is the recommendation information obtained by filling in different templates in a filling order; For each piece of initial recommendation information in the at least one initial recommendation information, determining style evaluation information corresponding to the initial recommendation information; Initial recommendation information whose corresponding style evaluation information satisfies a third filtering condition is filtered out from the at least one initial recommendation information as the target information to be recommended.

6. The method according to claim 5, characterized in that The method further comprises: Adding the target information to be recommended to an information distribution queue; Determining a distribution mode, wherein the distribution mode includes: an active distribution mode and a passive distribution mode; In response to the target information to be recommended being located at the first position of the information distribution queue, the target information to be recommended is sent to a target terminal corresponding to the target user through the target anthropomorphic agent according to the distribution mode.

7. A recommendation information generation device based on an anthropomorphic intelligent agent, characterized in that: include: A collection unit is configured to collect a set of behavior information corresponding to a target user, wherein the target user is a user who has authorized the corresponding user behavior collection permission and for whom information recommendation is to be made; The feature extraction unit is configured to extract features of a multidimensional personality dimension from the behavior information set to obtain a behavior feature set, wherein each behavior feature in the behavior feature set corresponds to a personality dimension, including: Dividing the time distribution corresponding to the behavior information set into time windows according to a preset time granularity to obtain at least one time window; grouping the behavior information in the behavior information set according to the at least one time window to obtain a behavior information group set, wherein each behavior information in the behavior information group corresponds to the same time window; For each behavior information group in the behavior information group set, the following processing steps are performed: generating a local heat map matrix corresponding to the behavior information group according to the content type of the behavior information in the behavior information group; Determining, based on the behavior content and content type included in the behavior information, a content domain label, an expression factor, and an emotion factor corresponding to each behavior information in the behavior information group; Determining an interactivity factor corresponding to each piece of behavior information in the behavior information group based on content access rights, content exposure information, and content comments included in the behavior information; Performing heat map normalization and splicing on at least one local heat map matrix obtained in time sequence to obtain a heat map matrix as the behavioral feature corresponding to the first personality dimension; Project the content domain labels, expression factors, emotional factors, and interactivity factors corresponding to the behavioral information onto the matrix space corresponding to the heat map matrix, respectively, to obtain the behavioral characteristics corresponding to the second personality dimension, the third personality dimension, the fourth personality dimension, and the fifth personality dimension; a first determining unit configured to determine a personality attribute label corresponding to the target user based on the set of behavioral characteristics, wherein the personality attribute label includes: a first personality attribute label and a second personality attribute label, wherein the first personality attribute label represents a personality attribute self-assessed by the target user using a personality attribute self-assessment form, and the second personality attribute label is a predicted personality attribute for the target user generated based on a pre-trained personality attribute label prediction model, the first personality attribute label, and the set of behavioral characteristics, wherein the personality attribute label prediction model includes five parallel sub-convolutional neural networks for independently performing deep feature extraction on behavioral characteristics corresponding to five different personality attribute labels, a feature stacking layer, and a multi-classifier; a matching unit configured to match a target anthropomorphic agent corresponding to the personality attribute label, wherein the target anthropomorphic agent is an anthropomorphic agent that recommends information according to the recommendation style corresponding to the personality attribute label, wherein the anthropomorphic agent acts as a repeater to select and recommend recommended information to users with the same personality attribute; a second determining unit configured to determine, based on the target anthropomorphic agent and the set of behavioral characteristics, information to be recommended and a recommendation type corresponding to the information to be recommended; The template filling unit is configured to perform template filling on the information to be recommended according to the recommendation template corresponding to the recommendation type to obtain target information to be recommended.

8. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer-readable medium, characterized in that A computer program is stored thereon, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images