CN112417275A

CN112417275A - Information providing method, device storage medium and electronic equipment

Info

Publication number: CN112417275A
Application number: CN202011290242.0A
Authority: CN
Inventors: 邢旺; 钟超
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2020-11-17
Filing date: 2020-11-17
Publication date: 2021-02-26

Abstract

The specification discloses an information providing method, an information providing device storage medium and electronic equipment, wherein according to the degree of association between the behavior of a user and target information to be provided, the association weight between any information characteristic and any behavior characteristic is determined, the behavior characteristics are subjected to association weighting by the association weight, different user historical behaviors are realized, different association weights are given according to the correlation between the behaviors and the target information, so that the speaking weight for predicting the target information is larger for the user behavior which is more closely related to the target information, and the accuracy of prediction is improved.

Description

Information providing method, device storage medium and electronic equipment

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method, an apparatus storage medium, and an electronic device for providing information.

Background

As more and more information is produced and distributed with the popularity of the internet, the explosive growth of information has made it increasingly difficult for users to obtain information that meets preferences. Personalized recommendation technology has become an indispensable basic technology in the internet field.

In the existing personalized recommendation technology, machine learning models for predicting preference information of users exist, preference prediction is often performed on the basis of user figures obtained according to user behaviors, but the accuracy of preference prediction obtained by the method is low.

Disclosure of Invention

The present specification provides an information providing method and apparatus to partially solve the above problems in the prior art.

The technical scheme adopted by the specification is as follows:

the present specification provides an information providing method including:

determining target information to be provided and various behavior data of a user;

extracting information features from each target information; extracting behavior characteristics from each behavior data; inputting all information characteristics and all behavior characteristics into a pre-trained information providing model, wherein the information providing model is used for determining an information providing mode of all target information;

for each input information characteristic, determining the association weight of the information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic according to the association weight to determine each behavior characteristic after the association weighting;

determining the target recommendation degree of the target information output by the information providing model according to the behavior characteristics after the associated weighting;

and determining an information providing mode of each target information according to the target recommendation degree of each target information, and providing at least part of target information for the user in the information providing mode.

Optionally, taking the information features extracted from each target information as first information features;

before determining the target recommendation degree of the target information output by the information providing model, the method further comprises:

determining each relevant information corresponding to the target information;

extracting information characteristics from each piece of relevant information corresponding to the target information to serve as second information characteristics;

inputting each information characteristic and each behavior characteristic into a pre-trained information providing model, which specifically comprises the following steps:

and inputting each first information characteristic, each second information characteristic and each behavior characteristic into a pre-trained information providing model.

Optionally, determining, by the information provision model, the associated weight between the information feature and each behavior feature specifically includes:

aiming at the input first information characteristic, determining a first association weight of the first information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic by using the first association weight to determine each behavior characteristic after the association weighting;

and aiming at the input second information characteristic, determining a second association weight of the second information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic by using the second association weight to determine each behavior characteristic after the association weighting.

Optionally, the information providing model comprises at least two sub-networks;

inputting each first information characteristic, each second information characteristic and each behavior characteristic into a pre-trained information providing model, and specifically comprising:

inputting each first information feature and each first behavior feature into a first subnetwork; inputting each second information characteristic and each second behavior characteristic into a second subnetwork; wherein the first behavior feature is a behavior feature extracted from behavior data in which the target information is an object, and the second behavior feature is a behavior feature extracted from behavior data in which the related information is an object.

Optionally, determining, by the information provision model, an association weight between the information feature and each behavior feature, and performing association weighting on the corresponding behavior feature by using the association weight to determine each behavior feature after the association weighting, specifically includes:

aiming at input first information characteristics, determining first association weights of the first information characteristics and each first behavior characteristic through the information providing model, performing association weighting on the corresponding first behavior characteristics by using the first association weights, and determining each first behavior characteristic after the association weighting;

and aiming at the input second information characteristics, determining second association weights of the second information characteristics and the second behavior characteristics through the information providing model, performing association weighting on the corresponding second behavior characteristics through the second association weights, and determining the second behavior characteristics after the association weighting.

Optionally, before determining the target recommendation degree of the target information output by the information providing model, the method further includes:

determining a first comprehensive characteristic of the target information according to the second behavior characteristics after the association weighting;

determining a second comprehensive characteristic of the target information according to each first behavior characteristic and each first comprehensive characteristic of the target information;

and determining the target recommendation degree of the target information according to the second comprehensive characteristics of the target information.

Optionally, providing at least part of the target information to the user in the information providing manner specifically includes:

and providing at least part of target information and at least part of related information corresponding to each provided target information for the user in the information providing mode.

Optionally, the target information specifically includes: information of the merchant;

the relevant information corresponding to the target information specifically includes: information about the goods offered by the merchant.

Optionally, the pre-training information providing model specifically includes:

determining an information providing model to be trained, and training samples for training the information providing model, the training samples including: target information and historical behavior data of users;

extracting information features from each target information; behavior characteristics are extracted from each behavior data and input into the information providing model to be trained;

determining the association weight of each information feature to each behavior feature, performing association weighting on the corresponding behavior feature according to the association weight, determining each behavior feature after association weighting, and determining the target recommendation degree of target information corresponding to the information feature according to each behavior feature after association weighting;

determining the labeling information of each target information according to the behavior data;

and determining the loss of the information providing model according to the target recommendation degree and the labeling information of each target information, and training the information providing model by taking the loss minimization as a target.

An information providing apparatus provided by the present specification includes:

the acquisition module is used for determining each target information to be provided and each behavior data of the user;

the characteristic input module is used for extracting information characteristics from each target information; extracting behavior characteristics from each behavior data; inputting all information characteristics and all behavior characteristics into a pre-trained information providing model, wherein the information providing model is used for determining an information providing mode of all target information;

the characteristic weighting module is used for determining the association weight of each input information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristics according to the association weight to determine each behavior characteristic after the association weighting;

the target recommendation module is used for determining the target recommendation degree of the target information output by the information providing model according to the associated weighted behavior characteristics;

and the information providing module is used for determining the information providing mode of each target information according to the target recommendation degree of each target information and providing at least part of the target information for the user in the information providing mode.

The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described information providing method.

The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the above information providing method when executing the program.

The technical scheme adopted by the specification can achieve the following beneficial effects:

in the information providing method provided by the present specification, according to the degree of association between the behavior of the user and the target information to be provided, the association weight between any information feature and any behavior feature is determined, the behavior feature is subjected to association weighting by the association weight, so that different user historical behaviors are realized, and different association weights are given according to the correlation between the behavior and the target information, so that a larger speaking weight for predicting the target information is given to the user behavior having a closer relationship with the target information, and compared with the case of non-association weights, the preference of the user to the target information is predicted according to each user behavior, and the accuracy is improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the specification and are incorporated in and constitute a part of this specification, illustrate embodiments of the specification and together with the description serve to explain the specification and not to limit the specification in a non-limiting sense. In the drawings:

FIG. 1 is a schematic flow chart of a model training method in the present specification;

FIG. 2 is a schematic flow chart of another model training method described herein;

FIG. 3 is a schematic diagram of an information providing model according to the present disclosure;

fig. 4 is a schematic flow chart of an information providing method in the present specification;

FIGS. 5A-5C are schematic diagrams of three related information providing methods of the present disclosure;

FIG. 6 is a schematic view of an information providing apparatus provided herein;

fig. 7 is a schematic diagram of an electronic device corresponding to fig. 4 provided in the present specification.

Detailed Description

In order to make the objects, technical solutions and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort belong to the protection scope of the present specification.

The information providing method described in this specification is intended to determine an information providing scheme and provide the information providing scheme to a user based on collected behavior data that can represent information preferences of the user, or to say, to select from a large amount of information and provide the selected information to the user in a specific manner, that is, to provide personalized information recommendation.

The technical solutions provided by the embodiments of the present description are described in detail below with reference to the accompanying drawings.

It should be noted that, since the embodiments of information provision described in this specification are based on the machine learning model used in the embodiments, that is, the improvement of the information provision model in this specification, in this context, the specification also provides an embodiment for training the information provision model.

The first embodiment is as follows:

fig. 1 is a schematic diagram of a model training process provided in an embodiment of the present specification, including:

s100: determining an information providing model to be trained, and training samples for training the information providing model, the training samples including: target information and historical behavior data of users.

In this specification, the information providing model to be trained (hereinafter referred to as the model to be trained) has the trained targets, that is, the functions to be expected to be realized: at least part of the object information is determined from the input object information, and is provided to the user in the determined providing mode. For example, the target information may be filtered and the filtering results may be sorted, but the present specification does not limit the filtered target information in any way, and may provide the target information in a non-ordered way.

The training samples used in this specification to train the model to be trained include: target information and historical behavior data of users. It can be known from the above description that the target information is used as the input of the model to be trained, and after the model to be trained is calculated and screened in a preset manner according to the historical behavior data of each user, the screened target information is also used as the output of the model to be trained.

Various types of information may be targeted according to the usage scenario. Generally speaking, in the field of electronic commerce, information of merchants or commodities can be used as target information; and for the content platform, information of the creator or the creator's work may be taken as target information.

It should be noted that, because each user often does not have the same preference, although the training sample includes behavior data in history of several users, for a single training (i.e., a single input and output of the model to be trained), only the behavior data in history of a single user is selected as the sample, each target information in the sample is labeled according to the behavior data of the single user, and a preference prediction result for the user is output.

Before training, the information providing model to be trained can be an initialized model, and all parameters in the model are preset or randomly generated; or a model that has been historically trained but failed to meet the usage requirements, in which case the model may be trained again. That is to say, the information providing model that needs to be trained to improve the model information providing performance may be the information providing model to be trained in this specification.

S102: extracting information features from each target information; and extracting behavior characteristics from each behavior data, and inputting the behavior characteristics into the information providing model to be trained.

The present specification does not limit the method of extracting information features from each target information and the method of extracting behavior features from each behavior data, and therefore, any existing feature extraction method may be selected and the extracted information features and behavior features may be input to the information provision model to be trained.

It should be noted that, in this specification, it is not limited that each target information/behavior data before extraction corresponds to one extracted information feature/behavior feature, and each behavior data may also be extracted as one behavior feature, and for convenience of description, each target information/behavior data before and after feature extraction corresponds to one information feature/behavior feature as an example in this specification.

S104: for each information feature, determining an associated weight of the information feature to each behavior feature.

Generally speaking, each user may have a plurality of behaviors, which are reflected in the embodiment of the present specification, that is, a large amount of behavior data may be acquired for each user, but not every behavior data may affect the preference of the user for the target information, for example, when the target information and the target of the user behavior are commodities, the acquired behavior data includes behavior data of commodities such as potato chips, earrings, cat food and the like for one user, and when the target commodity is a cat litter, generally, the preference of the user for the cat litter is less correlated with the behavior data of the user for the potato chips and the earrings, and is closely correlated with the behavior data of the user for the cat food.

Based on the information characteristic, for each information characteristic, according to the degree of closeness of association between the information characteristic and each behavior data, determining an association weight for the information characteristic for each behavior data, and characterizing the degree of closeness of association between the behavior data and the information characteristic, wherein the greater the association weight, the more closely the association between the behavior data and the information characteristic is, and conversely, the smaller the association weight, the more distant the association between the behavior data and the information characteristic is. In one embodiment of the present description, an attention mechanism may be employed to determine the associated weights between the information features and the behavior features in the form of attention weight assignment.

It is understood that, not only for one information feature, different behavior features have different degrees of association with the information feature, but also for different information features, the same behavior feature has different degrees of association with different information features, that is, any information feature has different degrees of association with any behavior feature, and the weights of association between the information features are different.

S106: and performing association weighting on the corresponding behavior characteristics according to the association weight, determining the behavior characteristics after each association weighting, and determining the target recommendation degree of the target information corresponding to the information characteristics according to each behavior characteristic after the association weighting.

It should be noted that the target recommendation degree of the target information output by the model to be trained can indicate the preference degree of the user for the target information considered by the model to be trained, and generally speaking, in the internet field, the prediction of the click rate usually represents the prediction of the preference degree of the user.

After the target recommendation degree of each target information is determined, the loss of the model to be trained can be directly determined according to the labeling information and the target recommendation degree of each target information; and determining each target information to be provided according to the target recommendation degree of each target information, and determining the loss of the model to be trained according to the marking information and the target recommendation degree of each target information to be provided.

When each behavior data is extracted as a behavior feature of one multi-dimension, it can be understood that the behavior feature may be subjected to correlation weighting, or may be a multi-dimension vector (or matrix) composed of a plurality of correlation weights.

S108: and determining the labeling information of each target information according to the behavior data.

The present specification only provides an exemplary labeling method, and any existing labeling method may be specifically adopted to label the target information, and details of other labeling methods are not repeated.

The historical behavior data of the user used in the training includes the degree of preference (e.g., click rate) of the user to each target information in the history, and according to the target recommendation degree of each target information, the target recommendation degree reflects the degree of preference (e.g., click rate prediction result) of the user to each target information considered by the model to be trained, in this scenario, the behavior data may be used as a label for the target information corresponding to the behavior data.

S110: and determining the loss of the information providing model according to the target recommendation degree and the labeling information of each target information, and training the information providing model by taking the loss minimization as a target.

And respectively determining the loss of the model to be trained according to the prediction results respectively output by the model to be trained and the labels of the corresponding samples, wherein the smaller the loss is, the closer the prediction result output by the model to be trained is to the label, and otherwise, the larger the difference between the prediction result and the label is. Gradients can be determined from the losses and a gradient descent algorithm can be used to adjust model parameters in the model to be trained. And when the adjusting times of the model parameters reach a preset time threshold value and/or the loss is less than a preset loss threshold value, the model to be trained is considered to be an information providing model after training is completed.

In the above example, for each target information, the real click rate of the user on the target information may be used as the label of the target information, the loss of the model to be trained is determined according to the label information and the click rate prediction result (i.e., the target recommendation degree) of the model to be trained on the target information, and each model parameter in the model to be trained is adjusted by adopting a gradient descent method until the adjustment times of the model parameters reach the preset time threshold, and/or the loss is smaller than the preset loss threshold.

The above describes a process of performing a single training on the model to be trained, where the single training refers to inputting and outputting the model to be trained once, and determining a loss according to the labeled information. However, in general, a model that can achieve the required prediction accuracy often requires multiple training sessions. For the purpose of enriching training samples, the behavior data of the same user is not selected as a training sample for each training of the model to be trained in the present specification.

Furthermore, how the user prefers the target information often depends not only on the target information itself, love and crow, and other information related to the target information, but also may include the user's preference for the target information.

For example, when the target information is a merchant, the preference of the user for the goods sold by the merchant is generally similar to the preference of the user for the merchant; when the target information is the creator, for example, the director, whether the user prefers the director largely depends on the movie work conducted by the director, so in this specification, the model to be trained may be trained by using the related information related to the target information.

It can be seen that there is a corresponding relationship between related information and target information, that is, related information only exists as related information of the target information, and also only affects the target information having the corresponding relationship, but there are several related information corresponding to one target information, and whether one related information may correspond to several target information.

In this situation, for the purpose of distinction, information features extracted from the target information are used as first information features; and taking the information features extracted from the related information as second information features, and inputting the first information features and the second information features into the model to be trained.

For each information feature (including the first information feature and the second information feature), determining an association weight between the information feature and each behavior feature, and performing association weighting on each behavior feature according to the determined association weight, specifically, the association weight determined by the first information feature and the behavior feature may be used as the first association weight; and the second information characteristic and the behavior characteristic are determined to be associated weight, and the second associated weight is used as the second associated weight.

The behavior features may be associated with each information feature without distinction, but when the behavior object of the behavior data can be determined, the behavior features extracted from the behavior data may be determined as the first behavior feature or the second behavior feature according to the behavior object of each behavior data, specifically, the first behavior feature may be a behavior feature extracted from the behavior data in which the target information is an object, and the second behavior feature may be a behavior feature extracted from the behavior data in which the related information is an object.

In this situation, the process of determining the association weight between the information feature and the behavior feature in steps S104 to S106 shown in fig. 1 and determining the target recommendation degree of the target information according to each behavior feature after the association weighting may be as shown in fig. 2.

Fig. 2 is a process of determining an association weight between information features and behavior features and determining a target recommendation degree of target information according to each behavior feature after the association weighting, which is provided in an embodiment of this specification, and specifically includes the following steps:

s200: and determining each first behavior characteristic after the associated weighting.

A first association weight between the first information feature (the information feature extracted from the target information) and each first behavior feature (that is, the behavior feature whose corresponding behavior data is the target information) is determined, and the first association weight is used to associate and weight the corresponding first behavior feature, thereby determining each first behavior feature after association and weighting.

S202: and determining each second behavior characteristic after the association weighting.

And determining a second association weight between each second information feature (the information feature extracted from the target information) and each second behavior feature (that is, the behavior feature of which the corresponding behavior data is the object of the related information), and performing association weighting on the corresponding second behavior feature by using the second association weight to determine each second behavior feature after the association weighting.

S204: and determining the target recommendation degree of the target information according to the first behavior characteristics after the association and the weighting and the second behavior characteristics after the association and the weighting.

Determining a first comprehensive characteristic of the target information according to the second behavior characteristics after the association weighting; determining a second comprehensive characteristic of the target information according to each first behavior characteristic and each first comprehensive characteristic of the target information; and determining the target recommendation degree of the target information according to the second comprehensive characteristics of the target information.

In step S204, the weighted second behavior features may be associated with each other by using any existing algorithm to determine a first comprehensive feature of the target information, such as any pooling method; each first behavior feature and the first comprehensive feature of the target information may adopt any existing algorithm to determine the second comprehensive feature of the target function, such as a Concat function, and how to determine the first comprehensive feature of the target information and the second comprehensive feature of the target function in the embodiments of the present specification is not described again.

It should be noted that, in steps S200 to S202, as to how to determine each first behavior feature and each second behavior feature after the association weighting by using the model to be trained, the association weights between the information feature and each behavior feature may be determined and associated and weighted sequentially for each information feature, or the association weights between the information feature and each behavior feature may be calculated and associated and weighted simultaneously and in parallel in a plurality of threads of a computer for each information feature.

For example, the model to be trained includes at least two sub-networks, which can be used as a first sub-network and a second sub-network, respectively, for differentiation. Wherein the first information characteristics, each first behavior characteristic, are input into a first subnetwork; the second information characteristics and the second behavior characteristics are input into the second sub-network.

In a first sub-network, determining first association weights of the first information features and the first behavior features, performing association weighting on the corresponding first behavior features by using the first association weights, and determining the first behavior features after the association weighting; in the second sub-network, for each second information characteristic, determining a second association weight of the second information characteristic and each second behavior characteristic, performing association weighting on the corresponding second behavior characteristic by using the second association weight, and determining each second behavior characteristic after the association weighting.

That is, steps S200 to S202 are completed such that the two subnetworks respectively calculate the features corresponding to the target information and the related information.

In this situation, in step S204, the first comprehensive characteristic of the target information may be determined according to the second behavior characteristics output by the second sub-network and weighted by the association corresponding to the second information characteristics; determining a second comprehensive characteristic of the target information according to the first behavior characteristics output by the first sub-network and weighted by the association and the first comprehensive characteristic of the target information; and determining the target recommendation degree of the target information according to the second comprehensive characteristics of the target information.

The structure of the model to be trained may be as shown in fig. 3, where in the subnetwork 1, an association weight 1A between the information feature 1 and the behavior feature a is determined, and the behavior feature a is subjected to association weighting by the association weight 1A, so as to obtain an association weighted behavior feature 1A; in the sub-network 2, the association weights 2B and 3B between the information features 2 and 3 and the behavior feature B are respectively determined, the behavior feature B is subjected to association weighting according to the association weights 2B and 3B to obtain association weighted behavior features 2B and 3B, a first comprehensive feature of the target information is determined according to the association weighted behavior features 2B and 3B, and a second comprehensive feature of the target information is determined according to the first comprehensive feature and the association weighted behavior feature 1A.

In general, besides the target, the related information, and the behavior data, there are other information that can determine the preference degree of the user for the target information, for example, profile information of the user, including gender, age, and the like, which can also be extracted as features and participate in the prediction of the preference of the user for the target information in the form of the information features 4 or the behavior features C as shown in fig. 3.

Further, the target function of the model to be trained may further include, after determining the providing manner of the target information, providing at least part of the relevant information corresponding to the target information for each of the provided target information.

The present specification does not limit how to train the model to be trained to provide relevant information corresponding to each target information, but provides three methods as examples:

the first method comprises the following steps: and determining each piece of relevant information corresponding to the target information aiming at each piece of provided target information, and determining the prediction result of each piece of relevant information according to each second behavior characteristic after the relevant weighting determined in the second step of each piece of relevant information.

And the second method comprises the following steps: for each determined and provided target information and each relevant information corresponding to each target information, the identities of the target information and the relevant information are exchanged, that is, for each relevant information, the relevant information is taken as the target information, the target information corresponding to the relevant information is taken as the relevant information, each behavior characteristic after the associated weighting is determined according to the exchanged target information and the relevant information, and the target recommendation degree of each exchanged target information (namely, the relevant information before the exchange) is determined according to each behavior characteristic after the associated weighting.

And aiming at each piece of target information to be provided before exchange, determining a prediction result of each piece of relevant information according to the target recommendation degree of each piece of relevant information corresponding to the target information and the target recommendation degree of each piece of relevant information.

And the third is that: and inputting the second comprehensive characteristic of the target information determined by the method into another fully-connected layer which is different from the fully-connected layer of the model to be trained so as to obtain a new model to be trained, wherein the target function of the new model to be trained is to determine the prediction result of each piece of relevant information.

Determining the loss of the model to be trained according to the prediction result of each piece of relevant information determined by any one of the methods and the marking information of the relevant information determined by the behavior data, and adjusting each model parameter in the model to be trained by taking the loss minimization as a target.

In addition, the present specification further provides a training method, which determines a loss of the model to be trained based on a result of predicting the target information and the related information by the model to be trained and label information of the target information and the behavior information determined according to the behavior data, and adjusts each model parameter in the model to be trained with the loss minimized as a target, thereby improving accuracy of predicting the target information and the related information by the model to be trained.

Based on the information providing model trained in the first embodiment, the present specification provides a corresponding information providing method, as in the second embodiment.

Example two:

fig. 4 is a schematic view of an information providing process provided in an embodiment of the present specification, including:

s400: and determining each target information to be provided and each behavior data of the user.

The usage scenario of the information providing model is determined, and various pieces of information that are likely to be provided to the user are targeted information. A user receiving the information provided by the information providing model is determined, and various behavior data of the user is determined.

In the embodiments of the present specification, it is considered that information providing is completed once each time information to be provided and a providing method are determined, and for convenience of description, each embodiment in a process of providing information once will be described below.

S402: extracting information features from each target information; extracting behavior characteristics from each behavior data; and inputting the information features and behavior features into a pre-trained information providing model.

The information providing model is a model trained by the model training method shown in fig. 1 and used for determining the information providing mode of each target information.

Extracting information characteristics from each target information by adopting the same method as that in the figure 1; it should be noted that, this specification does not limit the one-to-one correspondence relationship between the target information and the information features, and also does not limit the one-to-one correspondence relationship between the behavior data and the behavior features, and as can be understood by those skilled in the art, it is a common feature extraction means to combine each row of behavior data into a behavior sequence and extract the corresponding behavior features from the behavior sequence.

S404: and determining the associated weight of each information characteristic and each behavior characteristic by the information providing model for each input information characteristic.

The information providing model trained by the method shown in fig. 1 determines the associated weight of each information feature and each behavior feature.

S406: carrying out association weighting on the corresponding behavior characteristics according to the association weights, and determining each behavior characteristic after association weighting;

and performing association weighting on the corresponding behavior characteristics according to each determined association weight. It should be noted that, for each behavior feature, if a plurality of association weights are determined for the behavior feature and a plurality of information features, the behavior feature needs to be subjected to association weighting for a plurality of times.

However, in this case, the information feature only includes the information feature of the target information, and the behavior feature, which can be understood, is only performed once to perform the association weighting corresponding to the target information with the association weighting determined by the information feature.

S408: determining the target recommendation degree of the target information output by the information providing model according to the behavior characteristics after the associated weighting;

the description does not limit how to determine the target recommendation degree according to each behavior feature after the association weighting, and may be directly determined according to the behavior feature after the association weighting with the largest association weight of each information feature, or may be to sum up the behavior features after the association weighting of each information feature, and the description is not repeated for how to determine the target recommendation degree according to the behavior feature after the association weighting.

S410: and determining an information providing mode of each target information according to the target recommendation degree of each target information, and providing at least part of target information for the user in the information providing mode.

In this embodiment of the present specification, how to determine an information providing manner of each target information according to a target recommendation degree of each target information is not limited, and for example, the ranking of each target information may be determined according to the target recommendation degree, and N target information before the ranking is selected and provided to the user by using the ranking of each target recommendation degree as an order, where N is a preset ranking threshold.

Generally speaking, in the internet field, the prediction of the click rate usually represents the prediction of the preference degree of the user, and in this situation, the predicted click rate of the target information may be determined according to the recommendation degree of the target information, or the recommendation degree of the target information is the predicted click rate of the target information, so that it can be considered that the higher the target recommendation degree of the target information is, the more confident the user who receives the target information clicks on the target information, thereby realizing the personalized information provision.

As can be seen from the method described in fig. 4, the present specification provides an information providing method, in which an association weight between any information feature and any behavior feature is determined according to the degree of association between the behavior of a user and target information to be provided, the behavior features are associated and weighted by the association weight, so as to implement different user historical behaviors, and different association weights are given according to the correlation between the behaviors and the target information, so that a larger speaking weight for predicting the target information is given to a user behavior having a closer relationship with the target information, and compared with a case of a non-association weight, the user's preference for the target information is predicted according to each user behavior, and the accuracy is improved.

Fig. 4 in this specification illustrates an exemplary method for determining the target recommendation degree of one target information by using the information providing model, but since there are several target information to be provided, it is obviously necessary to calculate the target recommendation degree of each target information, but since the same information providing model can only calculate the target recommendation degree of one target information at the same time, the method shown in fig. 4 can be sequentially used to calculate the target recommendation degree of each target information, and the information providing model can also be run in multiple threads at the same time, that is, the target recommendation degrees of multiple target information are calculated in multiple threads at the same time.

On the other hand, the information providing model used in the information providing method provided in the present specification has a model structure as shown in fig. 3, and therefore, the present specification also provides the following information providing method based on this model.

The target information to be provided can be determined according to the target information, and the information features extracted from the target information can be used as first information features; determining each relevant information corresponding to the target information; extracting information characteristics from each piece of relevant information corresponding to the target information to serve as second information characteristics; and inputting each first information characteristic, each second information characteristic and each behavior characteristic into a pre-trained information providing model.

For distinguishing, the association weight determined by the first information characteristic and the behavior characteristic can be used as a first association weight; and the second information characteristic and the behavior characteristic are determined to be associated weight, and the second associated weight is used as the second associated weight. Aiming at the input first information characteristic, determining a first association weight of the first information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic by using the first association weight to determine each behavior characteristic after the association weighting; and aiming at the input second information characteristic, determining a second association weight of the second information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic by using the second association weight to determine each behavior characteristic after the association weighting.

The behavior features may also be determined as a first behavior model and a second behavior model according to the behavior data from which the behavior features are derived, the first behavior features being behavior features extracted from the behavior data targeting the target information, and the second behavior features being behavior features extracted from the behavior data targeting the related information.

Based on the information provision model of the structure shown in fig. 3, the association weight between the first information feature and the first behavior feature and the association weight between the second information feature and the second behavior feature may be calculated using at least two subnets, respectively, and each of the first information feature and each of the first behavior features may be input to the first subnetwork; the second information characteristics and the second behavior characteristics are input into a second subnetwork. Aiming at input first information characteristics, determining first association weights of the first information characteristics and each first behavior characteristic through the information providing model, performing association weighting on the corresponding first behavior characteristics by using the first association weights, and determining each first behavior characteristic after the association weighting; and aiming at the input second information characteristics, determining second association weights of the second information characteristics and the second behavior characteristics through the information providing model, performing association weighting on the corresponding second behavior characteristics through the second association weights, and determining the second behavior characteristics after the association weighting.

After the first behavior features after the association weighting and the second behavior features after the association weighting are calculated, the target recommendation degree of the target information can be calculated according to the first behavior features after the association weighting and the second behavior features after the association weighting, and the first comprehensive features of the target information are determined according to the second behavior features after the association weighting; determining a second comprehensive characteristic of the target information according to each first behavior characteristic and each first comprehensive characteristic of the target information; and determining the target recommendation degree of the target information according to the second comprehensive characteristics of the target information.

Furthermore, after the providing mode of the target information is determined, for each provided target information, at least part of related information corresponding to the target information is also provided.

In the following, the method of providing the relevant information will be described by taking the target information as the information of the merchant and the relevant information as the dish information of the merchant as an example.

Fig. 5A to 5C are schematic diagrams of three information providing manners of this specification, where 5A is a schematic diagram of providing only relevant information (merchants), and merchants a to C are merchants selected by the method of this specification and provided to the user in a ranking manner according to recommendation degrees; fig. 5B is an information providing manner for providing dish information on the basis of providing merchant information, in which A, B merchants are two merchants selected by the method in the embodiment of the present specification, and dishes 1 to 6 are reserved dishes belonging to A, B merchants selected by the embodiment of the present specification, where dishes 1 to 3 belong to a merchant a and dishes 4 to 6 belong to a merchant B; fig. 5C is another information providing manner for providing dish information on the basis of providing merchant information, where A, B merchants are two merchants selected by the method in the embodiment of the present specification, dishes 1 to 2 are reserved dishes belonging to A, B merchants selected by the embodiment of the present specification, and dishes 3 to 4 belong to a merchant a.

FIG. 5B differs from FIG. 5C in that FIG. 5B does not directly show merchant-to-dish correspondence, but only shows dishes; fig. 5C shows the correspondence between the merchants and the dishes. That is, the information providing method specified in the present specification includes not only the information to be provided but also the providing methods of the respective information: the information to be provided is obviously different in information providing mode, and the information provided is the same, and the information is provided in different mode, and is also different in information providing mode.

To this end, the present specification provides an information providing method that is based on modification of an information providing model that can predict user preferences and provide target information and related information related to the target information according to the predicted preferences and achieve a higher accuracy.

Based on the same idea, the information providing method provided by the embodiment of the present specification further provides a corresponding apparatus, a storage medium, and an electronic device.

Fig. 6 is a schematic structural diagram of an information providing apparatus provided in an embodiment of the present specification, where the apparatus includes:

an obtaining module 600, configured to determine each target information to be provided and each behavior data of a user;

a feature input module 602, configured to extract information features from each piece of target information; extracting behavior characteristics from each behavior data; inputting all information characteristics and all behavior characteristics into a pre-trained information providing model, wherein the information providing model is used for determining an information providing mode of all target information;

a feature weighting module 604, configured to determine, for each input information feature, an association weight between the information feature and each behavior feature through the information provision model, perform association weighting on the corresponding behavior feature according to the association weight, and determine each behavior feature after the association weighting;

a target recommendation module 606, configured to determine, according to each behavior feature after the associated weighting, a target recommendation degree of the target information output by the information provision model;

the information providing module 608 is configured to determine an information providing manner of each target information according to the target recommendation degree of each target information, and provide at least part of the target information to the user in the information providing manner.

The device further comprises: a model training module 610, configured to determine an information providing model to be trained, and a training sample for training the information providing model, where the training sample includes: target information and historical behavior data of users; extracting information features from each target information; behavior characteristics are extracted from each behavior data and input into the information providing model to be trained; determining the association weight of each information feature to each behavior feature, performing association weighting on the corresponding behavior feature according to the association weight, determining each behavior feature after association weighting, and determining the target recommendation degree of target information corresponding to the information feature according to each behavior feature after association weighting; determining the labeling information of each target information according to the behavior data; and determining the loss of the information providing model according to the target recommendation degree and the labeling information of each target information, and training the information providing model by taking the loss minimization as a target.

Optionally, the feature input module 602 is specifically configured to use an information feature extracted from each target information as a first information feature; determining each relevant information corresponding to the target information; extracting information characteristics from each piece of relevant information corresponding to the target information to serve as second information characteristics; and inputting each first information characteristic, each second information characteristic and each behavior characteristic into a pre-trained information providing model.

Optionally, the feature weighting module 604 is specifically configured to, for an input first information feature, determine a first association weight between the first information feature and each behavior feature through the information provision model, perform association weighting on the corresponding behavior feature by using the first association weight, and determine each behavior feature after the association weighting; and aiming at the input second information characteristic, determining a second association weight of the second information characteristic and each behavior characteristic through the information providing model, and performing association weighting on the corresponding behavior characteristic by using the second association weight to determine each behavior characteristic after the association weighting.

Optionally, the feature weighting module 604 includes at least two sub-networks, and is specifically configured to input each first information feature and each first behavior feature into the first sub-network; inputting each second information characteristic and each second behavior characteristic into a second subnetwork; wherein the first behavior feature is a behavior feature extracted from behavior data in which the target information is an object, and the second behavior feature is a behavior feature extracted from behavior data in which the related information is an object;

the feature weighting module 604 is specifically configured to, for an input first information feature, determine a first association weight between the first information feature and each first behavior feature through the information provision model, perform association weighting on the corresponding first behavior feature by using the first association weight, and determine each first behavior feature after the association weighting; and aiming at the input second information characteristics, determining second association weights of the second information characteristics and the second behavior characteristics through the information providing model, performing association weighting on the corresponding second behavior characteristics through the second association weights, and determining the second behavior characteristics after the association weighting.

Optionally, the target recommendation module 606 is specifically configured to determine, according to the second behavior features after the association weighting, a first comprehensive feature of the target information; determining a second comprehensive characteristic of the target information according to each first behavior characteristic and each first comprehensive characteristic of the target information; and determining the target recommendation degree of the target information according to the second comprehensive characteristics of the target information.

Optionally, the information providing module 608 is specifically configured to provide at least part of the target information and at least part of the related information corresponding to each provided target information to the user in the information providing manner.

Optionally, the target information specifically includes: information of the merchant; the relevant information corresponding to the target information specifically includes: information about the goods offered by the merchant.

The present specification also provides a computer-readable storage medium storing a computer program operable to execute the information providing method provided in fig. 4 described above.

This specification also provides a schematic block diagram of the electronic device shown in fig. 7. As shown in fig. 7, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, but may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs the computer program to implement the information providing method provided in fig. 4. Of course, besides the software implementation, the present specification does not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

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

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims

1. An information providing method, comprising:

2. The method according to claim 1, characterized in that an information feature extracted from each target information is taken as a first information feature;

determining each relevant information corresponding to the target information;

3. The method of claim 2, wherein determining the associated weight of the information feature and each behavior feature through the information providing model specifically comprises:

4. The method of claim 2, wherein the information providing model comprises at least two sub-networks;

5. The method as claimed in claim 4, wherein the determining, by the information providing model, the associated weight of the information feature and each behavior feature, and performing associated weighting on the corresponding behavior feature by using the associated weight to determine each behavior feature after the associated weighting specifically includes:

6. The method of claim 4, wherein prior to determining the target recommendation for the target information output by the information provision model, the method further comprises:

7. The method of claim 2, wherein providing at least part of the target information to the user in the information providing manner specifically comprises:

8. The method of claim 7, wherein the target information specifically includes: information of the merchant;

9. The method of any one of claims 1 to 8, wherein pre-training the information providing model specifically comprises:

10. An information providing apparatus, comprising:

11. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1 to 9.

12. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of claims 1 to 9 when executing the program.