WO2022148038A1 - Information recommendation method and device - Google Patents

Abstract

An information recommendation method and device, belonging to the technical field of the Internet. The information recommendation method comprises: in response to an information acquisition request of a target object, acquiring a long-term behavior data set of the target object (S201); determining feature information of each recommendation information in a recommendation information set (S203); determining target behavior data corresponding to each recommendation information from the long-term behavior data set on the basis of the feature information of each recommendation information (S205); inputting the target behavior data into an interest recognition network for interest recognition to obtain an interest index of the target object for each recommendation information (S207); and recommending target information in the recommendation information set to the target object on the basis of the interest index (S209).

Description

Information recommendation method and device

CROSS-REFERENCE TO RELATED APPLICATIONS

The present disclosure is based on a Chinese patent application with an application date of January 11, 2021 and an application number of 202110034339.3, and claims the priority of the Chinese patent application. The disclosure of the above Chinese patent application is hereby incorporated by reference in its entirety as a part of this disclosure.

technical field

The present disclosure relates to the field of Internet technologies, and in particular, to an information recommendation method, apparatus, electronic device, and storage medium.

Background technique

With the development of Internet technology, a large number of network platforms are constantly being upgraded. In addition to publishing some graphic information, individual users can also share daily short videos, etc., and how to accurately capture the interests of users is a large number of recommendations. system challenges.

SUMMARY OF THE INVENTION

The present disclosure provides an information recommendation method, device, electronic device and storage medium. The technical solutions of the present disclosure are as follows:

According to a first aspect of the embodiments of the present disclosure, an information recommendation method is provided, including:

In response to the information acquisition request of the target object, acquiring a long-term behavior data set of the target object, where the long-term behavior data set represents a plurality of historical behavior data of the target object within a preset time period;

Determine the feature information of each recommended information in the recommended information set;

Determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

Inputting the target behavior data into an interest recognition network for interest recognition, to obtain the target object's interest index for each recommendation information;

The target information in the recommendation information set is recommended to the target object based on the interest index.

In some embodiments, determining the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information includes:

Obtaining behavior feature information of each historical behavior data in the long-term behavior data set;

calculating the first similarity between the feature information of each recommendation information and the behavior feature information;

According to the first similarity, target behavior data corresponding to each recommendation information is determined from the long-term behavior data set.

In some embodiments, in response to the information acquisition request of the target object, acquiring the long-term behavior data set of the target object includes:

Receive the information acquisition request of the target object, and the information acquisition request includes the target object identification of the target object;

A long-term behavior data set of the target object is acquired based on the target object identifier.

In some embodiments, acquiring the long-term behavior data set of the target object based on the object identifier includes:

Obtaining a pre-built historical behavior database, the historical behavior database includes historical behavior data of multiple objects, and the mapping relationship between the object identifiers of the multiple objects and the corresponding historical behavior data;

Obtain historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

The corresponding historical behavior data is used as the long-term behavior data set of the target object.

In some embodiments, the determining the characteristic information of each recommendation information in the recommendation information set includes:

obtaining the subject tag of each of the recommended information;

generating feature information of each recommendation information based on the subject tag of each recommendation information;

or,

extracting the text information of each recommended information;

The feature information of each recommendation information is generated based on the text information of each recommendation information.

In some embodiments, the recommending the target information in the recommendation information set to the target object based on the interest index includes:

determining target information from the recommendation information set according to the interest index;

The target information is recommended to the target object.

In some embodiments, the method further includes:

Obtain information identification and sample feature information of multiple sample recommendation information;

Determine the historical recommendation object corresponding to each sample recommendation information and the interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

obtaining a long-term sample behavior data set of the historical recommendation object;

Determine the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of the each sample recommendation information;

Inputting the sample behavior data into the neural network for interest identification, and obtaining the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

determining a target loss according to the interest prediction index and the interest labeling index;

The neural network is trained based on the target loss to obtain the interest recognition network.

In some embodiments, the method further includes:

determining the recommendation time of the recommendation information for each sample;

Perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtered behavior data set;

The sample behavior data corresponding to each sample recommendation information determined from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information includes:

The sample behavior data corresponding to each sample recommendation information is determined from the corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.

According to a second aspect of the embodiments of the present disclosure, there is provided an information recommendation apparatus, including:

The first behavior data set acquisition module is configured to obtain a long-term behavior data set of the target object in response to an information acquisition request of the target object, and the long-term behavior data set represents the number of the target object in a preset time period. historical behavioral data;

a feature information determination module, configured to determine feature information of each recommendation information in the recommendation information set;

a target behavior data determination module, configured to determine target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

a first interest identification module, configured to input the target behavior data into an interest identification network for interest identification, and obtain an interest index of the target object for each recommendation information;

An information recommendation module, configured to recommend target information in the recommended information set to the target object based on the interest index.

In some embodiments, the target behavior data determination module includes:

a behavior feature information acquisition unit, configured to acquire behavior feature information of each historical behavior data in the long-term behavior data set;

a first similarity calculation unit, configured to calculate a first similarity between the feature information of each recommendation information and the behavior feature information;

The target behavior data determining unit is configured to determine the target behavior data corresponding to each recommendation information from the long-term behavior data set according to the first similarity.

In some embodiments, the first behavior data set acquisition module includes:

an information acquisition request receiving unit, configured to receive an information acquisition request of the target object, where the information acquisition request includes a target object identifier of the target object;

The long-term behavior data set acquisition unit is configured to acquire the long-term behavior data set of the target object based on the target object identifier.

In some embodiments, the long-term behavior data set acquisition unit includes:

A historical behavior database acquisition unit, configured to acquire a pre-built historical behavior database, the historical behavior database includes historical behavior data of a plurality of objects, and a mapping relationship between the object identifiers of the plurality of objects and the corresponding historical behavior data ;

a historical behavior data acquisition unit, configured to acquire historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

The long-term behavior data set determination unit is configured to use the corresponding historical behavior data as the long-term behavior data set of the target object.

In some embodiments, the feature information determination module includes:

a first topic tag acquiring unit, configured to acquire the topic tag of each of the recommended information;

a first feature information generating unit, configured to generate feature information of each recommendation information based on the subject tag of each recommendation information;

or,

a first text information extraction unit, configured to extract the text information of each recommended information;

The second feature information generating unit is configured to generate feature information of each recommendation information based on the text information of each recommendation information.

In some embodiments, the information recommendation module includes:

a target information determination unit, configured to determine target information from the recommended information set according to the interest index;

An information pushing unit, configured to recommend the target information to the target object.

In some embodiments, the apparatus further includes:

an information acquisition module, configured to acquire information identifiers and sample feature information of multiple sample recommendation information;

an object information determination module, configured to determine a historical recommendation object corresponding to each sample recommendation information and an interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

The second behavior data set obtaining module is configured to obtain the long-term sample behavior data set of the historical recommendation object;

a sample behavior number determination module, configured to determine sample behavior data corresponding to each sample recommendation information from a corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information;

The second interest identification module is configured to input the sample behavior data into the neural network for interest identification, so as to obtain the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

a target loss determination module, configured to determine a target loss according to the interest prediction index and the interest labeling index;

A network training module configured to train the neural network based on the target loss to obtain the interest recognition network.

In some embodiments, the apparatus further includes:

a recommendation time determination module, configured to determine the recommendation time of the recommended information for each sample;

a data filtering module, configured to perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtering behavior data set;

The sample behavior data determination module is further configured to determine sample behavior data corresponding to each sample recommendation information from a corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.

According to a third aspect of embodiments of the present disclosure, there is provided an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to achieve The method of any one of the first aspects above.

According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, when instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the first embodiment of the present disclosure. The method of any of the aspects.

According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform any of the methods described in the first aspect of the embodiments of the present disclosure.

In the process of information recommendation, the long-term behavior data set of the target object is obtained, which ensures the comprehensiveness of the behavior data and can effectively reflect the interest and preference of the object. The target behavior data used for interest recognition can effectively reflect the target object's real interest preference for recommended information, and can greatly reduce the amount of data in the process of interest recognition, effectively improve the efficiency and accuracy of interest recognition, and greatly improve the recommendation. Information recommendation accuracy and recommendation performance in the system.

Description of drawings

1 is a schematic diagram of an application environment according to some embodiments;

FIG. 2 is a flowchart of an information recommendation method according to some embodiments;

3 is a schematic flowchart of determining target behavior data corresponding to each recommendation information from a long-term behavior data set based on feature information of each recommendation information according to some embodiments;

4 is a schematic flowchart of a pre-trained interest recognition network according to some embodiments;

5 is a schematic flowchart of a pre-trained interest recognition network according to some embodiments;

6 is a block diagram of an information recommendation apparatus according to some embodiments;

Fig. 7 is a block diagram of an electronic device for information recommendation according to some embodiments.

Detailed ways

In order to make those skilled in the art better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first", "second" and the like in the description and claims of the present disclosure and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances such that the embodiments of the disclosure described herein can be practiced in sequences other than those illustrated or described herein. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure, as recited in the appended claims.

Please refer to FIG. 1, which is a schematic diagram of an application environment according to an exemplary embodiment. As shown in FIG. 1, the application environment may include a log database 01, a retrieval server 02, a historical behavior database 03, and a training server 04 , terminal 05 , content distribution server 06 and service server 07 .

In some embodiments, the log database 01 may be used to store recommendation log information generated during the recommendation process of the recommendation system. The historical behavior database 03 can be used to store long-term historical behavior data of a large number of objects in the recommender system. In some embodiments, the historical behavior database 03 uses the object identifier as the key (retrieval parameter), and the historical behavior data set corresponding to the object as the value (return data corresponding to the retrieval parameter) for data storage.

In some embodiments, the retrieval server 02 may initiate a long-term behavior data set acquisition request in combination with the recommendation log information in the log database 01, and acquire long-term samples of historical recommendation objects corresponding to the sample recommendation information recommended by the recommendation system from the historical behavior database 03 Behavior data set; and send the long-term sample behavior data set and sample recommendation information to the training server 04 as training data.

In some embodiments, the training server 04 may train the interest recognition network in combination with the long-term sample behavior data set and the sample recommendation information sent by the retrieval server 02 , and send the trained interest recognition network to the service server 07 .

In some embodiments, the terminal 05 may provide a user-oriented information recommendation service; accordingly, the user may trigger an information acquisition request based on the terminal 05 .

In some embodiments, the information acquisition request triggered by the terminal 05 may be sent to the content distribution server 06; correspondingly, the content distribution server 06 forwards the information acquisition request to the service server 07, and then the service server 07 combines the information acquisition request with the information contained in the request. The target object identifier of the target object obtains the target behavior data of the target object corresponding to the recommendation information from the historical behavior database 03, and combines the interest recognition network to perform interest recognition, and then information recommendation can be performed based on the result of the interest recognition.

In some embodiments, the retrieval server 02, the training server 04, the content distribution server 06, and the service server 07 may be independent physical servers, or may be server clusters or distributed systems composed of multiple physical servers, or may provide cloud services. Services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms Cloud servers for cloud computing services.

In some embodiments, the terminal 05 may include, but is not limited to, smartphones, desktop computers, tablet computers, laptop computers, smart speakers, digital assistants, augmented reality (AR)/virtual reality (VR) devices, Electronic devices such as smart wearable devices. In some embodiments, the operating system running on the electronic device may include, but is not limited to, the Android system, the IOS system, linux, windows, and the like.

In some embodiments, the log database 01 and the historical behavior database 03 may be relational databases.

In addition, it should be noted that what is shown in FIG. 1 is only an application environment provided by the present disclosure. In practical applications, other application environments may also be included. For example, the information acquisition request of the terminal 05 can be directly sent to the service server 07, In the embodiment of this specification, the content distribution server 06 set between the terminal 05 and the service server 07 can effectively reduce the request pressure of the service server, thereby improving the efficiency of information recommendation.

In the embodiment of this specification, the above-mentioned log database 01, retrieval server 02, historical behavior database 03, training server 04, terminal 05, content distribution server 06 and service server 07 can be directly or indirectly connected through wired or wireless communication. The disclosure is not limited here.

FIG. 2 is a flowchart of an information recommendation method according to an exemplary embodiment. As shown in FIG. 2 , the information recommendation method is used in electronic devices such as servers and edge computing nodes, and includes the following steps:

In S201, in response to the information acquisition request of the target object, a long-term behavior data set of the target object is acquired.

In some embodiments, the long-term behavioral data set represents a plurality of historical behavioral data of the target object within a preset time period. In some embodiments, the preset time period may be set in combination with actual application requirements, such as 1 year. In the embodiment of this specification, each historical behavior data may reflect the associated data in the process of the behavior of the corresponding object to the recommended information. In some embodiments, each historical behavior data may include: information identifier of historical recommendation information, publisher identifier of historical recommendation information, recommendation time, information viewing time, topic tag of historical recommendation information, behavior of objects on historical recommendation information Tags (e.g. click, like, follow, retweet, etc.).

In some embodiments, acquiring the long-term behavior data set of the target object in response to the information acquisition request of the target object includes: receiving an information acquisition request of the target object, where the information acquisition request includes the target object identifier of the target object; acquiring based on the target object identifier A long-term behavioral dataset of the target object.

In some embodiments, the above method may further include:

Pre-built historical behavior database.

In some embodiments, the historical behavior database may include historical behavior data of multiple objects, and a mapping relationship between object identifiers of the multiple objects and corresponding historical behavior data. Generally, each object often corresponds to a large amount of historical behavior data.

In practical applications, in order to store enough historical behavior data as soon as possible, the historical behavior database can support backfilling in two formats: historical data backfilling and automatic backfilling triggered by recommended events. In the process of historical data backfilling, historical behavior data can be extracted from a large amount of historical log data and backfilled into the historical behavior database. The automatic backfill triggered by the recommendation event can update the behavior data in the real-time information recommendation process to the historical behavior database. In addition, in order to ensure the consistency of the data, when updating, it is possible to perform deduplication processing in combination with the recommendation time and the information identifier of the historical recommendation information.

Correspondingly, the above-mentioned acquisition of the long-term behavior data set of the target object based on the object identifier may include:

Obtain a pre-built historical behavior database, the historical behavior database includes historical behavior data of multiple objects, and the mapping relationship between object identifiers of multiple objects and the corresponding historical behavior data;

Obtain the historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

The corresponding historical behavior data is taken as the long-term behavior data set of the target object.

In the above embodiment, by pre-constructing the historical behavior database including the historical behavior data of a large number of objects and the mapping relationship between the object identifiers of the large number of objects and the corresponding historical behavior data, it is convenient to receive the information acquisition request of the target object. Quickly query long-term behavioral data sets of objects directly combined with object identification.

In S203, the feature information of each recommendation information in the recommendation information set is determined.

In the embodiments of this specification, the recommended information set may include a large number of recommended information in the recommendation system, and in some embodiments, the recommended information may be videos, pictures, text information, and the like.

In some embodiments, determining the feature information of each recommendation information in the recommendation information set includes:

1) Get the hashtag of each recommendation;

2) generating feature information of each recommendation information based on the subject tag of each recommendation information;

In practical applications, the subject tag of the recommended information may be text information that can represent the subject content of the recommended information. In some embodiments, when a user publishes information in some platforms, the user is often asked to fill in the subject tag of the information in combination with the preset symbols. The hashtags are distinguished by means of preset symbols, and correspondingly, the hashtags can be obtained from the relevant text information of the information by matching the preset symbols. In some embodiments, the preset symbol may be "##", of course, other symbols or manners may also be used to distinguish hashtags in practical applications.

In some embodiments, a pre-trained hashtag recognition network can also be used to extract the hashtags of the recommended information. In some embodiments, the hashtag is obtained by the network performing hashtag recognition training on a preset neural network based on a large amount of sample recommendation information and hashtags corresponding to the sample recommendation information.

Further, generating the feature information of the recommendation information based on the topic tag of each recommendation information may include acquiring the word vector of the topic tag, and correspondingly, the word vector of the topic tag may be used as the feature information of the recommendation information.

In some embodiments, a word vector representation model obtained by pre-training a preset word vector model based on preset training text information may be pre-trained. In some embodiments, the preset training text information may be text information in a recommendation system.

In some embodiments, during the training of the word vector representation model, the preset training text information may be subjected to word segmentation processing, and the classification information after word segmentation processing may be input into the preset word vector model for training. Each word is mapped into a K-dimensional real number vector, and a word vector set representing the semantic similarity between words can be obtained while the word vector representation model is obtained. The preset word vector model is trained with the preset training text information in a certain system, and the obtained word vector representation model can effectively represent the semantic similarity between words in the system.

In some embodiments, when the word vector representation model has been trained, the hashtag can be segmented and input into the word vector representation model, and the word vector representation model can determine the above-mentioned hashtag based on the word vector in the word vector set The word vector of the corresponding word segmentation information. Further, in the case that the word segmentation information corresponding to the topic tag has multiple words, the mean value of the word vectors of the multiple words can be taken as the word vector of the topic tag; in the case that the word segmentation information corresponding to the topic tag has one word , the word vector of this word can be used as the word vector of the hashtag.

In the embodiment of this specification, the preset word vector model may include, but is not limited to, word2vec, BERT, glove and other models.

In some embodiments, determining the feature information of each recommendation information in the recommendation information set above may include:

1) Extract the text information of each recommendation information;

2) Generate feature information of each recommendation information based on the text information of each recommendation information.

In some embodiments, when the recommendation information itself is text information, the recommendation information may be used as text information. In some embodiments, when the recommended information is a video, the text information of the recommended information may include text information corresponding to the voice information in the video, text information of the video cover, title information of the video, and entity text (entity text) extracted from the video. It can be a person, an object, etc.) and the search information of the video (that is, the video can be recalled based on the search information in the recommender system), etc.

In some embodiments, generating the feature information of each recommendation information based on the text information of each recommendation information may be a word vector corresponding to the text information for generating the recommendation information, and correspondingly, the corresponding word vector recommends the feature information of the information. For the specific refinement of the word vector corresponding to the text information of the generated recommendation information, reference may be made to the above-mentioned relevant steps, which will not be repeated here.

In some embodiments, feature representation networks such as one-hot (one-hot) coding network and N-Gram (Chinese language model) can also be combined to generate a feature vector corresponding to the text information of the recommendation information. Correspondingly, the feature vector can be generated. The vector is used as the feature information of recommendation information.

In the above-mentioned embodiment, the feature information of the recommendation information is generated by the subject tag or text information of the recommendation information, which can realize the effective representation of the recommendation information, thereby ensuring the subsequent recall of behavior data related to the recommendation information.

In S205, target behavior data corresponding to each recommendation information is determined from the long-term behavior data set based on the feature information of each recommendation information.

In some embodiments, as shown in FIG. 3 , determining the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information may include the following steps:

In S2051, obtain behavior feature information of each historical behavior data in the long-term behavior data set;

In S2053, calculate the first similarity between the feature information of each recommendation information and the behavior feature information;

In S2055, according to the first similarity, target behavior data corresponding to each recommendation information is determined from the long-term behavior data set.

In some embodiments, the behavior feature information of the historical behavior data is a feature vector corresponding to the historical behavior data generated by a feature representation network such as a one-hot coding network, an N-Gram (Chinese language model), and the like.

In some embodiments, the first similarity between the behavior feature information and the feature information of the recommendation information may represent the degree of association between the historical behavior data and the recommendation information. In some embodiments, the relationship between the behavior feature information and the feature information of the recommendation information The higher the first similarity, the higher the degree of correlation between the corresponding historical behavior data and the recommendation information; conversely, the lower the first similarity between the behavior feature information and the feature information of the recommendation information, the lower the corresponding historical behavior data and the recommendation information. The lower the degree of correlation between the information.

In some embodiments, the first similarity between the behavior feature information and the feature information of the recommendation information may include, but is not limited to, cosine distance, Euclidean distance, and Manhattan distance between the behavior feature information and the feature information of the recommendation information.

In some embodiments, historical behavior data corresponding to behavior feature information whose first degree of similarity to the feature information of the recommendation information is greater than or equal to a preset threshold may be used as the target behavior data corresponding to the recommendation information.

In some embodiments, the first similarity between the behavior feature information of each historical behavior data in the long-term behavior data set and the feature information of the recommendation information can be sorted in descending order, and correspondingly, a preset number of behaviors can be sorted first. The historical behavior data corresponding to the feature information is used as the target behavior data corresponding to the recommendation information.

In the embodiment of this specification, the preset threshold and the preset number can be set in advance in combination with practical applications to set the information recommendation accuracy requirements (the higher the degree of association between the recommended information and the historical behavior data, the better the information recommendation accuracy).

In the above-mentioned embodiment, through the similarity between the feature information of the recommendation information and the behavior feature information of the historical behavior data, the historical behavior data with a higher degree of correlation can be selected from the long-term behavior data set of the target object for each recommendation information. As the target behavior data corresponding to the recommendation information, it ensures that the behavior data used for interest identification can fully reflect the interests and preferences of the object; at the same time, combined with the feature information of different recommendation information, the historical behavior data is screened in a targeted manner, so that the target behavior The data can effectively reflect the interest and preference of the target object to the recommended information, greatly reduce the amount of data in the process of interest identification, and effectively improve the efficiency and accuracy of interest identification.

In S207, the target behavior data is input into the interest recognition network for interest recognition, and the target object's interest index for each recommendation information is obtained.

In some embodiments, the interest index can represent the preference of the target object to the recommended information. In some embodiments, the interest recognition network may be pre-trained, and accordingly, the above method may further include the step of pre-training the interest recognition network. As shown in Figure 4, pre-training an interest recognition network can include the following steps:

In S401, information identifiers and sample feature information of multiple sample recommendation information are acquired.

In the embodiment of this specification, the multiple sample recommendation information may be information recommended in the recommendation system. In some embodiments, the recommendation log information stored in the log database may include sample recommendation information corresponding to multiple historical recommendation events, information identifiers of the sample recommendation information corresponding to multiple historical recommendation events, and recommendation time and other basic information about historical recommendation events. information. Correspondingly, in the embodiment of this specification, information identifiers of multiple sample recommendation information may be obtained from recommendation log information.

In some embodiments, the sample feature information of the above-mentioned multiple sample recommendation information may be obtained in the following manner:

Get the hashtag of each sample recommendation information;

Generate sample feature information of each sample recommendation information based on the subject tag of each sample recommendation information;

or,

Extract the text information of each sample recommendation information;

The sample feature information of each sample recommendation information is generated based on the text information of each sample recommendation information.

In the embodiment of this specification, for the specific refinement of the steps related to obtaining the sample feature information of the sample recommendation information, reference may be made to the specific refinement of the related steps for determining the feature information of each recommendation information in the recommendation information set above, which will not be repeated here.

In the above embodiment, the sample feature information of the sample recommendation information is generated by the subject tag or text information of the sample recommendation information, which can realize the effective representation of the sample recommendation information, thereby ensuring the subsequent recall of behavior data related to the sample recommendation information.

In S403, a historical recommendation object corresponding to each sample recommendation information and an interest labeling index corresponding to the historical recommendation object are determined based on the information identifier of each sample recommendation information.

In practical applications, a large amount of historical behavior data stored in the historical behavior database can reflect the behavior of the corresponding object after the recommendation information is recommended. Correspondingly, each historical behavior data can include the information identifier of the recommendation information (sample recommendation information) and The object's behavior label for this recommendation.

In the embodiment of this specification, the object corresponding to the historical behavior data pair including the information identifier of the sample recommendation information may be used as the historical recommendation object corresponding to the sample recommendation information. Correspondingly, the interest labeling index can be generated in combination with the behavior label of the object's recommendation information about the sample in the historical behavior data.

In practical applications, in different business scenarios, behaviors that characterize an object's preference for a recommendation information may be different, and in the same business scenario, there may also be multiple behaviors that characterize an object's preference for a recommendation information. Correspondingly, in the embodiment of the present specification, the interest labeling index may include, in combination with the actual application scenario, the preference of the object on the recommendation information reflected by one or more behaviors. In some embodiments, taking the click behavior to reflect the object's preference for the recommended information as an example, in response to the behavior label in the historical behavior data being a click, in some embodiments, the interest labeling indicator may be 1 (represented by 1). clicked, 0 means not clicked).

In S405, a long-term sample behavior data set of the historical recommendation object is acquired.

In some embodiments, in the case of determining the historical recommendation object corresponding to the sample recommendation information, the object identifier of the historical recommendation object may be obtained. In some embodiments, corresponding historical behavior data may be obtained from a historical behavior database in combination with the object identifier, as a long-term sample behavior data set of the historical recommendation object.

In S407, the sample behavior data corresponding to each sample recommendation information is determined from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information.

In some embodiments, determining the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information may include:

Obtain the sample behavior feature information of each sample behavior data in the long-term sample behavior data set;

calculating the second similarity between the sample behavior feature information and the sample feature information of each sample recommendation information;

According to the second similarity, the sample behavior data corresponding to each sample recommendation information is determined from the long-term sample behavior data set.

In the embodiment of this specification, for the specific refinement of the relevant steps for determining the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information, please refer to the above-mentioned based on each sample behavior data. The feature information of the recommendation information is determined from the long-term behavior data set to determine the specific refinement of the relevant steps of the target behavior data corresponding to each recommendation information, which will not be repeated here.

In the above-mentioned embodiment, through the similarity between the sample feature information of the sample recommendation information and the behavior feature information of the historical behavior data corresponding to the historical recommendation object, each sample recommendation information can be selected from the long-term sample behavior data set of the historical recommendation object. The historical behavior data with a high degree of correlation is used as the sample behavior data corresponding to the sample recommendation information, which ensures that the sample behavior data used for interest identification can fully reflect the interests and preferences of the object; at the same time, combined with different sample recommendation information, targeted The screening of historical behavior data enables the sample behavior data to effectively reflect the interest and preference of historical recommendation objects to sample recommendation information, and greatly reduces the amount of data in the process of interest recognition, thereby improving the efficiency and accuracy of interest recognition.

In S409, the sample behavior data is input into the neural network for interest identification, so as to obtain the interest prediction index of each historical recommendation object to the corresponding sample recommendation information.

In some embodiments, the neural network may identify the network for the interest to be trained. In some embodiments, neural networks may include, but are not limited to, convolutional neural networks, recurrent neural networks, and the like. In some embodiments, the interest prediction index may represent the preference of each historical recommendation object predicted by the trained neural network to the corresponding sample recommendation information. In some embodiments, the interest predictor may be a value greater than or equal to 0 and less than or equal to 1. Correspondingly, the larger the value of the interest predictor is, the greater the value of the interest predictor, represents the difference between the historical recommendation objects predicted by the trained neural network and the corresponding sample recommendation information. more like.

In S411, the target loss is determined according to the interest prediction index and the interest labeling index.

In some embodiments, according to the interest prediction index and the interest labeling index, determining the target loss may include calculating the loss between the interest prediction index corresponding to each historical recommendation object and the corresponding interest labeling index based on a preset loss function, and analyzing multiple The losses corresponding to the historical recommendation objects corresponding to the sample recommendation information are summed to obtain the above target loss.

In the embodiment of this specification, the preset loss function may include, but is not limited to, a cross-entropy loss function, a logistic loss function, a Hinge (hinge) loss function, an exponential loss function, and the like, and the embodiment of this specification is not limited to the above.

In S413, a neural network is trained based on the target loss to obtain an interest recognition network.

In some embodiments, training a neural network based on the target loss to obtain an interest recognition network may include

When the target loss does not meet the preset conditions, update the network parameters in the neural network;

The target loss is updated based on the updated neural network, until the target loss meets the preset condition, and the current neural network is used as the above-mentioned interest recognition network.

In some embodiments, the target loss satisfying the preset condition may be that the target loss is less than or equal to a specified threshold, or the difference between the corresponding target losses in the two training processes before and after is less than a certain threshold. In the embodiment of this specification, the specified threshold and a certain threshold may be set in combination with actual training requirements.

In addition, it should be noted that, in practical applications, in response to the neural network being a multi-task neural network to be trained, correspondingly, the interest annotation information may be task annotation information corresponding to multiple tasks, and the interest prediction indicators may be multiple Task prediction information corresponding to the task.

In the above embodiment, in the training process of the interest recognition network, the identification information of the sample recommendation information is obtained to determine the historical recommendation object and the interest labeling index corresponding to the historical recommendation object, and the long-term sample behavior data set of the historical recommendation object is obtained, which ensures the The comprehensiveness of the behavior data in the training process can effectively reflect the interests and preferences of the objects, and at the same time, combined with different sample recommendation information, the historical behavior data is screened in a targeted manner, so that the sample behavior data used for interest identification can effectively reflect the historical recommendation objects. The real interest preference of sample recommendation information can greatly reduce the amount of data in the interest recognition process, thereby effectively improving the interest recognition efficiency and recognition accuracy of the trained interest recognition network.

In some embodiments, before determining the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information, as shown in FIG. 5 , the above method further includes:

In S415, the recommendation time of each sample recommendation information is determined;

In S417, data filtering is performed on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtering behavior data set;

Correspondingly, determining the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information may include:

The sample behavior data corresponding to each sample recommendation information is determined from the corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.

In some embodiments, the recommendation time of the sample recommendation information may be obtained from the corresponding recommendation log information.

In some embodiments, performing data filtering on the corresponding long-term sample behavior data set based on the recommendation time, and obtaining the filtered behavior data set may include filtering out the behavior time from the long-term sample behavior data set of the historical recommendation object corresponding to each sample recommendation information After the sample behavior data of the recommendation information of each sample, the above filtering behavior data set is obtained.

In the above embodiment, data filtering is performed on the corresponding long-term sample behavior data set in combination with the recommendation time of the sample recommendation information, which can effectively prevent data traversal, realize the filtering of historical behavior data generated after the recommendation time, and ensure that the information is used for interest purposes. Validity of identified behavioral data.

In S209, the target information in the recommendation information set is recommended to the target object based on the interest index.

In some embodiments, recommending the target information in the recommendation information set to the target object based on the interest index may include: determining the target information from the recommendation information set according to the interest index; recommending the target information to the target object.

In some embodiments, the interest index represents the preference of the target object for each recommendation information in the recommendation information set. In some embodiments, the interest index may be a numerical value proportional to the degree of preference, or may be a characterized representation representing the degree of preference of the target object for each recommendation information in the recommendation information set, such as "medium". In some embodiments, the character symbol representation can be quantified into corresponding numerical values in combination with certain rules.

In some embodiments, a confidence threshold may be set in advance in combination with the information recommendation accuracy requirement (generally, the higher the confidence threshold, the more accurate the recommended information). Correspondingly, the value corresponding to the interest index is greater than or equal to the confidence threshold. Recommendation information can be used as target information.

It can be seen from the technical solutions provided by the above embodiments of this specification that in the embodiments of this specification, in the process of information recommendation, the long-term behavior data set of the target object is obtained, which ensures the comprehensiveness of the behavior data, and can effectively reflect the interests and preferences of the object. Combined with the feature information of different recommendation information, the historical behavior data is screened in a targeted manner, so that the target behavior data used for interest identification can effectively reflect the target object's real interest preference for the recommendation information, and can greatly reduce the interest identification process. It can effectively improve the efficiency and accuracy of interest recognition, and then greatly improve the accuracy and performance of information recommendation in the recommendation system.

Fig. 6 is a block diagram of an information recommendation apparatus according to an exemplary embodiment. Referring to Figure 6, the device includes:

The first behavior data set obtaining module 610 is configured to obtain the long-term behavior data set of the target object in response to the information obtaining request of the target object;

A feature information determination module 620, configured to determine feature information of each recommendation information in the recommendation information set;

The target behavior data determination module 630 is configured to determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

The first interest identification module 640 is configured to input the target behavior data into the interest identification network for interest identification, and obtain the interest index of the target object for each recommendation information;

The information recommendation module 650 is configured to recommend the target information in the recommended information set to the target object based on the interest index.

In some embodiments target behavior data determination module 630 includes:

a behavior feature information acquisition unit, configured to acquire behavior feature information of each historical behavior data in the long-term behavior data set;

a first similarity calculation unit, configured to calculate a first similarity between the feature information of each recommendation information and the behavior feature information;

The target behavior data determining unit is configured to determine the target behavior data corresponding to each recommendation information from the long-term behavior data set according to the first similarity.

In some embodiments, the first behavior data set acquisition module 610 includes:

an information acquisition request receiving unit, configured to receive an information acquisition request of a target object, where the information acquisition request includes a target object identifier of the target object;

The long-term behavior data set acquisition unit is configured to acquire the long-term behavior data set of the target object based on the target object identifier.

In some embodiments, the long-term behavioral data set acquisition unit includes:

The historical behavior database acquisition unit is configured to acquire a pre-built historical behavior database, the historical behavior database includes historical behavior data of multiple objects, and a mapping relationship between object identifiers of multiple objects and corresponding historical behavior data;

The historical behavior data acquisition unit is configured to acquire historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

The long-term behavior data set determining unit is configured to take the corresponding historical behavior data as the long-term behavior data set of the target object.

In some embodiments, characteristic information determination module 620 includes:

a first topic tag acquiring unit, configured to acquire a topic tag of each recommended information;

a first feature information generating unit, configured to generate feature information of each recommendation information based on the subject tag of each recommendation information;

or,

a first text information extraction unit, configured to extract the text information of each recommendation information;

The second feature information generating unit is configured to generate feature information of each recommendation information based on the text information of each recommendation information.

In some embodiments, the information recommendation module 650 includes:

a target information determination unit, configured to determine target information from the recommended information set according to the interest index;

The information pushing unit is configured to recommend target information to the target object.

In some embodiments, the above-mentioned apparatus further comprises:

an information acquisition module, configured to acquire information identifiers and sample feature information of multiple sample recommendation information;

The object information determination module is configured to determine the historical recommendation object corresponding to each sample recommendation information and the interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

The second behavior data set acquisition module is configured to acquire long-term sample behavior data sets of historical recommendation objects;

The sample behavior number determination module is configured to determine the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information;

The second interest identification module is configured to input the sample behavior data into the neural network for interest identification, and obtain the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

The target loss determination module is configured to determine the target loss according to the interest prediction index and the interest labeling index;

The network training module is configured to train the neural network based on the target loss to obtain the interest recognition network.

In some embodiments, the above-mentioned apparatus further comprises:

a recommendation time determination module, configured to determine the recommendation time of each sample recommendation information;

The data filtering module is configured to perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain the filtering behavior data set;

The sample behavior data determination module is further configured to determine sample behavior data corresponding to each sample recommendation information from the corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.

Regarding the apparatus in the above-mentioned embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be described in detail here.

FIG. 7 is a block diagram of an electronic device for information recommendation according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as shown in FIG. 7 . The electronic device includes a processor, memory, and a network interface connected by a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a computer-readable storage medium and an internal memory. The computer-readable storage medium stores an operating system and a computer program. The internal memory provides an environment for the execution of the operating system and computer programs in the computer-readable storage medium. The network interface of the electronic device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements an information recommendation method.

Those skilled in the art can understand that the structure shown in FIG. 7 is only a block diagram of a partial structure related to the solution of the present disclosure, and does not constitute a limitation on the electronic device to which the solution of the present disclosure is applied. The specific electronic device may be Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

In an exemplary embodiment, an electronic device is also provided, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to the instructions to implement embodiments as disclosed in the present disclosure The information in the recommended method.

In an exemplary embodiment, a computer-readable storage medium is also provided, when the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can perform the information recommendation in the embodiments of the present disclosure method. In some embodiments, the computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

In an exemplary embodiment, there is also provided a computer program product containing instructions, which, when executed on a computer, cause the computer to execute the information recommendation method in the embodiments of the present disclosure.

All the embodiments of the present disclosure may be implemented independently or in combination with other embodiments, which are all regarded as the protection scope required by the present disclosure.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. During execution, the processes of the embodiments of the above-mentioned methods may be included. Wherein, any reference to memory, storage, database or other medium used in the various embodiments provided in this application may include non-volatile and/or volatile memory.

Claims (26)

1. An information recommendation method including:

   In response to the information acquisition request of the target object, acquiring a long-term behavior data set of the target object, where the long-term behavior data set represents a plurality of historical behavior data of the target object within a preset time period;

   Determine the feature information of each recommended information in the recommended information set;

   Determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

   Inputting the target behavior data into an interest recognition network for interest recognition to obtain the target object's interest index for each recommendation information;

   The target information in the recommendation information set is recommended to the target object based on the interest index.
2. The information recommendation method according to claim 1, wherein the determining the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information comprises:

   Obtaining behavior feature information of each historical behavior data in the long-term behavior data set;

   calculating the first similarity between the feature information of each recommendation information and the behavior feature information;

   According to the first similarity, target behavior data corresponding to each recommendation information is determined from the long-term behavior data set.
3. The information recommendation method according to claim 1, wherein, in response to the information acquisition request of the target object, acquiring the long-term behavior data set of the target object comprises:

   receiving an information acquisition request from the target object, where the information acquisition request includes a target object identifier of the target object;

   A long-term behavior data set of the target object is acquired based on the target object identifier.
4. The information recommendation method according to claim 3, wherein the acquiring the long-term behavior data set of the target object based on the object identifier comprises:

   Obtaining a pre-built historical behavior database, the historical behavior database includes historical behavior data of multiple objects, and the mapping relationship between the object identifiers of the multiple objects and the corresponding historical behavior data;

   Obtain historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

   The corresponding historical behavior data is used as the long-term behavior data set of the target object.
5. The information recommendation method according to claim 1, wherein the determining the characteristic information of each recommendation information in the recommendation information set comprises:

   obtaining the subject tag of each of the recommended information;

   generating feature information of each recommendation information based on the subject tag of each recommendation information;

   or,

   extracting the text information of each recommended information;

   The feature information of each recommendation information is generated based on the text information of each recommendation information.
6. The information recommendation method according to claim 1, wherein the recommending the target information in the recommended information set to the target object based on the interest index comprises:

   determining target information from the recommendation information set according to the interest index;

   The target information is recommended to the target object.
7. The information recommendation method according to any one of claims 1 to 6, further comprising:

   Obtain information identification and sample feature information of multiple sample recommendation information;

   Determine the historical recommendation object corresponding to each sample recommendation information and the interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

   obtaining a long-term sample behavior data set of the historical recommendation object;

   Determine the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of the each sample recommendation information;

   Inputting the sample behavior data into the neural network for interest identification, and obtaining the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

   determining a target loss according to the interest prediction index and the interest labeling index;

   The neural network is trained based on the target loss to obtain the interest recognition network.
8. The information recommendation method according to claim 7, further comprising:

   determining the recommendation time of the recommendation information for each sample;

   Perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtered behavior data set;

   The sample behavior data corresponding to each sample recommendation information determined from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information includes:

   The sample behavior data corresponding to each sample recommendation information is determined from the corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.
9. An information recommendation device, comprising:

   The first behavior data set acquisition module is configured to obtain a long-term behavior data set of the target object in response to an information acquisition request of the target object, and the long-term behavior data set represents the number of the target object in a preset time period. historical behavioral data;

   a feature information determination module, configured to determine feature information of each recommendation information in the recommendation information set;

   a target behavior data determination module, configured to determine target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

   a first interest identification module, configured to input the target behavior data into an interest identification network for interest identification, and obtain an interest index of the target object for each recommendation information;

   An information recommendation module, configured to recommend target information in the recommended information set to the target object based on the interest index.
10. The information recommendation device according to claim 9, wherein the target behavior data determination module comprises:

    a behavior feature information acquisition unit, configured to acquire behavior feature information of each historical behavior data in the long-term behavior data set;

    a first similarity calculation unit, configured to calculate a first similarity between the feature information of each recommendation information and the behavior feature information;

    The target behavior data determining unit is configured to determine the target behavior data corresponding to each recommendation information from the long-term behavior data set according to the first similarity.
11. The information recommendation device according to claim 9, wherein the first behavior data set acquisition module comprises:

    an information acquisition request receiving unit, configured to receive an information acquisition request of the target object, where the information acquisition request includes a target object identifier of the target object;

    The long-term behavior data set acquisition unit is configured to acquire the long-term behavior data set of the target object based on the target object identifier.
12. The information recommendation device according to claim 11, wherein the long-term behavior data set acquisition unit comprises:

    A historical behavior database acquisition unit, configured to acquire a pre-built historical behavior database, the historical behavior database includes historical behavior data of a plurality of objects, and a mapping relationship between the object identifiers of the plurality of objects and the corresponding historical behavior data ;

    a historical behavior data acquisition unit, configured to acquire historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

    The long-term behavior data set determination unit is configured to use the corresponding historical behavior data as the long-term behavior data set of the target object.
13. The information recommendation device according to claim 9, wherein the feature information determination module comprises:

    a first topic tag acquiring unit, configured to acquire the topic tag of each of the recommended information;

    a first feature information generating unit, configured to generate feature information of each recommendation information based on the subject tag of each recommendation information;

    or,

    a first text information extraction unit, configured to extract the text information of each recommended information;

    The second feature information generating unit is configured to generate feature information of each recommendation information based on the text information of each recommendation information.
14. The information recommendation device according to claim 9, wherein the information recommendation module comprises:

    a target information determination unit, configured to determine target information from the recommended information set according to the interest index;

    An information pushing unit, configured to recommend the target information to the target object.
15. The information recommendation device according to any one of claims 9 to 14, wherein the device further comprises:

    an information acquisition module, configured to acquire information identifiers and sample feature information of multiple sample recommendation information;

    an object information determination module, configured to determine a historical recommendation object corresponding to each sample recommendation information and an interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

    The second behavior data set obtaining module is configured to obtain the long-term sample behavior data set of the historical recommendation object;

    a sample behavior number determination module, configured to determine sample behavior data corresponding to each sample recommendation information from a corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information;

    The second interest identification module is configured to input the sample behavior data into the neural network for interest identification, and obtain the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

    a target loss determination module, configured to determine a target loss according to the interest prediction index and the interest labeling index;

    A network training module configured to train the neural network based on the target loss to obtain the interest recognition network.
16. The information recommendation device according to claim 15, further comprising:

    a recommendation time determination module, configured to determine the recommendation time of the recommended information for each sample;

    a data filtering module, configured to perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtering behavior data set;

    The sample behavior data determination module is further configured to determine sample behavior data corresponding to each sample recommendation information from a corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.
17. An electronic device comprising:

    processor;

    a memory for storing the processor-executable instructions;

    wherein the processor is configured to:

    In response to the information acquisition request of the target object, acquiring a long-term behavior data set of the target object, where the long-term behavior data set represents a plurality of historical behavior data of the target object within a preset time period;

    Determine the feature information of each recommended information in the recommended information set;

    Determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

    Inputting the target behavior data into an interest recognition network for interest recognition to obtain the target object's interest index for each recommendation information;

    The target information in the recommendation information set is recommended to the target object based on the interest index.
18. The electronic device of claim 17, the processor configured to:

    Obtaining behavior feature information of each historical behavior data in the long-term behavior data set;

    calculating the first similarity between the feature information of each recommendation information and the behavior feature information;

    According to the first similarity, target behavior data corresponding to each recommendation information is determined from the long-term behavior data set.
19. The electronic device of claim 17, the processor configured to:

    receiving an information acquisition request from the target object, where the information acquisition request includes a target object identifier of the target object;

    A long-term behavior data set of the target object is acquired based on the target object identifier.
20. The electronic device of claim 19, the processor configured to:

    Obtaining a pre-built historical behavior database, the historical behavior database includes historical behavior data of multiple objects, and the mapping relationship between the object identifiers of the multiple objects and the corresponding historical behavior data;

    Obtain historical behavior data corresponding to the target object identifier from the historical behavior database according to the mapping relationship;

    The corresponding historical behavior data is used as the long-term behavior data set of the target object.
21. The electronic device of claim 17, the processor configured to:

    obtaining the subject tag of each of the recommended information;

    generating feature information of each recommendation information based on the subject tag of each recommendation information;

    or,

    extracting the text information of each recommended information;

    The feature information of each recommendation information is generated based on the text information of each recommendation information.
22. The electronic device of claim 17, the processor configured to:

    determining target information from the recommendation information set according to the interest index;

    The target information is recommended to the target object.
23. The electronic device of claim 17, the processor further configured to:

    Obtain information identification and sample feature information of multiple sample recommendation information;

    Determine the historical recommendation object corresponding to each sample recommendation information and the interest labeling index corresponding to the historical recommendation object based on the information identifier of each sample recommendation information;

    obtaining a long-term sample behavior data set of the historical recommendation object;

    Determine the sample behavior data corresponding to each sample recommendation information from the corresponding long-term sample behavior data set based on the sample feature information of the each sample recommendation information;

    Inputting the sample behavior data into the neural network for interest identification, and obtaining the interest prediction index of each historical recommendation object to the corresponding sample recommendation information;

    determining a target loss according to the interest prediction index and the interest labeling index;

    The neural network is trained based on the target loss to obtain the interest recognition network.
24. The electronic device of claim 23, the processor further configured to:

    determining the recommendation time of the recommendation information for each sample;

    Perform data filtering on the corresponding long-term sample behavior data set based on the recommended time to obtain a filtered behavior data set;

    The sample behavior data corresponding to each sample recommendation information determined from the corresponding long-term sample behavior data set based on the sample feature information of each sample recommendation information includes:

    The sample behavior data corresponding to each sample recommendation information is determined from the corresponding filtering behavior data set based on the sample feature information of each sample recommendation information.
25. A computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enabling the electronic device to execute:

    In response to the information acquisition request of the target object, acquiring a long-term behavior data set of the target object, where the long-term behavior data set represents a plurality of historical behavior data of the target object within a preset time period;

    Determine the feature information of each recommended information in the recommended information set;

    Determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

    Inputting the target behavior data into an interest recognition network for interest recognition to obtain the target object's interest index for each recommendation information;

    The target information in the recommendation information set is recommended to the target object based on the interest index.
26. A computer program product, comprising a computer program, wherein the computer program implements an information recommendation method when executed by a processor,

    This information recommendation method includes:

    In response to the information acquisition request of the target object, acquiring a long-term behavior data set of the target object, where the long-term behavior data set represents a plurality of historical behavior data of the target object within a preset time period;

    Determine the feature information of each recommended information in the recommended information set;

    Determine the target behavior data corresponding to each recommendation information from the long-term behavior data set based on the feature information of each recommendation information;

    Inputting the target behavior data into an interest recognition network for interest recognition to obtain the target object's interest index for each recommendation information;

    The target information in the recommendation information set is recommended to the target object based on the interest index.

Images