An information recommendation method and device, electronic equipment and storage medium

Abstract

The application discloses an information recommendation method and device, electronic equipment and a storage medium, and belongs to the field of data processing. The information recommendation method comprises the following steps: in response to the feedback behavior of a target user to recommended information, acquiring the characteristics of the recommended information, and recalling a plurality of candidate information according to the characteristics of the recommended information; determining the relationship characteristics between the target user and the publishing user of the recommended information; determining the interest object of the target user according to the characteristics of the recommended information and the relationship characteristics; determining the order of the plurality of candidate information based on the interest object of the target user; and recommending the candidate information to the target user according to the order of the plurality of candidate information.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of data processing, and particularly relates to an information recommendation method and device, electronic equipment and a storage medium. BACKGROUND

[0002] After entering the mobile network era, information flow and short video have become the most important channel for users to obtain information. A recommendation system can filter in a large amount of information flow according to user portraits and behavior data, and provide personalized recommendations for each user.

[0003] The social subscription distribution scenario refers to a scenario in which, in an information flow distribution scenario, a user subscribes to a publishing user, and information published by the publishing user is recommended to the user. Then, the user generates feedback behavior on the recommended information, or generates social behavior on the publishing user of the recommended information.

[0004] The recommendation method in the related art learns historical behavior data of a user and features of recommended information to recommend information to the user, and the obtained interests of the user are not comprehensive, which cannot meet the information recommendation in the social subscription distribution scenario, and affects the accuracy of information recommendation. SUMMARY

[0005] The embodiments of the application provide an information recommendation method, device, electronic equipment and storage medium, which can meet the information recommendation in the subscription distribution scenario with social needs, and improve the accuracy of information recommendation.

[0006] In order to achieve the above technical effects, the application is implemented as follows:

[0007] In a first aspect, the embodiments of the application provide an information recommendation method, comprising: in response to a behavior generated by a target user on recommended information, acquiring features of the recommended information, and recalling a plurality of candidate information according to the features of the recommended information; determining a relationship feature between the target user and a publishing user of the recommended information; determining an interest degree of the target user to an interest object according to the features of the recommended information and the relationship feature, the interest object including content of the recommended information and the publishing user of the recommended information; determining an order of the plurality of candidate information based on the interest degree of the target user to the interest object; and performing information recommendation to the target user according to the order of the plurality of candidate information.

[0008] In a second aspect, an embodiment of the present application provides an information recommendation device, comprising: an acquisition module, configured to acquire a feature of recommendation information in response to a behavior of a target user to the recommendation information, and recall a plurality of candidate information according to the feature of the recommendation information; a first operation module, configured to determine a relationship feature between the target user and a publishing user of the recommendation information; a recall module, configured to determine an interest degree of the target user to an interest object according to the feature of the recommendation information and the relationship feature, the interest object comprising content of the recommendation information and the publishing user of the recommendation information; a second operation module, configured to determine an order of the plurality of candidate information based on the interest degree of the target user to the interest object; and a recommendation module, configured to recommend information to the target user according to the order of the plurality of candidate information.

[0009] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory, wherein the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the information recommendation method according to the first aspect.

[0010] In a fourth aspect, an embodiment of the present application provides a readable storage medium, wherein the readable storage medium stores programs or instructions, and the programs or instructions are executed by a processor to implement the steps of the information recommendation method according to the first aspect.

[0011] In the embodiments of the present application, the feature of the recommendation information is acquired in response to the behavior of the target user to the recommendation information, and the plurality of candidate information is recalled according to the feature of the recommendation information; the interest degree of the target user to the interest object is determined according to the feature of the recommendation information and the relationship feature between the target user and the publishing user of the recommendation information, the interest object comprising the content of the recommendation information and the publishing user of the recommendation information; and the order of the plurality of candidate information is determined based on the interest degree of the target user to the interest object, so as to recommend information to the target user. The scheme can determine the real intention of the behavior of the target user to the recommendation information according to the behavior of the target user to the recommendation information and the relationship feature between the target user and the publishing user of the recommendation information, and recommend a plurality of blog information meeting the interest of the target user to the target user in real time, so as to meet the information recommendation in the subscription distribution scene with social needs and improve the accuracy of the information recommendation. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 FIG. 1 shows a flow diagram of an information recommendation method according to an embodiment of the present application;

[0013] Figure 2 FIG. 2 shows another flow diagram of an information recommendation method according to an embodiment of the present application;

[0014] Figure 3 FIG. 1 shows a structural schematic diagram of an information recommendation device provided by an embodiment of the present application;

[0015] Figure 4 FIG. 2 shows a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art belong to the scope of protection of the present application.

[0017] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be exchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a category and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the front and rear associated objects are in an "or" relationship.

[0018] As described previously, the recommendation method in the related art performs information recommendation on a user by learning historical behavior data of the user and features of recommended information. Since the relationship features between the user and a publisher of the information are ignored, the obtained interests of the user are not comprehensive, and the information recommendation cannot meet the information recommendation in a social subscription distribution scenario, affecting the accuracy of the information recommendation.

[0019] In view of this, the embodiments of the present application provide an information recommendation method, an information recommendation device, an electronic device, and a readable storage medium, to improve the accuracy of information recommendation.

[0020] The information recommendation method, the information recommendation device, the electronic device, and the readable storage medium provided by the embodiments of the present application will be described in detail below with reference to the drawings, through specific embodiments and application scenarios.

[0021] Figure 1This diagram illustrates a flowchart of an information recommendation method provided by an embodiment of this application. This method can be executed by an electronic device, such as a terminal device or a server device. In other words, the method can be executed by software or hardware installed on the terminal device or server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. (Refer to...) Figure 1 The method may include the following steps:

[0022] S101: In response to the target user's behavior toward the recommendation information, obtain the features of the recommendation information, and recall multiple candidate information based on the features of the recommendation information.

[0023] Specifically, the behaviors of target users towards recommended information typically include positive feedback behaviors and negative feedback behaviors. Positive feedback behaviors include, but are not limited to, clicking, commenting, forwarding, liking, sharing, and reading, reflecting that the target user has a certain interest in the recommended information. Negative feedback behaviors include marking as uninterested and complaining, reflecting that the recommended information fails to accurately meet the target user's interests. It can be understood that this application embodiment determines the characteristics of recommended information that the target user is interested in by responding to positive feedback behaviors such as clicking, commenting, forwarding, liking, sharing, and reading generated by the target user towards recommended information.

[0024] The characteristics of the aforementioned recommendation information include the characteristics of the content of the recommendation information or the characteristics of the associated users, such as the tags or topics carried by the recommendation information, the users who published the recommendation information, etc.

[0025] As an example, a target user's positive feedback behavior towards recommended information can be seen as indicating that the target user has a certain interest in the recommended information. Candidate information recalled based on the features of the recommended information aligns with the target user's interests in at least one aspect. For instance, in a news recommendation scenario, if a target user likes and comments on a news article tagged with "food safety," this tag can be considered a feature of that recommended information, thus recalling multiple candidate articles with the same "food safety" tag; or, using "news" as a feature of the recommended information, multiple candidate articles of the same news type can be recalled.

[0026] Optionally, to avoid recommending duplicate information to the target user, the recalled candidate information is blog posts in the recommended information that the target user has not engaged in.

[0027] S102: Determine the relationship characteristics between the target user and the user who published the recommendation information.

[0028] Optionally, the relationship feature includes, but is not limited to, the number of interactions between the target user and the publishing user, the number of feedback behaviors of the target user to the information published by the publishing user, the closeness feature between the target user and the publishing user, and the identity relationship between the target user and the publishing user. For example, the identity relationship between the target user and the publishing user can be colleagues, relatives, friends, mutual attention, etc.

[0029] S103: determining the interest degree of the target user to the interest object according to the feature of the recommendation information and the relationship feature.

[0030] Specifically, the interest object includes the content of the recommendation information and the publishing user of the recommendation information; the feedback behavior of the target user to the recommendation information is affected by the interest degree of the target user to the interest object. It should be understood that in the social subscription distribution scenario, the blog information published by the bloggers or subscription numbers followed by the user will be distributed to the user as the recommendation information. The intention of the user to generate behavior to the recommendation information is different each time, sometimes the user is interested in the content of the recommendation information, and sometimes the user is interested in the bloggers or subscription numbers. The interest degree is used to represent the real intention of the target user to generate behavior based on the target user to the recommendation information, that is, in the case that the target user generates behavior to the recommendation information, the target user is interested in the content of the recommendation information or the publishing user of the recommendation information.

[0031] For example, the user generates a like behavior to the blog information published by the followed blogger "someone" about "X", and the like behavior of the user can be due to a high interest degree in the topic content "X" or a high support or interest degree in "someone".

[0032] By determining the interest degree of the target user to the interest object, the reason for the target user to generate behavior to the recommendation information in the subscription distribution scenario with social needs can be more accurately judged, and information more consistent with the interest of the target user can be recommended, thereby improving the accuracy of information recommendation.

[0033] S104: determining the order of the plurality of candidate information based on the interest degree of the target user to the interest object.

[0034] Specifically, based on the interest degree of the target user to the interest object, instant feedback is made on the candidate information, so that the order of the plurality of candidate information is associated with the interest object with higher interest degree of the target user. For example, after the target user generates a behavior on the recommended information with the "food safety" label, the order of the information related to "X" in the candidate information is advanced, so that the target user obtains the recommended information related to "X" preferentially when refreshing to obtain more recommended information. In this way, the target user can obtain more recommended information that meets his interest in time, and the accuracy of information recommendation is improved. The frequency and duration of the behavior of the target user can be increased.

[0035] Optionally, the order of the plurality of candidate information is positively correlated with the interest object of the target user, that is, the higher the positive correlation between the candidate information and the interest object of the target user, the higher the order of the candidate information.

[0036] S105: According to the order of the plurality of candidate information, information is recommended to the target user.

[0037] According to the order of the candidate information determined in step S104, information is recommended to the target user, so that the target user can obtain information that meets his interest preferentially.

[0038] Optionally, according to the order of the plurality of candidate information, a preset number of candidate information is recommended to the target user. For example, the top 20 candidate information is recommended to the target user.

[0039] Optionally, due to the size limitation of the mobile device, the plurality of candidate information is recommended to the target user in batches. For example, in the case that the target user uses a smart phone to view recommended information, 4-6 blog information is displayed in the current page each time.

[0040] In the embodiment of the present application, by responding to the behavior generated by the target user on the recommended information, the characteristics of the recommended information are obtained, and according to the characteristics of the recommended information, a plurality of candidate information is recalled; the relationship characteristics between the target user and the publisher of the recommended information are determined; according to the characteristics of the recommended information and the relationship characteristics, the interest degree of the target user to the interest object is determined, the interest object including the content of the recommended information and the publisher of the recommended information; based on the interest degree of the target user to the interest object, the order of the plurality of candidate information is determined; according to the order of the plurality of candidate information, information is recommended to the target user, which can recommend a plurality of blog information that meets the interest of the target user according to the behavior generated by the target user and the relationship characteristics between the user and the publisher of the information, meet the information recommendation in the subscription distribution scene with social needs, and improve the accuracy of information recommendation.

[0041] Figure 2 Another flowchart of the information recommendation method provided by the embodiments of the present application is shown, referring to Figure 2 The method can include the following steps:

[0042] S201: In response to the behavior of the target user generated by the recommended information, the characteristics of the recommended information are obtained, and according to the characteristics of the recommended information, a plurality of candidate information is recalled.

[0043] In an implementation manner, the characteristics of the recommended information include the content characteristics of the recommended information and the user characteristics of the publishing user of the recommended information.

[0044] Optionally, the content characteristics of the recommended information include the classification label of the recommended information, whether the recommended information is related to a hot spot, the interaction data or the exposure times of the recommended information, such as the number of comments, the number of forwarding, etc. The user characteristics of the publishing user of the recommended information include the professional field of the publishing user, whether the publishing user is authenticated, the number of subscriptions of the publishing user, etc.

[0045] S202: Determine the relationship characteristics between the target user and the publishing user of the recommended information.

[0046] The description of step S202 can refer to Figure 1 Step S102 in the embodiments, which will not be repeated here.

[0047] S203: The content characteristics of the recommended information and the user characteristics of the publishing user of the recommended information, and the relationship characteristics are input into a first model to determine the interest degree of the target user to the content of the recommended information, and the first interest value of the target user to the content of the recommended information is obtained.

[0048] Specifically, the first model is used to determine the interest degree of the target user to the content of the recommended information. The first interest value can represent the interest degree of the target user to the content of the recommended information. The greater the first interest value, the greater the interest degree of the target user to the content of the recommended information.

[0049] As an example, the user has a like behavior on the blog information about "X" published by the blogger "someone". The first model can determine that the user's like behavior is because of the high interest degree on the topic content "X" by determining the content features of the recommendation information, the user features of the publishing user of the recommendation information, and the relationship features.

[0050] S204: Determine a second interest value of the target user on the publishing user of the recommendation information based on the first interest value of the target user on the content of the recommendation information.

[0051] The second interest value can represent the interest degree of the target user on the publishing user of the recommendation information.

[0052] Specifically, the first interest value and the second interest value are negatively correlated. That is, the greater the first interest value, the smaller the second interest value. As an example, the user has a comment behavior on a blog information of a travel photo published by a friend "A". The first model determines that the user has a low interest degree on the topic content such as "travel" or "daily life record". Because "travel" is not a hot topic at present, it belongs to the "daily life record" of the blogger "A". The user also does not have behaviors on the blog information related to such topics. The user generates a comment behavior because of the high interest degree on the publishing user "A". Because the blogger "A" is an unauthenticated account with a small number of fans and a relatively low frequency of publishing blog information, and the relationship feature between the user and the blogger "A" is mutual attention and friendship, the user has interacted with the blogger "A" by sending messages many times in the past. Based on the above features, it can be inferred that the user has a high interest degree on the blogger, i.e., the publishing user of the recommendation information.

[0053] Optionally, the first interest value can be a probability value between 0 and 1, and the sum of the interest degree of the target user on the content of the recommendation information and the interest degree on the publishing user of the recommendation information is 1.

[0054] S205: Input the features of the recommendation information and the content features of the candidate information into a second model to predict the probability of the target user generating a behavior on the candidate information, and obtain an output behavior probability value corresponding to the candidate information.

[0055] Specifically, the second model is configured to determine a probability of the target user generating a behavior in response to the candidate information. As an example, the second model can be a CTR (Click-Through Rate) estimation model. The higher the behavior probability value of the candidate information, the higher the interest of the target user in the candidate information, and the greater the possibility of generating a positive feedback behavior.

[0056] Optionally, before inputting the features of the recommended information and the content features of the candidate information into the second model, the features of the recommended information and the features of the candidate information are respectively subjected to feature data processing, such as binary conversion of all features into 0 and 1, discretization of continuous features, smoothing of feature values, or vectorization of multiple features, so that the noise in the feature data has better robustness.

[0057] S206: According to the behavior probability value corresponding to the candidate information, the first interest value of the target user in the recommended information, and the second interest value of the target user in the publishing user of the recommended information, determine an interest score of the target user in the candidate information.

[0058] As an example, the interest score can be determined by multiplying the behavior probability value of the candidate information and the first interest value or the second interest value of the target user. For example, the probability score of the candidate information is 0.7, the first interest value or the second interest value of the target user is 0.8, and the interest score of the candidate information is 0.7*0.8=0.56. .

[0059] S207: According to the interest score of the target user in the candidate information, determine the order of the plurality of candidate information.

[0060] Specifically, the interest score reflects the degree of interest of the target user in the candidate information. As an example, the higher the interest score, the earlier the order of the candidate information corresponding to the interest score.

[0061] S208: According to the order of the plurality of candidate information, recommend information to the target user.

[0062] This step can refer to the description of step S104 in the embodiment, which will not be repeated here. Figure 1

[0063] ​In the embodiment of the present application, the feature of the recommendation information is obtained by responding to the behavior of the target user to the recommendation information, and a plurality of candidate information is recalled according to the feature of the recommendation information; the relationship feature between the target user and the publishing user of the recommendation information is determined; the content feature of the recommendation information and the user feature of the publishing user of the recommendation information, and the relationship feature are input into the first model to determine the interest degree of the target user to the content of the recommendation information, and the first interest value of the target user to the content of the recommendation information is obtained as output; the second interest value of the target user to the publishing user of the recommendation information is determined based on the first interest value of the target user to the content of the recommendation information; the content feature of the candidate information is input into the second model to predict the probability of the behavior of the target user to the candidate information, and the behavior probability value corresponding to the candidate information is obtained as output; the interest score of the target user to the candidate information is determined according to the behavior probability value corresponding to the candidate information, the first interest value of the target user to the recommendation information, and the second interest value of the target user to the publishing user of the recommendation information; the order of the plurality of candidate information is determined according to the interest score of the target user to the candidate information; and the information recommendation is performed on the target user according to the order of the plurality of candidate information, which can determine the intention of the user to generate the behavior according to the behavior generated by the target user and the relationship feature between the user and the publishing user of the information, and can recommend a plurality of blog information meeting the interest of the target user in time, thereby meeting the information recommendation in the subscription distribution scene with social needs and improving the accuracy of the information recommendation.

[0064] In an implementation manner, the plurality of candidate information includes at least one first candidate information and at least one second candidate information; the first candidate information is candidate information recalled based on the content of the recommendation information, and the second candidate information is candidate information recalled based on the publishing user of the recommendation information.

[0065] The interest score of the target user to the candidate information is determined according to the behavior probability value corresponding to the candidate information, the first interest value of the target user to the content of the recommendation information, and the second interest value of the target user to the publishing user of the recommendation information, and the order of the plurality of candidate information is determined according to the interest score of the target user to the candidate information; the information recommendation is performed on the target user according to the order of the plurality of candidate information, which can determine the intention of the user to generate the behavior according to the behavior generated by the target user and the relationship feature between the user and the publishing user of the information, and can recommend a plurality of blog information meeting the interest of the target user in time, thereby meeting the information recommendation in the subscription distribution scene with social needs and improving the accuracy of the information recommendation.

[0066] In the case that the candidate information is the first candidate information, the interest score of the target user to the first candidate information is determined according to the behavior probability value corresponding to the first candidate information and the first interest value.

[0067] In the case that the candidate information is the second candidate information, the interest score of the target user to the second candidate information is determined according to the behavior probability value corresponding to the second candidate information and the second interest value.

[0068] Optionally, a first candidate sequence corresponding to the first candidate information and a second candidate sequence corresponding to the second candidate information are determined respectively. Or the interest score of the target user on the candidate information is taken as a unified standard to determine a third candidate sequence including the first candidate information and the second candidate information, so that the target user obtains multiple recommended information meeting the interest thereof.

[0069] The method of the embodiments of the present application can recall candidate information based on different interest objects of users, can increase the diversity of the recalled candidate information, meet the information recommendation in the subscription distribution scene with social needs, and make the target user more accurately obtain multiple recommended information meeting the interest thereof.

[0070] In an implementation manner, the features of the recommended information further include an embedding feature corresponding to the content of the recommended information and an embedding feature corresponding to the publishing user of the recommended information.

[0071] The recalling multiple candidate information according to the features of the recommended information includes at least one of the following:

[0072] (1) According to the content feature of the recommended information, a blog information with a same content feature as the recommended information is recalled as the first candidate information; specifically, the recalled first candidate information has a same content feature as the recommended information. For example, according to the classification label (such as “funny and humorous”) of the recommended information, a first candidate information with a same / similar classification label (such as a candidate information with “funny”, “humorous” or “funny and humorous”) as the recommended information is recalled.

[0073] (2) According to the embedding feature corresponding to the content of the recommended information, a blog information similar to the content of the recommended information is recalled as the first candidate information.

[0074] In an implementation manner, the recalling, according to the embedding feature corresponding to the content of the recommended information, a blog information similar to the content of the recommended information as the first candidate information includes:

[0075] Constructing a first vector index according to the embedding feature corresponding to the content of the recommended information.

[0076] Constructing a third vector index according to the embedding feature corresponding to the content of the blog information to be recalled.

[0077] Calculating the content similarity between the recommended information and the blog information to be recalled based on the first vector index and the third vector index.

[0078] The blog information with a high content similarity is recalled as the first candidate information.

[0079] Specifically, the embedding feature corresponding to the content of the recommendation information is obtained by mapping the content of the recommendation information to a high-dimensional vector space. The vector index is a data index structure with high efficiency in time and space, which is constructed by a mathematical quantization model based on the embedding feature. The vector index is used to more accurately obtain a vector or embedding feature with a high similarity, so as to recall blog information with a high similarity as the first candidate information, and improve the accuracy and efficiency of recalling blog information.

[0080] For example, the first candidate information is recalled according to the embedding feature corresponding to the content of the recommendation information, and includes blog information of the same event, the same hot spot, or the same classification label as the content of the recommendation information.

[0081] (3) The blog information published by the user who publishes the recommendation information is recalled as the second candidate information according to the user feature of the user who publishes the recommendation information.

[0082] Specifically, the other blog information published by the user who publishes the recommendation information is determined as the second candidate information to be recalled. For example, the user has just performed a click behavior on a blog information published by a blogger “someone”, and N blog information published by “someone” and not acted on by the user is recalled as the second candidate information.

[0083] (4) The blog information published by other users similar to the user who publishes the recommendation information is recalled as the second candidate information according to the embedding feature corresponding to the user who publishes the recommendation information.

[0084] In an implementation manner, the blog information published by other users similar to the user who publishes the recommendation information is recalled as the second candidate information according to the embedding feature corresponding to the user who publishes the recommendation information, including:

[0085] A second vector index is constructed according to the embedding feature corresponding to the user who publishes the recommendation information.

[0086] A fourth vector index is constructed according to the embedding feature corresponding to the other user to be recalled.

[0087] The user similarity between the user who publishes the recommendation information and the other user is calculated based on the second vector index and the fourth vector index.

[0088] The blog information published by a preset number of other users with a high user similarity is recalled as the second candidate information.

[0089] Specifically, the identity feature vector of the publishing user is obtained by mapping the user feature of the publishing user of the recommendation information through a high-dimensional vector space.

[0090] For example, the other user similar to the publishing user of the recommendation information includes a blogger having the same social relationship or the same type of friend circle as the publishing user.

[0091] Optionally, the embedding feature is composed of multi-dimensional float type (float) numerical values, such as [0.11, 0.32, 0.24, 0.56, …, 0.19], [0.35, 0.27, 0.46, 0.36, …, 0.28], etc.

[0092] In the embodiments of the present application, the blog information having the same type of content feature as the recommendation information is recalled as the first candidate information according to the content feature of the recommendation information; the blog information similar to the content of the recommendation information is recalled as the first candidate information according to the embedding feature corresponding to the content of the recommendation information; the blog information published by the publishing user is recalled as the second candidate information according to the user feature of the publishing user of the recommendation information; and the blog information published by the other user similar to the publishing user is recalled as the second candidate information according to the embedding feature corresponding to the publishing user of the recommendation information. At least one of the above-mentioned recalling methods is used to realize multi-path recall of the candidate information, increase the diversity of information recommendation, and improve the accuracy of candidate information recall through the high-dimensional vector embedding feature as the representation method of the feature.

[0093] Figure 3 FIG. 1 is a structural schematic diagram of an information recommendation device according to an embodiment of the present application. As shown in FIG. 1, the information recommendation device 30 includes an acquisition module 31, a first operation module 32, a second operation module 33, a third operation module 34, and a recommendation module 35. Figure 3

[0094] ​The acquisition module 31 is configured to acquire a feature of the recommendation information in response to a behavior of a target user generated for the recommendation information, and recall a plurality of candidate information according to the feature of the recommendation information; the first operation module 32 is configured to determine a relationship feature between the target user and a publishing user of the recommendation information; the second operation module 33 is configured to determine an interest degree of the target user to an interest object according to the feature of the recommendation information and the relationship feature, the interest object including content of the recommendation information and the publishing user of the recommendation information; the third operation module 34 is configured to determine an order of the plurality of candidate information based on the interest degree of the target user to the interest object; and the recommendation module 35 is configured to recommend information to the target user according to the order of the plurality of candidate information.

[0095] In an implementation manner, the feature of the recommendation information includes a content feature of the recommendation information and a user feature of the publishing user of the recommendation information; the second operation module 33 is further configured to input the content feature of the recommendation information and the user feature of the publishing user, and the relationship feature into a first model to determine the interest degree of the target user to the content of the recommendation information, and obtain an output first interest value of the target user to the content of the recommendation information; determine a second interest value of the target user to the publishing user of the recommendation information based on the first interest value of the target user to the content of the recommendation information; and the greater the first interest value is, the smaller the second interest value is.

[0096] In an implementation manner, the third operation module 34 is further configured to input the feature of the recommendation information and a content feature of the candidate information into a second model to perform a probability prediction of a behavior of the target user generated for the candidate information before determining the order of the plurality of candidate information based on the interest degree of the target user to the interest object, and obtain an output behavior probability value corresponding to the candidate information. In an implementation manner, the third operation module 34 is further configured to determine an interest score of the target user to the candidate information according to the behavior probability value corresponding to the candidate information, the first interest value of the target user to the content of the recommendation information, and the second interest value of the target user to the publishing user of the recommendation information; and determine the order of the plurality of candidate information according to the interest score of the target user to the candidate information.

[0097] In an implementation manner, the plurality of candidate information includes at least one first candidate information and at least one second candidate information; the first candidate information is candidate information recalled based on content of the recommendation information, and the second candidate information is candidate information recalled based on a publishing user of the recommendation information; the third operation module 34 is further configured to, in a case where the candidate information is the first candidate information, determine an interest score of the target user for the first candidate information according to a behavior probability value corresponding to the first candidate information and the first interest value; and in a case where the candidate information is the second candidate information, determine an interest score of the target user for the second candidate information according to a behavior probability value corresponding to the second candidate information and the second interest value.

[0098] In an implementation manner, the features of the recommendation information further include an embedding feature corresponding to content of the recommendation information and an embedding feature corresponding to a publishing user of the recommendation information; and the obtaining module 31 is further configured to recall a plurality of candidate information according to at least one of the following: according to the content feature of the recommendation information, recall a blog information having a same content feature as the recommendation information as the first candidate information; according to the embedding feature corresponding to the content of the recommendation information, recall a blog information similar to the content of the recommendation information as the first candidate information; according to a user feature of the publishing user of the recommendation information, recall a blog information published by the publishing user as the second candidate information; and according to the embedding feature corresponding to the publishing user of the recommendation information, recall a blog information published by another user similar to the publishing user as the second candidate information.

[0099] In an implementation manner, the obtaining module 31 is further configured to construct a first vector index according to the embedding feature corresponding to the content of the recommendation information; construct a third vector index according to an embedding feature corresponding to content of a blog information to be recalled; calculate a content similarity between the recommendation information and the blog information to be recalled based on the first vector index and the third vector index; and recall a preset number of blog information with a higher content similarity as the first candidate information.

[0100] In an implementation manner, the obtaining module 31 is further configured to construct a second vector index according to the embedding feature corresponding to the publishing user of the recommendation information; construct a fourth vector index according to an embedding feature corresponding to another user to be recalled; calculate a user similarity between the publishing user and the another user based on the second vector index and the fourth vector index; and recall a preset number of blog information published by the another user with a higher user similarity as the second candidate information.

[0101] The device 30 provided by the embodiments of the present application can execute the methods described in the foregoing method embodiments, and achieve the functions and advantages of the methods described in the foregoing method embodiments, which will not be repeated here.

[0102] Figure 4 A hardware structure schematic diagram of an electronic device 400 executing the information recommendation method provided by the embodiments of the present application is shown. Referring to the diagram, at the hardware level, the electronic device 400 includes a processor 401, and optionally, an internal bus 404, a network interface 402, and a memory 403. The memory 403 can include an internal memory, for example, a high-speed random-access memory 403 (RAM), and can also include a non-volatile memory 403, for example, at least one disk memory 403, etc. Of course, the electronic device 400 can also include other hardware required by other services.

[0103] The processor 401, the network interface 402, and the memory 403 can be connected to each other through the internal bus 404, which can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one bidirectional arrow is used in the diagram, but it does not mean that there is only one bus or only one type of bus.

[0104] The memory 403 is used to store programs. Specifically, the programs can include program codes including computer operation instructions. The memory 403 can include an internal memory and a non-volatile memory 403, and provide instructions and data to the processor 401.

[0105] The processor 401 reads the corresponding computer programs from the non-volatile memory into the internal memory and then runs, and forms the device for positioning target users at the logical level. The processor 401 executes the programs stored in the memory 403, and is specifically used to execute the methods described in the foregoing method embodiments Figures 1-2 and achieve the functions and advantages of the methods described in the foregoing method embodiments Figures 1-2 , which will not be repeated here.

[0106] The above as described in the embodiments of the present application Figures 1-2The information recommendation method disclosed by the embodiment shown can be applied to or implemented by the processor 401. The processor 401 can be an integrated circuit chip with processing capability. In the implementation process, each step of the information recommendation method described above can be completed by the integrated logic circuit of hardware in the processor 401 or by the instruction in the form of software. The processor 401 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; or a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Each information recommendation method, step and logic block diagram disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor 401 can also be any conventional processor. The steps of the information recommendation method disclosed in combination with the embodiment of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 403, and the processor 401 reads the information in the memory 403, and combines the hardware to complete the steps of the information recommendation method described above.

[0107] The electronic device can also perform the information recommendation method in the foregoing method embodiment Figures 1-2 The electronic device can also perform the information recommendation method in the foregoing method embodiment Figures 1-2 The functions and advantages of each method described above are not repeated here.

[0108] Of course, in addition to the software implementation, the electronic device of the present application does not exclude other implementation manners, such as logic devices or software and hardware combined manner, etc. That is, the execution subject of the following processing flow is not limited to each logic unit, but also can be hardware or logic device.

[0109] The embodiment of the present application also proposes a computer readable storage medium, the computer readable medium stores one or more programs, the one or more programs when being executed by the electronic device including a plurality of application programs, make the electronic device execute the information recommendation method in the foregoing method embodiment Figures 1-2 The electronic device can also perform the information recommendation method in the foregoing method embodimentFigures 1-2 The functions and advantages of the methods described above are not repeated here.

[0110] The computer readable storage medium includes a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, etc.

[0111] Further, the embodiment of the present application further provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the functions of the above method embodiments are realized. Figures 1-2 The functions and advantages of the information recommendation methods described above are not repeated here.

[0112] In summary, the above only describes the preferred embodiments of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0113] The system, device, module or unit illustrated in the above embodiments can be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer. Specifically, the computer may, for example, be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, 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.

[0114] The computer readable medium includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules 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 technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, the computer readable medium does not include transitory computer readable media, such as modulated data signals and carriers.

[0115] It has to be noted that, as used herein, the terms "includes" and / or "contains", or any other tautological variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements is not limited to those elements, but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. In other words, without further restriction, the element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element. Furthermore, it is pointed out that the scope of the methods and apparatus of the present embodiments is not limited to performing functions in the order recited or discussed, but can include performing functions in a substantially simultaneous manner or in the reverse order, e.g., the described methods can be performed in a different order than described, and various steps can be added, omitted, or combined, and features described with respect to certain examples can be combined in other examples.

[0116] From the above description of the embodiments, it is apparent that the above-described method of the embodiments can be realized by means of software and the necessary universal hardware platform, of course, also by hardware, but in many cases the former is the better embodiment. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the method described in the various embodiments of the present application.

[0117] The embodiments of the present application are described above in conjunction with the drawings, but the present application is not limited to the above-described specific embodiments, which are merely illustrative, not restrictive, and a person of ordinary skill in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the scope protected by the claims.

Claims

1. An information recommendation method characterized by comprising: The method comprises the following steps: In response to the behavior of the target user to the recommended information, the characteristics of the recommended information are obtained, and a plurality of candidate information is recalled according to the characteristics of the recommended information, wherein the characteristics of the recommended information include the content characteristics of the recommended information and the user characteristics of the publishing user of the recommended information; The relationship characteristics between the target user and the publishing user of the recommended information are determined; According to the characteristics of the recommended information and the relationship characteristics, the interest degree of the target user to the interest object is determined, wherein the interest object includes the content of the recommended information and the publishing user of the recommended information; Based on the interest degree of the target user to the interest object, the order of the plurality of candidate information is determined; According to the order of the plurality of candidate information, information is recommended to the target user; The method comprises the following steps: The content characteristics of the recommended information and the user characteristics of the publishing user of the recommended information, and the relationship characteristics are input into a first model to determine the interest degree of the target user to the content of the recommended information, and a first interest value of the target user to the content of the recommended information is obtained as output; Based on the first interest value of the target user to the content of the recommended information, a second interest value of the target user to the publishing user of the recommended information is determined; wherein the greater the first interest value, the smaller the second interest value, the interest degree includes the first interest value and the second interest value, the first interest value represents the interest degree of the target user to the content of the recommended information, and the second interest value represents the interest degree of the target user to the publishing user of the recommended information.

2. The information recommendation method according to claim 1, characterized by, Before the step of determining the order of the plurality of candidate information based on the interest degree of the target user to the interest object, the method further comprises the following steps: The characteristics of the recommended information and the content characteristics of the candidate information are input into a second model to predict the probability of the target user to generate behavior to the candidate information, and a behavior probability value corresponding to the candidate information is obtained as output.

3. The information recommendation method according to claim 2, characterized by, The method comprises the following steps: According to the behavior probability value corresponding to the candidate information, the first interest value of the target user to the content of the recommended information, and the second interest value of the target user to the publishing user of the recommended information, the interest score of the target user to the candidate information is determined; According to the interest score of the target user to the candidate information, the order of the plurality of candidate information is determined.

4. The information recommendation method according to claim 3, characterized by, The plurality of candidate information includes at least one first candidate information and at least one second candidate information; the first candidate information is candidate information recalled based on the content of the recommended information, and the second candidate information is candidate information recalled based on the publishing user of the recommended information; The interest score of the target user for the candidate information is determined according to a behavior probability value corresponding to the candidate information, a first interest value of the target user for content of the recommendation information, and a second interest value of the target user for a publishing user of the recommendation information, and the interest score of the target user for the candidate information is determined according to a behavior probability value corresponding to the candidate information, a first interest value of the target user for content of the recommendation information, and a second interest value of the target user for a publishing user of the recommendation information. In a case where the candidate information is first candidate information, the interest score of the target user for the first candidate information is determined according to the behavior probability value corresponding to the first candidate information and the first interest value. In a case where the candidate information is second candidate information, the interest score of the target user for the second candidate information is determined according to the behavior probability value corresponding to the second candidate information and the second interest value.

5. The information recommendation method according to claim 4, characterized by, The features of the recommendation information further include an embedding feature corresponding to content of the recommendation information and an embedding feature corresponding to a publishing user of the recommendation information. The plurality of candidate information is recalled according to the features of the recommendation information, including at least one of the following: The blog information with the same content feature as the recommendation information is recalled as the first candidate information according to the content feature of the recommendation information. The blog information similar to the content of the recommendation information is recalled as the first candidate information according to the embedding feature corresponding to the content of the recommendation information. The blog information published by the publishing user of the recommendation information is recalled as the second candidate information according to the user feature of the publishing user of the recommendation information. The blog information published by other users similar to the publishing user of the recommendation information is recalled as the second candidate information according to the embedding feature corresponding to the publishing user of the recommendation information.

6. The information recommendation method according to claim 5, characterized by, The blog information similar to the content of the recommendation information is recalled as the first candidate information according to the embedding feature corresponding to the content of the recommendation information, including: A first vector index is constructed according to the embedding feature corresponding to the content of the recommendation information. A third vector index is constructed according to the embedding feature corresponding to the content of the blog information to be recalled. The content similarity between the recommendation information and the blog information to be recalled is calculated based on the first vector index and the third vector index. The blog information of a preset number of blog information with the highest content similarity is recalled as the first candidate information.

7. The information recommendation method according to claim 5, characterized by, The blog information published by other users similar to the publishing user of the recommendation information is recalled as the second candidate information according to the embedding feature corresponding to the publishing user of the recommendation information, including: A second vector index is constructed according to the embedding feature corresponding to the publishing user of the recommendation information. A fourth vector index is constructed according to the embedding feature corresponding to the other user to be recalled. The user similarity between the publishing user and the other user is calculated based on the second vector index and the fourth vector index. The blog information published by a preset number of other users with the highest user similarity is recalled as the second candidate information.

8. An information recommendation device characterized by comprising: ​ The acquisition module is configured to acquire a feature of the recommendation information in response to a behavior generated by a target user to the recommendation information, and recall a plurality of candidate information according to the feature of the recommendation information, wherein the feature of the recommendation information comprises a content feature of the recommendation information and a user feature of a publishing user of the recommendation information. The first operation module is configured to determine a relationship feature between the target user and the publishing user of the recommendation information. The second operation module is configured to determine an interest degree of the target user to an interest object according to the feature of the recommendation information and the relationship feature, wherein the interest object comprises a content of the recommendation information and the publishing user of the recommendation information. The third operation module is configured to determine an order of the plurality of candidate information based on the interest degree of the target user to the interest object. The recommendation module is configured to recommend information to the target user according to the order of the plurality of candidate information. The second operation module is specifically configured to input the content feature of the recommendation information and the user feature of the publishing user of the recommendation information, and the relationship feature into a first model to determine the interest degree of the target user to the content of the recommendation information, and obtain an output first interest value of the target user to the content of the recommendation information; and determine a second interest value of the target user to the publishing user of the recommendation information based on the first interest value of the target user to the content of the recommendation information, wherein the first interest value is greater, the second interest value is smaller, the interest degree comprises the first interest value and the second interest value, the first interest value represents the interest degree of the target user to the content of the recommendation information, and the second interest value represents the interest degree of the target user to the publishing user of the recommendation information.

9. An electronic device, comprising: The processor, the memory, and the program or the instructions stored on the memory and executable on the processor are included, and the program or the instructions are executed by the processor to implement the steps of the information recommendation method according to any one of claims 1 to 7.

10. A readable storage medium, characterized by, The program or the instructions are stored on the readable storage medium, and the program or the instructions are executed by the processor to implement the steps of the information recommendation method according to any one of claims 1 to 7.

Images