CN114048390A - Content recommendation method and device, electronic equipment and storage medium - Google Patents
Content recommendation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114048390A CN114048390A CN202210024223.6A CN202210024223A CN114048390A CN 114048390 A CN114048390 A CN 114048390A CN 202210024223 A CN202210024223 A CN 202210024223A CN 114048390 A CN114048390 A CN 114048390A
- Authority
- CN
- China
- Prior art keywords
- content
- user
- expert
- behavior
- behaviors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 230000003993 interaction Effects 0.000 claims abstract description 108
- 238000012216 screening Methods 0.000 claims abstract description 28
- 230000006399 behavior Effects 0.000 claims description 263
- 230000002452 interceptive effect Effects 0.000 claims description 75
- 238000004590 computer program Methods 0.000 claims description 10
- 238000010586 diagram Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 239000002360 explosive Substances 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- 230000002411 adverse Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a content recommendation method, a content recommendation device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user; screening relevant experts relevant to the user from the expert database based on the user representation of the user and the expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert; determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user; the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content. The invention improves the accuracy, efficiency and quality of content recommendation.
Description
Technical Field
The present invention relates to the field of information processing technologies, and in particular, to a content recommendation method and apparatus, an electronic device, and a storage medium.
Background
The content community is an important way for users to share knowledge and obtain answers, and is a general or vertical question-answer community, which is widely developed at home and abroad. The home page of the content community directly influences the impression of the user on the community, and a content recommendation strategy based on similar users in home page recommendation is also an important strategy in each company recommendation system. The content includes articles published by users in the community, answers to questions and other information.
The core idea of the UserCF recommendation strategy is that a user is interested in similar contents to the user. The implementation of the existing UserCF algorithm is generally divided into two steps: 1. taking an article with a behavior generated by a user as a user characteristic, and calculating a user similarity matrix; 2. and obtaining K users similar to the recommended user according to the user similarity matrix, accumulating the similarity as a weight, and recommending the top N items to the user from the items of which the similar user generates the behavior. However, the conventional UserCF recommendation strategy regularly generates user vectors every day, and the generation of the vectors requires the use of the interaction behavior of the user and the content in the latest period of time, so that the vector generated by the user with less recent behavior has poor effect, and the recommendation effect is poor. In addition, in a scene in which the algorithm is applied, the greater the number of users, the more difficult it is to calculate the user similarity matrix, and the difficulty is reflected in time complexity and space complexity, resulting in insufficient efficiency of the recommendation algorithm. Secondly, for a content community with high-quality content, the content recommendation accuracy in the professional field is important, and the current recommendation mode cannot meet the requirements.
Disclosure of Invention
The invention provides a content recommendation method, a content recommendation device, electronic equipment and a storage medium, which are used for solving the defects of poor recommendation effect and low calculation efficiency when the number of users is large in the prior art.
The invention provides a content recommendation method, which comprises the following steps:
acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user;
screening relevant experts relevant to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert;
determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user;
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
According to the content recommendation method provided by the invention, the historical behavior characteristics of the user or the expert are determined based on the following steps:
acquiring historical interaction behaviors of the user or the expert;
determining the content corresponding to the historical interactive behaviors and the popularity of the content corresponding to the historical interactive behaviors;
randomly discarding the content with the popularity higher than a preset threshold value in the content corresponding to the historical interaction behavior;
and counting the behaviors of the residual contents based on the user or the expert to obtain the historical behavior characteristics.
According to the content recommendation method provided by the invention, the popularity of the content corresponding to the historical interactive behavior is determined, and the method specifically comprises the following steps:
acquiring the interaction amount and/or ranking information of the content corresponding to the historical interaction behavior in a preset time period;
determining popularity of the content corresponding to the historical interaction behaviors based on the interaction amount and/or ranking information of the content corresponding to the historical interaction behaviors in a preset time period; the popularity of any content is higher as the interaction amount and/or ranking information of the content in a preset time period is higher.
According to the content recommendation method provided by the invention, the expert is determined based on the following steps:
calculating a content score of each content based on the interaction index of each content; the interaction index comprises at least one of a praise rate, a collection rate, a sharing rate, a thank you rate and a comment rate;
determining the latest interactive content corresponding to each mature user based on the interactive behavior of each mature user in the latest preset time period;
and screening out experts from each mature user based on the content score of the corresponding latest interactive content of each mature user.
According to the content recommendation method provided by the invention, the screening of experts from each mature user based on the content score of the latest interactive content corresponding to each mature user specifically comprises the following steps:
determining a user score of any mature user based on a content score of a most recent interactive content corresponding to the mature user and the number of users who pay attention to the mature user and are of a high-quality user type;
and screening out experts from the various mature users based on the user scores of the various mature users.
According to the content recommendation method provided by the invention, the determining of the content to be recommended based on the recent interactive behavior of the relevant experts specifically comprises the following steps:
acquiring the content related to the latest interactive behavior of the related experts; the recent interaction behaviors comprise click-type behaviors and attention-type behaviors;
determining recommendation scores of the contents related to the click-type behaviors based on click rates of the contents related to the click-type behaviors in the latest interactive behaviors in all experts in the expert database;
determining recommendation scores of contents related to the attention class behaviors based on behavior time of the attention class behaviors in the recent interactive behaviors;
and determining the content to be recommended based on the recommendation score of the content associated with the clicking type behavior and the recommendation score of the content associated with the attention type behavior.
According to the content recommendation method provided by the invention, the determining of the recommendation score of the content associated with the attention class behavior based on the behavior time of the attention class behavior in the recent interactive behavior specifically includes:
and if the content related to the attention class behaviors corresponds to a plurality of attention class behaviors of the relevant experts, taking the latest behavior time in the attention class behaviors as the recommendation score of the content related to the attention class behaviors.
The present invention also provides a content recommendation apparatus, comprising:
a user representation acquisition unit for acquiring a user representation of a user; the user representation is determined based on user attribute information of the user or based on user attribute characteristics and historical behavior characteristics of the user;
an expert screening unit for screening relevant experts related to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert;
the content pushing unit is used for determining the content to be recommended based on the recent interaction behavior of the relevant experts and pushing the content to be recommended to the user;
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
The present invention also provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and operable on the processor, wherein the processor implements the steps of any of the content recommendation methods described above when executing the program.
The invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the content recommendation method as described in any of the above.
According to the content recommendation method, the content recommendation device, the electronic equipment and the storage medium, the user representation of the user is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user, the problems that the number of new user interaction behaviors is small and modeling is difficult are solved, relevant experts are selected from the expert database according to the user representation, the content to be recommended is determined according to the recent interaction behaviors of the relevant experts, and the content recommendation efficiency is improved; in addition, through comprehensively considering the interaction behavior of each mature user to each content and the content score of each content, the experts are screened out, the representativeness of the experts in the content community is improved, the experts are selected by the method for content recommendation, the accuracy and the quality of content recommendation can be improved, and the method is more suitable for content recommendation of the content community.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a flow chart of a content recommendation method provided by the present invention;
FIG. 2 is a detailed diagram of a content recommendation method provided by the present invention;
FIG. 3 is a schematic structural diagram of a content recommendation apparatus provided in the present invention;
fig. 4 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a content recommendation method according to an embodiment of the present invention, as shown in fig. 1, the method includes:
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
Specifically, a user needing content recommendation is determined, and a user representation of the user is obtained. The user representation of the user is used for representing user portrait information of the user, and when certain interactive behavior information exists in the community, the user representation can also represent interest tendency information of the user on community contents. Here, considering that the user may be a new user who has just registered or a regression user who has rarely accessed the community in recent time, and at this time, the user may lack an interactive behavior, so that the user representation of the user may be determined by using a graph model based on the user attribute characteristics of the user by using the characteristic that the interests and hobbies of users with similar attributes may be similar, and the problems that the interactive behavior of the new user is few and the modeling is difficult are effectively solved. The user attribute features are text information containing personal information of the user, such as age group, gender, school, academic calendar, city, login equipment system, login equipment model and the like, which can be obtained by obtaining registration information of the user in the community database and personal information displayed on other platform (such as social platform and the like) pages associated with the user account.
After the user generates interactive behaviors in the community, the user attribute characteristics and the historical behavior characteristics of the user can be combined, and the interest tendency implied by the user representation and the interest tendency indicated by the actual behaviors of the user are integrated to model the user, so that the user representation of the user is obtained. The historical behavior features are statistical texts containing the interaction behaviors of the user in the community in the recent period of time, such as user praise times, collection times, sharing times, author concerned times, answer click rate, article click rate and the like, and can be obtained according to user behavior statistics recorded in the click log table data. The frequency of updating the user representation may be determined based on the registration duration of the user within the community. When the user registers for 7 days, the number of the historical behavior characteristics is 0 or less, as long as the user logs in the community, the user representation of the user can be calculated once in real time for content recommendation, and after the user registers for more than 7 days, the updating frequency of the user representation can be reduced, for example, the user representation can be updated regularly once every week.
Subsequently, considering that the reference value of the interest of most common users in the community is not high when making content recommendation, in order to reduce the calculation amount of similar users and improve the efficiency of content recommendation, relevant experts related to the user can be screened from the expert database based on the user representation of the user and the expert representations of the experts in the expert database. Wherein the expert representation of any expert is determined based on the user attribute characteristics and historical behavior characteristics of the expert. The user attribute characteristics and the historical behavior characteristics of the expert respectively comprise personal information of the expert and statistical characteristics of the interaction behaviors in the community in the recent period of time. Relevant experts similar to the user may be screened from the expert repository by calculating similarities between the user representation of the user and expert representations of all experts. If the user is a new user or the historical behavior features are less, relevant experts, namely experts with similar attributes to the user, can deduce possible interest tendency of the user when the user lacks interactive behaviors by screening out the relevant experts.
Here, the experts in the expert database are bright light users having a certain authority in a certain field in the community, and the behaviors thereof should be more representative than those of most general users. In order to extract experts with more representative behaviors from a large number of users in a community, the embodiment of the invention comprehensively considers the interactive behaviors of each mature user to each content and the content scores of each content, and screens out the mature users which have the content which generates the interactive behaviors recently and is high-quality content as the experts. The content score of each content is determined based on the interaction index corresponding to each content, and the higher the interaction index of a certain content is, the higher the degree of attention and approval of the content is, the better the content is. It should be noted that, unlike the existing method of determining the expert user only according to the user's own attributes (e.g., liveness, influence, etc.), in the embodiment of the present invention, the content score of the interactive content is fully utilized when determining the expert user, so that a mature user whose content that has recently generated an interactive behavior is a high-quality content is selected as an expert, and the representativeness of the expert in the content community is improved.
Then, content voting can be performed according to the screened latest interaction behaviors of the relevant experts, and the content which has the latest access and the most access is selected as the content to be recommended, so that the content to be recommended is pushed to the user.
The method provided by the embodiment of the invention determines the user representation of the user based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user, solves the problems of few new user interaction behaviors and difficult modeling, selects relevant experts from an expert database according to the user representation, determines the content to be recommended according to the recent interaction behaviors of the relevant experts, and improves the content recommendation efficiency; in addition, through comprehensively considering the interaction behavior of each mature user to each content and the content score of each content, the experts are screened out, the representativeness of the experts in the content community is improved, the experts are selected by the method for content recommendation, the accuracy and the quality of content recommendation can be improved, and the method is more suitable for content recommendation of the content community.
Based on any of the above embodiments, the historical behavior characteristics of the user or the expert are determined based on the following steps:
acquiring historical interaction behaviors of the user or the expert;
determining the content corresponding to the historical interactive behaviors and the popularity of the content corresponding to the historical interactive behaviors;
randomly discarding the content with the popularity higher than a preset threshold value in the content corresponding to the historical interaction behavior;
and counting the behaviors of the residual contents based on the user or the expert to obtain the historical behavior characteristics.
Specifically, some explosive head contents are ignored in the current recommendation algorithm, although most users click to view the contents, the interests of the users are not the same, nor are the users really interested in similar contents, so that the explosive contents are improperly recommended to the users who are not interested, or two users who view the same explosive contents are mistakenly considered to be similar, and the contents accessed by one user are pushed to the other user, thereby resulting in poor recommendation effect.
In contrast, when modeling is performed on a user or an expert, adverse effects possibly caused by the content of the exploding money are planed, so that the accuracy of content recommendation is improved. Specifically, historical interactive behaviors of a user or an expert can be obtained, and content corresponding to the historical interactive behaviors and popularity of the content corresponding to the historical interactive behaviors can be determined. The higher the popularity of the content, that is, the greater the access traffic, the more likely it is that the exploded content is viewed and accessed by most users. For content with a popularity higher than a preset threshold, i.e. for explosive content, it may be discarded to avoid adverse effects when acquiring a user representation or an expert representation, resulting in a misspeculation of the user's or expert's interest, or misjudging the similarity between the user and the expert.
After the explosive content is discarded, the behavior of the remaining content can be counted based on the user or the expert to obtain corresponding historical behavior characteristics.
Based on any of the above embodiments, determining the popularity of the content corresponding to the historical interaction behavior specifically includes:
acquiring the interaction amount and/or ranking information of the content corresponding to the historical interaction behavior in a preset time period;
determining popularity of the content corresponding to the historical interaction behaviors based on the interaction amount and/or ranking information of the content corresponding to the historical interaction behaviors in a preset time period; the popularity of any content is higher as the interaction amount and/or ranking information of the content in a preset time period is higher.
Specifically, the amount of interaction and/or ranking information of the content corresponding to the historical interaction behavior of the user or the expert in a preset time period (for example, about 180 days) may be obtained. The mutual quantity may be a reading quantity, a collection quantity, a forwarding quantity, a comment quantity, and the like of the corresponding content, and the ranking information may be a ranking condition of the corresponding content in the whole community or a ranking condition of a specific field, and the like.
And determining the popularity of the content corresponding to the historical interactive behaviors according to the interactive amount and/or ranking information of the content corresponding to the historical interactive behaviors in a preset time period. Wherein, the popularity of any content is in proportion to the amount of interaction and/or ranking information of the content in a preset time period.
In any of the above embodiments, the expert is determined based on the following steps:
calculating a content score of each content based on the interaction index of each content; the interaction index comprises at least one of a praise rate, a collection rate, a sharing rate, a thank you rate and a comment rate;
determining the latest interactive content corresponding to each mature user based on the interactive behavior of each mature user in the latest preset time period;
and screening out experts from each mature user based on the content score of the corresponding latest interactive content of each mature user.
Specifically, the content score of each content may be calculated based on the interaction index of each content within the community. The interaction indexes comprise one or more of the like rate, the collection rate, the sharing rate, the thank you rate and the comment rate. And carrying out weighted summation on the values of the interactive indexes according to the weights corresponding to the interactive indexes of any content, so as to calculate the content score of the content.
In addition, based on the interaction behavior of each mature user in the last preset time period (for example, last 90 days), for example, like, collection, thank you, comment, creation, etc., the content related to the interaction behavior can be used as the last interaction content corresponding to each mature user.
And screening out experts from each mature user based on the content score of the corresponding latest interactive content of each mature user. The content score of the latest interactive content corresponding to any mature user can be obtained, and the average value of the content scores of the latest interactive content is calculated. And screening a plurality of mature users with the highest average value as experts according to the average value of the content scores of the latest interactive contents corresponding to the mature users.
Based on any of the above embodiments, the screening of experts from the respective mature users based on the content scores of the latest interactive content corresponding to the respective mature users specifically includes:
determining a user score of any mature user based on a content score of a most recent interactive content corresponding to the mature user and the number of users who pay attention to the mature user and are of a high-quality user type;
and screening out experts from the various mature users based on the user scores of the various mature users.
Specifically, for any mature user, a user list concerning the mature user can be obtained, and the number of users in the user list which are of a high-quality user type is screened out. Wherein, the user whose content is operated for more than 3 times and VIP user can be regarded as the high-quality user type. In addition, weights may be set for the average of the content scores of the latest interactive content corresponding to the mature user and the number of users who pay attention to the mature user and are of high-quality user types, and the user scores of the mature user may be obtained by weighted addition of the weights. Based on the user scores of the respective mature users, the mature users having the user scores higher than a preset score may be acquired as experts.
Based on any of the above embodiments, step 130 specifically includes:
acquiring the content related to the latest interactive behavior of the related experts; the recent interaction behaviors comprise click-type behaviors and attention-type behaviors;
determining recommendation scores of the contents related to the click-type behaviors based on click rates of the contents related to the click-type behaviors in the latest interactive behaviors in all experts in the expert database;
determining recommendation scores of contents related to the attention class behaviors based on behavior time of the attention class behaviors in the recent interactive behaviors;
and determining the content to be recommended based on the recommendation score of the content associated with the clicking type behavior and the recommendation score of the content associated with the attention type behavior.
Specifically, the latest interaction behavior of the relevant experts and the content related to the latest interaction behavior are obtained. The recent interactive behaviors include click-type behaviors, that is, behaviors in which a user clicks to view a certain content, and attention-type behaviors, such as praise, collection, and sharing. Wherein, the creation time of the content associated with the click-like behavior in the latest interactive behavior should be within a preset time period (for example, 120 days). For the click-through behavior in the recent interactive behavior, the click rate of the content associated with the behavior in each expert in the expert database can be obtained as the recommendation score of the content associated with the click-through behavior. For an attention class behavior in the recent interactive behaviors, the behavior time of the class behavior can be acquired as the recommendation score of the content associated with the attention class behavior. And after calculating the recommendation scores of the contents related to all the click behaviors and the recommendation scores of the contents related to all the attention behaviors, obtaining the recommendation score of each content related to the latest interactive behaviors. After the recommendation scores of each content related to the latest interaction behavior of each relevant expert are sorted, the content of the TopN can be pushed as the content to be recommended.
Based on any of the above embodiments, the determining, based on the behavior time of the attention class behavior in the recent interaction behaviors, the recommendation score of the content associated with the attention class behavior specifically includes:
and if the content related to the attention class behaviors corresponds to a plurality of attention class behaviors of the relevant experts, taking the latest behavior time in the attention class behaviors as the recommendation score of the content related to the attention class behaviors.
Specifically, if a relevant expert performs multiple attention class behaviors on content associated with a certain attention class behavior, for example, praise, collect, and share the same article, the latest behavior time of the multiple attention class behaviors is taken as the recommendation score of the content associated with the attention class behavior.
Based on any of the above embodiments, fig. 2 is a detailed schematic diagram of a content recommendation method provided by an embodiment of the present invention, as shown in fig. 2, the method includes:
calculating the content score of each content in the community, calculating the user score of each user by combining the behavior of each user on the content, selecting the user with the score of more than or equal to 0.7 as an expert to form an expert database, wherein the number of the experts is about 40 ten thousand. And generating expert representations of the experts by associating the experts with the AI embedding offline table, and establishing an online zann index for the expert representation of each expert. Here, the expert and the expert representation may be determined by the expert determination method and the expert representation determination method given in the above embodiments, and details thereof are not described herein. In addition, the content involved in the most recent interactive behavior of each expert is recorded in the redis index. The expert list in the expert database and the expert representation in the zann index can be updated periodically every week, the content related to the latest interactive behavior in the redis can be updated by monitoring the kafka message in real time, when the expert generates a new behavior, and the data expiration time can be set to be 30 days when the new behavior is written in the redis index, and the behavior data can be automatically deleted after 30 days.
The redis index is used for storing the latest interactive behaviors of experts and related contents thereof, wherein the followUser correspondingly writes in praise, collection and sharing behaviors of the user, the creation time of the corresponding contents is not limited, the behavior time is the most recommended score of the corresponding contents, and the maximum number of behavior lists is 300. The userClick corresponds to the click behavior written in the user, the creation time of the corresponding content is within 120 days, the length of the behavior list is not limited, and the click rate of the corresponding content in all the special persons is obtained according to the recommendation score of the corresponding content.
And when the online recall is carried out, generating a user representation of the user through the user attribute characteristics and the historical behavior characteristics in the community. The user indicates that the user directly calls the online embedding interface to obtain when the user is used on the user line, and compared with the traditional mode that the user embedding is generated once every day, the method is more friendly to the newly registered user, and the user coverage rate is improved. And then, calculating 1200 experts with the highest similarity from the zann, acquiring 100 latest interaction behaviors from the redis index aiming at each screened expert, calculating the recommendation score of the content related to the latest interaction behaviors, and taking the content with the recommendation score TopN as the content to be recommended to obtain the recommendation result of the final recall queue.
Each functional module in the diagram (i.e. each rectangular module in fig. 2) is independently responsible for its own execution, so that it is convenient to quickly locate the problem when a fault occurs, and it is also beneficial to expand the function and optimize the algorithm. Compared with the traditional userCF recall, the recall queue can generate user vectors after users are created, is more friendly to newly created users, and simultaneously uses an expert user online voting mechanism to better accord with the preference of the requesting users for the content recommended to the users.
The content recommendation device provided by the present invention is described below, and the content recommendation device described below and the content recommendation device method described above may be referred to in correspondence with each other.
Based on any of the above embodiments, fig. 3 is a schematic structural diagram of a content recommendation device according to an embodiment of the present invention, and as shown in fig. 3, the device includes: the user represents the acquisition unit 310, the expert filtering unit 320, and the content push unit 330.
Wherein, the user representation acquiring unit 310 is configured to acquire a user representation of a user; the user representation is determined based on user attribute information of the user or based on user attribute characteristics and historical behavior characteristics of the user;
the expert screening unit 320 is used for screening the relevant experts related to the user from the expert database based on the user representation of the user and the expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert;
the content pushing unit 330 is configured to determine content to be recommended based on the recent interaction behavior of the relevant experts, and push the content to be recommended to the user;
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
The device provided by the embodiment of the invention determines the user representation of the user based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user, solves the problems of few new user interaction behaviors and difficulty in modeling, selects relevant experts from the expert database according to the user representation, determines the content to be recommended according to the recent interaction behaviors of the relevant experts, and improves the content recommendation efficiency; in addition, through comprehensively considering the interaction behavior of each mature user to each content and the content score of each content, the experts are screened out, the representativeness of the experts in the content community is improved, the experts are selected by the method for content recommendation, the accuracy and the quality of content recommendation can be improved, and the method is more suitable for content recommendation of the content community.
Based on any of the above embodiments, the historical behavior characteristics of the user or the expert are determined based on the following steps:
acquiring historical interaction behaviors of the user or the expert;
determining the content corresponding to the historical interactive behaviors and the popularity of the content corresponding to the historical interactive behaviors;
randomly discarding the content with the popularity higher than a preset threshold value in the content corresponding to the historical interaction behavior;
and counting the behaviors of the residual contents based on the user or the expert to obtain the historical behavior characteristics.
Based on any of the above embodiments, determining the popularity of the content corresponding to the historical interaction behavior specifically includes:
acquiring the interaction amount and/or ranking information of the content corresponding to the historical interaction behavior in a preset time period;
determining popularity of the content corresponding to the historical interaction behaviors based on the interaction amount and/or ranking information of the content corresponding to the historical interaction behaviors in a preset time period; the popularity of any content is higher as the interaction amount and/or ranking information of the content in a preset time period is higher.
In any of the above embodiments, the expert is determined based on the following steps:
calculating a content score of each content based on the interaction index of each content; the interaction index comprises at least one of a praise rate, a collection rate, a sharing rate, a thank you rate and a comment rate;
determining the latest interactive content corresponding to each mature user based on the interactive behavior of each mature user in the latest preset time period;
and screening out experts from each mature user based on the content score of the corresponding latest interactive content of each mature user.
Based on any of the above embodiments, the screening of experts from the respective mature users based on the content scores of the latest interactive content corresponding to the respective mature users specifically includes:
determining a user score of any mature user based on a content score of a most recent interactive content corresponding to the mature user and the number of users who pay attention to the mature user and are of a high-quality user type;
and screening out experts from the various mature users based on the user scores of the various mature users.
Based on any one of the above embodiments, the determining the content to be recommended based on the recent interaction behavior of the relevant expert specifically includes:
acquiring the content related to the latest interactive behavior of the related experts; the recent interaction behaviors comprise click-type behaviors and attention-type behaviors;
determining recommendation scores of the contents related to the click-type behaviors based on click rates of the contents related to the click-type behaviors in the latest interactive behaviors in all experts in the expert database;
determining recommendation scores of contents related to the attention class behaviors based on behavior time of the attention class behaviors in the recent interactive behaviors;
and determining the content to be recommended based on the recommendation score of the content associated with the clicking type behavior and the recommendation score of the content associated with the attention type behavior.
Based on any of the above embodiments, the determining, based on the behavior time of the attention class behavior in the recent interaction behaviors, the recommendation score of the content associated with the attention class behavior specifically includes:
and if the content related to the attention class behaviors corresponds to a plurality of attention class behaviors of the relevant experts, taking the latest behavior time in the attention class behaviors as the recommendation score of the content related to the attention class behaviors.
Fig. 4 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 4: a processor (processor)410, a communication Interface 420, a memory (memory)430 and a communication bus 440, wherein the processor 410, the communication Interface 420 and the memory 430 are communicated with each other via the communication bus 440. Processor 410 may invoke logic instructions in memory 430 to perform a content recommendation method comprising: acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user; screening relevant experts relevant to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert; determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user; the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
In addition, the logic instructions in the memory 430 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product including a computer program, the computer program being storable on a non-transitory computer-readable storage medium, the computer program being capable of executing, when executed by a processor, a content recommendation method provided by the above methods, the method including: acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user; screening relevant experts relevant to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert; determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user; the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which when executed by a processor, implements a method for content recommendation provided by the above methods, the method comprising: acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user; screening relevant experts relevant to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert; determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user; the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A content recommendation method, comprising:
acquiring a user representation of a user; the user representation is determined based on the user attribute characteristics of the user or the user attribute characteristics and the historical behavior characteristics of the user;
screening relevant experts relevant to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert;
determining the content to be recommended based on the recent interaction behavior of the relevant experts, and pushing the content to be recommended to the user;
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
2. The content recommendation method according to claim 1, wherein the historical behavior characteristics of the user or the expert are determined based on the steps of:
acquiring historical interaction behaviors of the user or the expert;
determining the content corresponding to the historical interactive behaviors and the popularity of the content corresponding to the historical interactive behaviors;
randomly discarding the content with the popularity higher than a preset threshold value in the content corresponding to the historical interaction behavior;
and counting the behaviors of the residual contents based on the user or the expert to obtain the historical behavior characteristics.
3. The content recommendation method according to claim 2, wherein determining the popularity of the content corresponding to the historical interaction behavior specifically comprises:
acquiring the interaction amount and/or ranking information of the content corresponding to the historical interaction behavior in a preset time period;
determining popularity of the content corresponding to the historical interaction behaviors based on the interaction amount and/or ranking information of the content corresponding to the historical interaction behaviors in a preset time period; the popularity of any content is higher as the interaction amount and/or ranking information of the content in a preset time period is higher.
4. The content recommendation method according to claim 1, wherein said expert is determined based on the steps of:
calculating a content score of each content based on the interaction index of each content; the interaction index comprises at least one of a praise rate, a collection rate, a sharing rate, a thank you rate and a comment rate;
determining the latest interactive content corresponding to each mature user based on the interactive behavior of each mature user in the latest preset time period;
and screening out experts from each mature user based on the content score of the corresponding latest interactive content of each mature user.
5. The content recommendation method according to claim 4, wherein the screening of experts from the mature users based on the content scores of the most recent interactive content corresponding to the mature users comprises:
determining a user score of any mature user based on a content score of a most recent interactive content corresponding to the mature user and the number of users who pay attention to the mature user and are of a high-quality user type;
and screening out experts from the various mature users based on the user scores of the various mature users.
6. The content recommendation method according to claim 1, wherein the determining the content to be recommended based on the recent interactive behavior of the relevant expert specifically comprises:
acquiring the content related to the latest interactive behavior of the related experts; the recent interaction behaviors comprise click-type behaviors and attention-type behaviors;
determining recommendation scores of the contents related to the click-type behaviors based on click rates of the contents related to the click-type behaviors in the latest interactive behaviors in all experts in the expert database;
determining recommendation scores of contents related to the attention class behaviors based on behavior time of the attention class behaviors in the recent interactive behaviors;
and determining the content to be recommended based on the recommendation score of the content associated with the clicking type behavior and the recommendation score of the content associated with the attention type behavior.
7. The content recommendation method according to claim 6, wherein the determining a recommendation score of the content associated with the attention class behavior based on the behavior time of the attention class behavior in the recent interactive behaviors specifically comprises:
and if the content related to the attention class behaviors corresponds to a plurality of attention class behaviors of the relevant experts, taking the latest behavior time in the attention class behaviors as the recommendation score of the content related to the attention class behaviors.
8. A content recommendation apparatus characterized by comprising:
a user representation acquisition unit for acquiring a user representation of a user; the user representation is determined based on user attribute information of the user or based on user attribute characteristics and historical behavior characteristics of the user;
an expert screening unit for screening relevant experts related to the user from an expert database based on the user representation of the user and expert representations of the experts in the expert database; the expert representation is determined based on the user attribute characteristics and the historical behavior characteristics of the expert;
the content pushing unit is used for determining the content to be recommended based on the recent interaction behavior of the relevant experts and pushing the content to be recommended to the user;
the expert is determined based on the interaction behavior of each mature user to each content and the content score of each content, and the content score of each content is determined based on the interaction index corresponding to each content.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the content recommendation method according to any of claims 1 to 7 are implemented when the processor executes the program.
10. A non-transitory computer readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the content recommendation method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210024223.6A CN114048390A (en) | 2022-01-11 | 2022-01-11 | Content recommendation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210024223.6A CN114048390A (en) | 2022-01-11 | 2022-01-11 | Content recommendation method and device, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114048390A true CN114048390A (en) | 2022-02-15 |
Family
ID=80213668
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210024223.6A Pending CN114048390A (en) | 2022-01-11 | 2022-01-11 | Content recommendation method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114048390A (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140344861A1 (en) * | 2013-05-14 | 2014-11-20 | Tivo Inc. | Method and system for trending media programs for a user |
CN112784608A (en) * | 2021-02-24 | 2021-05-11 | 科大讯飞股份有限公司 | Test question recommendation method and device, electronic equipment and storage medium |
CN113158033A (en) * | 2021-03-19 | 2021-07-23 | 浙江工业大学 | Collaborative recommendation model construction method based on knowledge graph preference propagation |
-
2022
- 2022-01-11 CN CN202210024223.6A patent/CN114048390A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140344861A1 (en) * | 2013-05-14 | 2014-11-20 | Tivo Inc. | Method and system for trending media programs for a user |
CN112784608A (en) * | 2021-02-24 | 2021-05-11 | 科大讯飞股份有限公司 | Test question recommendation method and device, electronic equipment and storage medium |
CN113158033A (en) * | 2021-03-19 | 2021-07-23 | 浙江工业大学 | Collaborative recommendation model construction method based on knowledge graph preference propagation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108334575B (en) | Recommendation result sorting correction method and device and electronic equipment | |
CN103377250B (en) | Top k based on neighborhood recommend method | |
CN110442790A (en) | Recommend method, apparatus, server and the storage medium of multi-medium data | |
CN108256119A (en) | A kind of construction method of resource recommendation model and the resource recommendation method based on the model | |
US11216494B2 (en) | Virtual artificial intelligence based consultant | |
CN112612942B (en) | Social big data-based fund recommendation system and method | |
CN111510783B (en) | Method, device, electronic equipment and storage medium for determining video exposure | |
CN106921795A (en) | A kind of contact data management method and its system | |
CN113435627A (en) | Work order track information-based electric power customer complaint prediction method and device | |
CN107633257B (en) | Data quality evaluation method and device, computer readable storage medium and terminal | |
CN111581486A (en) | Information distribution method and device | |
WO2019242453A1 (en) | Information processing method and device, storage medium, and electronic device | |
CN111160699A (en) | Expert recommendation method and system | |
CN111104422A (en) | Training method, device, equipment and storage medium of data recommendation model | |
CN114048390A (en) | Content recommendation method and device, electronic equipment and storage medium | |
CN113254788B (en) | Big data based recommendation method and system and readable storage medium | |
CN111309774A (en) | Data processing method and device, electronic equipment and storage medium | |
CN116091133A (en) | Target object attribute identification method, device and storage medium | |
Noble et al. | Conditional language models for community-level linguistic variation | |
CN111127059B (en) | User quality analysis method and device | |
CN113469819A (en) | Recommendation method of fund product, related device and computer storage medium | |
CN112734463A (en) | Service information sending method and device, electronic equipment and storage medium | |
CN114722266A (en) | Questionnaire pushing method and device, electronic equipment and readable storage medium | |
CN110619547A (en) | Method and device for determining cyber red commercial tenant, electronic equipment and storage medium | |
Hees et al. | BetterRelations: Detailed Evaluation of a Game to Rate Linked Data Triples |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220215 |