A personalized content recommendation cold start method, system, device and storage medium

Abstract

The embodiment of the present invention discloses a personalized content recommendation cold start method, system, device and storage medium. The method performs a quality assessment on each content in the database to obtain an assessment result, then selects the target content from the database, and then uses the content identifier of the target content and the corresponding assessment result to generate a basic content queue. The number of behaviors corresponding to the label of each user's related content is counted from the database to obtain the user's behavior weight score and personalized ratio. Based on the behavior weight score and personalized ratio, elements are selected from the basic content queue and re-sorted and combined to generate a content recommendation queue. Finally, the content data in the content recommendation queue is recommended to the user. The embodiment of the present invention achieves the recommendation of high-quality and diversified content for new users, avoids user loss due to poor quality of recommended content or recommended content that users are not interested in, expands the types of recommended content, and effectively improves the quality of recommended content.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of recommendation systems, and in particular to a personalized content recommendation cold start method, system, device, and storage medium. Background Art

[0002] Most existing content recommendation systems rely on users' historical behavior. For new users without historical behavior, content recommendation cannot be performed based on historical behavior. In this case, a cold start method is needed to perform content recommendation.

[0003] Existing conventional cold start methods will recommend popular content to new users, but the types of popular content are limited by platform attributes. For example, the popular content on a recipe platform is almost all recipes, making it difficult to expand the content scope.

[0004] Existing popular content generally only considers the number of times users browse it or the proportion of effective browsing, which cannot fully reflect the quality of the content and the degree of popularity among platform users.

[0005] Since new content released by the platform has low views and no historical user evaluations of this content, if new content is randomly recommended to users, the quality of the content and the degree of user preference for the new content cannot be guaranteed, which may cause harm to users and lead to user loss. Summary of the Invention

[0006] To this end, embodiments of the present invention provide a personalized content recommendation cold start method, system, device and storage medium to solve the problem that existing content recommendation relies on user historical behavior and cannot recommend representative personalized content for new users or newly released content.

[0007] In order to achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0008] According to a first aspect of an embodiment of the present invention, a personalized content recommendation cold start method is provided, the method comprising:

[0009] Perform quality assessment on each content in the database and obtain assessment results;

[0010] Selecting target content from the database, and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result;

[0011] Counting the number of actions of each user with a tag corresponding to the relevant content from the database to obtain the user's action weight score and personalized ratio;

[0012] Selecting elements from the basic content queue and reordering and combining them according to the behavior weight score and the personalization ratio to generate a content recommendation queue;

[0013] Recommending content data in the content recommendation queue to users.

[0014] Furthermore, the quality of each content in the database is evaluated to obtain the evaluation results, including:

[0015] According to a first preset period, using the database, querying the number of each related behavior of each content;

[0016] Performing a division operation by the number of related behaviors of the content and the total number of exposures to obtain the relevant behavior rate of each content;

[0017] performing a weighted operation on each of the related behavior rates according to a first preset weight of the related behavior rate to obtain each weighted behavior rate of the content;

[0018] The weighted behavior rates are summed to obtain an evaluation result for each content.

[0019] Furthermore, target content is selected from the database, and a basic content queue is generated using the content identifier of the target content and the corresponding evaluation result, including:

[0020] Initializing the basic content queue according to a second preset period;

[0021] Defining a first preset number of first content queues according to platform user attributes, each of the first content queues corresponding to at least one first content tag;

[0022] For each first content tag in the database, define a corresponding tag content set and a single extraction quantity of elements corresponding to each tag content set;

[0023] Query the database for content under all content tags;

[0024] Determining whether the number of views of the content is greater than a preset number of views;

[0025] If the page views of the content are greater than a preset page views, the content is used as the first content;

[0026] According to the first content tag corresponding to the first content, storing the content identifier of the first content and the corresponding evaluation result as elements in a corresponding tag content set;

[0027] Reordering the elements in the tag content set according to the evaluation results of the corresponding contents;

[0028] For the first content tag corresponding to the first content queue, extract the corresponding single extraction quantity of elements from the head of the corresponding sorted tag content set, and place them into the first intermediate queue;

[0029] Deleting the extracted elements from the sorted tag content set to obtain an updated tag content set;

[0030] Reorder the elements in the first intermediate queue according to the evaluation results of the corresponding contents;

[0031] Determine whether there is a sorted tag content set without element extraction;

[0032] If there is a sorted tag content set that has not been subjected to element extraction, then extract elements from the next sorted tag content set;

[0033] If all sorted tag content sets have been subjected to element extraction, the data in the sorted first intermediate queue is placed into the second intermediate queue, and the first intermediate queue is cleared;

[0034] Determine whether the number of elements in the current second intermediate queue reaches a second preset number;

[0035] If the number of elements in the current second intermediate queue does not reach the second preset number, then re-extracting the corresponding single extraction number of elements from the head of each updated tag content set;

[0036] If the number of elements in the current second intermediate queue reaches a second preset number, the second intermediate queue is used as the first content queue.

[0037] Furthermore, target content is selected from the database, and a basic content queue is generated using the content identifier of the target content and the corresponding evaluation result, further comprising:

[0038] Using each content tag in the database as a second content tag, defining a corresponding second content queue;

[0039] If the page views of the content are less than or equal to the preset page views, the content is used as the second content;

[0040] storing the content identifier and the evaluation result of the second content in a corresponding second content queue according to the second content tag corresponding to the second content;

[0041] Utilize new content in the database as tertiary content;

[0042] According to the content tag of the third content, storing the content identifier and the evaluation result of the third content into the tail of the corresponding second content queue;

[0043] The basic content queue is obtained by using the first content queue and / or the second content queue.

[0044] Furthermore, the number of behaviors of each user with the corresponding tag of the relevant content is counted from the database to obtain the behavior weight score and personalized ratio of the user, including:

[0045] According to a third preset period, using the database, querying the number of user-related content-related behavior tags;

[0046] Using the second preset weight, a weighted sum operation is performed on the number of behaviors of the tags corresponding to the relevant content to obtain the behavior weight score corresponding to each tag of the user;

[0047] Sum the behavior weight scores corresponding to each tag of the user to obtain the total behavior weight score of the user;

[0048] A personalized ratio of the user is obtained according to the total behavior weight score, wherein the personalized ratio is less than a preset ratio.

[0049] Furthermore, according to the behavior weight score and the personalized ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, including:

[0050] Receive content requests initiated by users;

[0051] Determining whether the content request is the first request of the user within a fourth preset period;

[0052] If the content request is the first request of the user within a fourth preset period, initializing an element extraction position of the first content queue of the user;

[0053] If the content request is not the first request of the user within the fourth preset period, the element extraction position of the first content queue of the user is not initialized;

[0054] Obtaining a corresponding personalized value according to the personalized ratio of the user, wherein the personalized value is less than a preset value;

[0055] According to the personalized value of the user, a corresponding number of elements to be extracted is obtained;

[0056] Extract elements of the element extraction quantity backward from the element extraction position of the first content queue, put them into the third intermediate queue, and update the element extraction position using the element extraction quantity.

[0057] Furthermore, according to the behavior weight score and the personalized ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, which also includes:

[0058] For the second content tag in the database, determining whether the behavior weight score corresponding to the second content tag of the user is greater than a preset weight score;

[0059] If the behavior weight score corresponding to the second content tag is greater than the preset weight score, taking a third preset number of content identifiers and evaluation results from the head of the corresponding second content queue and placing them into a third intermediate queue;

[0060] If the behavior weight score corresponding to the second content tag is less than or equal to the preset weight score, the content identifier and the evaluation result are not extracted from the second content queue and are placed in the third intermediate queue;

[0061] The elements in the third intermediate queue are reordered according to the evaluation results of the corresponding contents to obtain a content recommendation queue.

[0062] According to a second aspect of an embodiment of the present invention, a personalized content recommendation cold start system is provided, the system comprising:

[0063] The content quality assessment module is used to perform quality assessment on each content in the database and obtain assessment results;

[0064] A basic content queue generation module is used to select target content from the database and generate a basic content queue using the content identifier of the target content and the corresponding evaluation result;

[0065] A personalized building module is used to count the number of behaviors of each user with the corresponding tag of the relevant content from the database to obtain the behavior weight score and personalized ratio of the user;

[0066] a content recommendation queue generation module, configured to select elements from the basic content queue and re-order and combine them according to the behavior weight score and the personalization ratio to generate a content recommendation queue;

[0067] The content recommendation module is used to recommend the content data in the content recommendation queue to the user.

[0068] Furthermore, the quality of each content in the database is evaluated to obtain the evaluation results, including:

[0069] According to a first preset period, using the database, querying the number of each related behavior of each content;

[0070] Performing a division operation by the number of related behaviors of the content and the total number of exposures to obtain the relevant behavior rate of each content;

[0071] performing a weighted operation on each of the related behavior rates according to a first preset weight of the related behavior rate to obtain each weighted behavior rate of the content;

[0072] The weighted behavior rates are summed to obtain an evaluation result for each content.

[0073] Furthermore, target content is selected from the database, and a basic content queue is generated using the content identifier of the target content and the corresponding evaluation result, including:

[0074] Initializing the basic content queue according to a second preset period;

[0075] Defining a first preset number of first content queues according to platform user attributes, each of the first content queues corresponding to at least one first content tag;

[0076] For each first content tag in the database, define a corresponding tag content set and a single extraction quantity of elements corresponding to each tag content set;

[0077] Query the database for content under all content tags;

[0078] Determining whether the number of views of the content is greater than a preset number of views;

[0079] If the page views of the content are greater than a preset page views, the content is used as the first content;

[0080] According to the first content tag corresponding to the first content, storing the content identifier of the first content and the corresponding evaluation result as elements in a corresponding tag content set;

[0081] Reordering the elements in the tag content set according to the evaluation results of the corresponding contents;

[0082] For the first content tag corresponding to the first content queue, extract the corresponding single extraction quantity of elements from the head of the corresponding sorted tag content set, and place them into the first intermediate queue;

[0083] Deleting the extracted elements from the sorted tag content set to obtain an updated tag content set;

[0084] Reorder the elements in the first intermediate queue according to the evaluation results of the corresponding contents;

[0085] Determine whether there is a sorted tag content set without element extraction;

[0086] If there is a sorted tag content set that has not been subjected to element extraction, then extract elements from the next sorted tag content set;

[0087] If all sorted tag content sets have been subjected to element extraction, the data in the sorted first intermediate queue is placed into the second intermediate queue, and the first intermediate queue is cleared;

[0088] Determine whether the number of elements in the current second intermediate queue reaches a second preset number;

[0089] If the number of elements in the current second intermediate queue does not reach the second preset number, then re-extracting the corresponding single extraction number of elements from the head of each updated tag content set;

[0090] If the number of elements in the current second intermediate queue reaches a second preset number, the second intermediate queue is used as the first content queue.

[0091] Furthermore, target content is selected from the database, and a basic content queue is generated using the content identifier of the target content and the corresponding evaluation result, further comprising:

[0092] Using each content tag in the database as a second content tag, defining a corresponding second content queue;

[0093] If the page views of the content are less than or equal to the preset page views, the content is used as the second content;

[0094] storing the content identifier and the evaluation result of the second content in a corresponding second content queue according to the second content tag corresponding to the second content;

[0095] Utilize new content in the database as tertiary content;

[0096] According to the content tag of the third content, storing the content identifier and the evaluation result of the third content into the tail of the corresponding second content queue;

[0097] The basic content queue is obtained by using the first content queue and / or the second content queue.

[0098] Furthermore, the number of behaviors of each user with the corresponding tag of the relevant content is counted from the database to obtain the behavior weight score and personalized ratio of the user, including:

[0099] According to a third preset period, using the database, querying the number of user-related content corresponding to the tag;

[0100] Using the second preset weight, a weighted sum operation is performed on the number of behaviors of the tags corresponding to the relevant content to obtain the behavior weight score corresponding to each tag of the user;

[0101] Sum the behavior weight scores corresponding to each tag of the user to obtain the total behavior weight score of the user;

[0102] A personalized ratio of the user is obtained according to the total behavior weight score, wherein the personalized ratio is less than a preset ratio.

[0103] Furthermore, according to the behavior weight score and the personalized ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, including:

[0104] Receive content requests initiated by users;

[0105] Determining whether the content request is the first request of the user within a fourth preset period;

[0106] If the content request is the first request of the user within a fourth preset period, initializing an element extraction position of the first content queue of the user;

[0107] If the content request is not the first request of the user within the fourth preset period, the element extraction position of the first content queue of the user is not initialized;

[0108] Obtaining a corresponding personalized value according to the personalized ratio of the user, wherein the personalized value is less than a preset value;

[0109] According to the personalized value of the user, a corresponding number of elements to be extracted is obtained;

[0110] Extract elements of the element extraction quantity backward from the element extraction position of the first content queue, put them into the third intermediate queue, and update the element extraction position using the element extraction quantity.

[0111] Furthermore, according to the behavior weight score and the personalized ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, which also includes:

[0112] For the second content tag in the database, determining whether the behavior weight score corresponding to the second content tag of the user is greater than a preset weight score;

[0113] If the behavior weight score corresponding to the second content tag is greater than the preset weight score, taking a third preset number of content identifiers and evaluation results from the head of the corresponding second content queue and placing them into a third intermediate queue;

[0114] If the behavior weight score corresponding to the second content tag is less than or equal to the preset weight score, the content identifier and the evaluation result are not extracted from the second content queue and are placed in the third intermediate queue;

[0115] The elements in the third intermediate queue are reordered according to the evaluation results of the corresponding contents to obtain a content recommendation queue.

[0116] According to a third aspect of an embodiment of the present invention, there is provided a personalized content recommendation cold start device, the device comprising: a processor and a memory;

[0117] The memory is used to store one or more program instructions;

[0118] The processor is configured to run one or more program instructions to execute the steps of any one of the above described personalized content recommendation cold start methods.

[0119] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a personalized content recommendation cold start method as described in any one of the above items are implemented.

[0120] The embodiments of the present invention have the following advantages:

[0121] The embodiment of the present invention discloses a personalized content recommendation cold start method, system, device and storage medium. The method performs a quality assessment on each content in the database to obtain an assessment result, then selects the target content from the database, and then uses the content identifier of the target content and the corresponding assessment result to generate a basic content queue. The number of behaviors corresponding to the label of each user's related content is counted from the database to obtain the user's behavior weight score and personalized ratio. Based on the behavior weight score and personalized ratio, elements are selected from the basic content queue and re-sorted and combined to generate a content recommendation queue. Finally, the content data in the content recommendation queue is recommended to the user. The embodiment of the present invention achieves the recommendation of high-quality and diversified content for new users, avoids user loss due to poor quality of recommended content or recommended content that users are not interested in, expands the types of recommended content, and effectively improves the quality of recommended content. BRIEF DESCRIPTION OF THE DRAWINGS

[0122] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0123] The structures, proportions, sizes, etc. illustrated in this specification are intended only to complement the contents disclosed herein and to facilitate understanding and reading by persons familiar with the art. They are not intended to limit the conditions under which the present invention may be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportions, or adjustments in sizes, without affecting the efficacy and objectives of the present invention, shall still fall within the scope of the technical contents disclosed herein.

[0124] Figure 1 A schematic diagram of the logical structure of a personalized content recommendation cold start system provided by an embodiment of the present invention;

[0125] Figure 2 A flowchart of a cold-start method for personalized content recommendation provided by an embodiment of the present invention;

[0126] Figure 3 A schematic diagram of a process for performing quality assessment on each content in a database provided by an embodiment of the present invention;

[0127] Figure 4 A schematic diagram of a process for generating a basic content queue according to an embodiment of the present invention;

[0128] Figure 5 A schematic diagram of a process for obtaining a user's behavior weight score and personalized ratio according to an embodiment of the present invention;

[0129] Figure 6 A schematic diagram of a process for generating a content recommendation queue according to an embodiment of the present invention. DETAILED DESCRIPTION

[0130] The following describes the implementation of the present invention using specific embodiments. Those skilled in the art will readily understand the other advantages and benefits of the present invention from the disclosure herein. Obviously, the embodiments described are only a portion of the present invention, not all of it. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0131] refer to Figure 1 The embodiment of the present invention provides a personalized content recommendation cold start system, which specifically includes: a content quality assessment module 1, a basic content queue generation module 2, a personalized construction module 3, a content recommendation queue generation module 4 and a content recommendation module 5.

[0132] Furthermore, the content quality assessment module 1 is used to perform quality assessment on each content in the database to obtain an assessment result; the basic content queue generation module 2 is used to select target content from the database, and generate a basic content queue using the content identifier of the target content and the corresponding assessment result; the personalized construction module 3 is used to count the number of behaviors of the label corresponding to each user-related content from the database to obtain the user's behavior weight score and personalized ratio; the content recommendation queue generation module 4 is used to select elements from the basic content queue and re-sort and combine them according to the behavior weight score and personalized ratio to generate a content recommendation queue; the content recommendation module 5 is used to recommend content data in the content recommendation queue to the user.

[0133] The embodiment of the present invention discloses a personalized content recommendation cold start system. The system performs a quality assessment on each content in the database to obtain an assessment result. The system then selects target content from the database. The system then uses the content identifier of the target content and the corresponding assessment result to generate a basic content queue. The system then counts the number of behaviors corresponding to the label of each user's relevant content from the database to obtain the user's behavior weight score and personalized ratio. Based on the behavior weight score and personalized ratio, the system selects elements from the basic content queue and re-sorts and combines them to generate a content recommendation queue. Finally, the content data in the content recommendation queue is recommended to the user. The embodiment of the present invention achieves the recommendation of high-quality and diversified content for new users, avoids user loss due to poor quality of recommended content or recommended content that users are not interested in, expands the types of recommended content, and effectively improves the quality of recommended content.

[0134] Corresponding to the personalized content recommendation cold start system disclosed above, an embodiment of the present invention further discloses a personalized content recommendation cold start method. The following describes in detail the personalized content recommendation cold start method disclosed in an embodiment of the present invention in conjunction with the personalized content recommendation cold start system described above.

[0135] refer to Figure 2 The following describes the specific steps of a personalized content recommendation cold start method provided by an embodiment of the present invention.

[0136] The content quality assessment module 1 performs quality assessment on each content in the database to obtain an assessment result.

[0137] refer to Figure 3The above steps specifically include: according to a first preset period, querying various related behavior times of the content through the database, the above related behavior times include: effective browsing times, comment times, like times, collection times and sharing times; dividing each related behavior time of the content by the total exposure times of the content to obtain each related behavior rate of the content, the related behavior rate includes: effective browsing rate, comment rate, like rate, collection rate and sharing rate; setting a first preset weight for the above related behavior rate, and then weighting the related behavior rate to obtain each weighted behavior rate of the content, accumulating each weighted behavior rate of the content to obtain an evaluation result of the content; and storing the obtained evaluation result in the database.

[0138] The embodiment of the present invention uses quality assessment to comprehensively evaluate users' effective browsing, comments, likes, collections, and sharing of content, rather than just considering the number of views, so that the evaluation results can more reasonably measure the quality of the content.

[0139] The basic content queue generation module 2 selects target content from the database and generates a basic content queue using the content identifier of the target content and the corresponding evaluation result.

[0140] refer to Figure 4, the above steps specifically include: initializing the existing basic content queue according to the second preset period and clearing the data in the basic content queue; defining a first preset number of first content queues according to the platform user attributes, each first content queue corresponding to at least one first content tag; for each first content tag in the database, defining a corresponding tag content set and the number of elements to be extracted at a time corresponding to each tag content set; using each content tag in the database as a second content tag, defining a corresponding second content queue; querying the content under all content tags from the database; judging whether the number of views of each content is greater than the preset number of views; if the number of views of the content is greater than the preset number of views, treating the content as the first content, and according to the first content tag corresponding to the first content, storing the content identifier and the corresponding evaluation result of the first content as elements in the corresponding tag content set; sorting the elements in the tag content set in descending order according to the evaluation results of the corresponding content; for the first content tag corresponding to the first content queue, extracting the content from the corresponding sorted tag content Extract the corresponding number of elements extracted at a time from the head of the content set and put them into the first intermediate queue; delete the extracted elements from the sorted label content set to obtain an updated label content set; re-sort the elements in the first intermediate queue according to the evaluation results of the corresponding content; determine whether there is a sorted label content set that has not been extracted; if there is a sorted label content set that has not been extracted, extract elements from the sorted label content set that has not been extracted; if all sorted label content sets have been extracted, put the data in the sorted first intermediate queue into the second intermediate queue, and clear the sorted first intermediate queue; determine whether the number of elements in the current second intermediate queue has reached the second preset number; if the number of elements in the current second intermediate queue has not reached the second preset number, re-extract the corresponding number of elements extracted at a time from the head of the updated label content set; if the number of elements in the current second intermediate queue has reached the second preset number, use the second intermediate queue as the first content queue.

[0141] If the number of views of the content is less than or equal to the preset number of views, the content will be used as the second content; according to the second content tag corresponding to the second content, the content identifier and evaluation result of the second content will be stored in the corresponding second content queue; at the same time, the new content in the database will be used as the third content in real time; according to the content tag of the third content, the content identifier and evaluation result of the third content will be stored at the end of the corresponding second content queue; and the first content queue and / or the second content queue will be used to generate a basic content queue.

[0142] The basic content queue generated by the embodiment of the present invention includes a first content queue and / or a second content queue. Each first content queue corresponds to a type of content tag. By constructing a first preset number of first content queues, the content surface is expanded, allowing users to try more diverse content types.

[0143] The personalized construction module 3 counts the number of behaviors of each user with the corresponding tag of the relevant content from the database to obtain the user's behavior weight score and personalized ratio.

[0144] refer to Figure 5 The above steps specifically include: according to the third preset period, querying the number of behaviors of the tags corresponding to the user-related content from the database; setting a second preset weight for various behaviors of the user; using the second preset weight to perform a weighted sum operation on the number of behaviors of the tags corresponding to the user-related content to obtain the behavior weight score corresponding to each tag of the user; accumulating the behavior weight score corresponding to each tag of the user to obtain the total behavior weight score of the user; defining the personalized ratio corresponding to different total behavior weight score segments, the personalized ratio and the total behavior weight score have a positive proportional relationship, and the personalized ratio of the user is obtained according to the total behavior weight score; setting the upper limit of the above personalized ratio to the preset ratio, and the personalized ratio is less than the preset ratio.

[0145] The content recommendation queue generation module 4 selects elements from the basic content queue and re-sorts and combines them according to the behavior weight score and the personalization ratio to generate a content recommendation queue.

[0146] refer to Figure 6 , the above steps specifically include: receiving a content request sent by a user; determining whether the content request is the first request of the user within a fourth preset period; if the content request is the first request of the user within the fourth preset period, initializing the element extraction position of the user's first content queue; if the content request is not the first request of the user within the fourth preset period, not initializing the element extraction position of the user's first content queue; storing the element extraction position in the KV storage space of the corresponding redis database to facilitate fast access to data; obtaining a corresponding personalized value according to the user's personalized ratio, wherein the personalized value is less than a preset value; obtaining a corresponding element extraction quantity according to the user's personalized value, and the element extraction quantity is inversely proportional to the personalized value; extracting elements of the element extraction quantity from the element extraction position of the first content queue, placing them into the third intermediate queue, and updating the element extraction position using the element extraction quantity.

[0147] For each second content tag in the database, determine whether the behavior weight score corresponding to the user's second content tag is greater than the preset weight score; if the behavior weight score corresponding to the user's second content tag is greater than the preset weight score, extract a third preset number of content identifiers and evaluation results from the head of the corresponding second content queue and place them in a third intermediate queue; if the behavior weight score corresponding to the user's second content tag is less than or equal to the preset weight score, do not extract the content identifiers and evaluation results from the second content queue and place them in the third intermediate queue; sort the elements in the third intermediate queue obtained after the above process in descending order according to the corresponding evaluation results to obtain a content recommendation queue.

[0148] The following example illustrates the above steps:

[0149] The first content queues D1, D2, D3, and the second content queue D4 have been constructed. After receiving a content request initiated by a user, the user's personalization ratio is used to obtain the user's personalization value x, which is less than 8. If the content request is the user's first request that day, the element extraction position i[D][user] in the user's first content queue is initialized to i[D][user] = 0. Based on the user's personalization value x, the corresponding element extraction number y is obtained. The relationship between the personalization value x and the element extraction number y is shown in Table 1:

[0150] condition y[D1] y[D2] y[D3] x<2 9-x 0 0 2<＝x<4 4 6-x 0 4<＝x<8 1 1 8-x

[0151] Table 1: Correspondence between personalized values ​​and number of elements extracted

[0152] Take the corresponding elements from the i[D1][user]th position of the first content queues D1, D2, and D3, respectively, and extract a number y of elements into the third intermediate queue, setting i[D1][user] = i[D1][user] + y; for each second content tag in the database, determine whether there is a user whose second content tag corresponds to a behavior weight score greater than a preset weight score; if the behavior weight score of a certain second content tag of the user exceeds the preset weight score, take a content identifier and the corresponding evaluation result from the head of the second content queue D4 corresponding to the second content tag and put them into the third intermediate queue, and set i[D4][user] = i[D4][user] + 1; sort the elements in the third intermediate queue obtained after the above steps in descending order according to the corresponding evaluation results to obtain a content recommendation queue.

[0153] The content recommendation module 5 recommends the content data in the content recommendation queue to the user.

[0154] The above steps specifically include: according to the content identifiers in the content recommendation queue, returning the content corresponding to the content identifiers in the database to the client in order, and recommending them to the user who sent the content request.

[0155] The embodiment of the present invention discloses a personalized content recommendation cold start method, which performs a quality assessment on each content in the database to obtain an assessment result, then selects the target content from the database, and then uses the content identifier of the target content and the corresponding assessment result to generate a basic content queue. The number of behaviors corresponding to the label of each user's related content is counted from the database to obtain the user's behavior weight score and personalized ratio. Based on the behavior weight score and personalized ratio, elements are selected from the basic content queue and re-sorted and combined to generate a content recommendation queue. Finally, the content data in the content recommendation queue is recommended to the user. The embodiment of the present invention achieves the recommendation of high-quality and diversified content for new users, avoids user loss due to poor quality of recommended content or recommended content that users are not interested in, expands the types of recommended content, and effectively improves the quality of recommended content.

[0156] In addition, an embodiment of the present invention also provides a personalized content recommendation cold start device, which includes: a processor and a memory; the memory is used to store one or more program instructions; the processor is used to run one or more program instructions to execute the steps of a personalized content recommendation cold start method as described in any of the above items.

[0157] In addition, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the personalized content recommendation cold start method as described in any one of the above items are implemented.

[0158] In the embodiments of the present invention, the processor may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0159] The methods, steps, and logic diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads the information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned methods.

[0160] The storage medium may be a memory and may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0161] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0162] Volatile memory may be random access memory (RAM), which is used as an external cache memory. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM).

[0163] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0164] Those skilled in the art will appreciate that in one or more of the above examples, the functions described herein can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media includes any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0165] Although the present invention has been described in detail above using general descriptions and specific embodiments, it will be apparent to those skilled in the art that modifications and improvements may be made thereto. Therefore, such modifications and improvements, without departing from the spirit of the present invention, are intended to be within the scope of protection claimed herein.

Claims

1. A cold start method for personalized content recommendation, characterized in that: The method comprises: Perform quality assessment on each content in the database and obtain assessment results; Selecting target content from the database, and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result; Counting the number of actions of each user with a tag corresponding to the relevant content from the database to obtain the user's action weight score and personalized ratio; Selecting elements from the basic content queue and reordering and combining them according to the behavior weight score and the personalization ratio to generate a content recommendation queue; recommending content data in the content recommendation queue to the user; Selecting target content from the database and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result includes: Initializing the basic content queue according to a second preset period; Defining a first preset number of first content queues according to platform user attributes, each of the first content queues corresponding to at least one first content tag; For each first content tag in the database, define a corresponding tag content set and a single extraction quantity of elements corresponding to each tag content set; Query the database for content under all content tags; Determining whether the number of views of the content is greater than a preset number of views; If the page views of the content are greater than a preset page views, the content is used as the first content; According to the first content tag corresponding to the first content, storing the content identifier of the first content and the corresponding evaluation result as elements in a corresponding tag content set; Reordering the elements in the tag content set according to the evaluation results of the corresponding contents; For the first content tag corresponding to the first content queue, extract the corresponding single extraction quantity of elements from the head of the corresponding sorted tag content set, and place them into the first intermediate queue; Deleting the extracted elements from the sorted tag content set to obtain an updated tag content set; Reorder the elements in the first intermediate queue according to the evaluation results of the corresponding contents; Determine whether there is a sorted tag content set without element extraction; If there is a sorted tag content set that has not been subjected to element extraction, then extract elements from the next sorted tag content set; If all sorted tag content sets have been subjected to element extraction, the data in the sorted first intermediate queue is placed into the second intermediate queue, and the first intermediate queue is cleared; Determine whether the number of elements in the current second intermediate queue reaches a second preset number; If the number of elements in the current second intermediate queue does not reach the second preset number, then re-extracting the corresponding single extraction number of elements from the head of each updated tag content set; If the number of elements in the second intermediate queue currently reaches a second preset number, using the second intermediate queue as the first content queue; Selecting target content from the database, and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result, further comprising: Using each content tag in the database as a second content tag, defining a corresponding second content queue; If the page views of the content are less than or equal to the preset page views, the content is used as the second content; storing the content identifier and the evaluation result of the second content in a corresponding second content queue according to the second content tag corresponding to the second content; The basic content queue is obtained by using the first content queue and / or the second content queue.

2. The personalized content recommendation cold start method according to claim 1, characterized in that: Perform quality assessment on each content in the database and obtain assessment results, including: According to a first preset period, using the database, querying the number of each related behavior of each content; Performing a division operation by the number of related behaviors of the content and the total number of exposures to obtain the relevant behavior rate of each content; performing a weighted operation on each of the related behavior rates according to a first preset weight of the related behavior rate to obtain each weighted behavior rate of the content; The weighted behavior rates are summed to obtain an evaluation result for each content.

3. The personalized content recommendation cold start method according to claim 1, characterized in that: Counting the number of actions of each user with the corresponding tag in the relevant content from the database to obtain the user's action weight score and personalized ratio includes: According to a third preset period, using the database, querying the number of user-related content-related behavior tags; Using the second preset weight, a weighted sum operation is performed on the number of behaviors of the tags corresponding to the relevant content to obtain the behavior weight score corresponding to each tag of the user; Sum the behavior weight scores corresponding to each tag of the user to obtain the total behavior weight score of the user; A personalized ratio of the user is obtained according to the total behavior weight score, wherein the personalized ratio is less than a preset ratio.

4. The personalized content recommendation cold start method according to claim 3, characterized in that: According to the behavior weight score and the personalization ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, including: Receive content requests initiated by users; Determining whether the content request is the first request of the user within a fourth preset period; If the content request is the first request of the user within a fourth preset period, initializing an element extraction position of the first content queue of the user; If the content request is not the first request of the user within the fourth preset period, the element extraction position of the first content queue of the user is not initialized; Obtaining a corresponding personalized value according to the personalized ratio of the user, wherein the personalized value is less than a preset value; According to the personalized value of the user, a corresponding number of elements to be extracted is obtained; Extract elements of the element extraction quantity backward from the element extraction position of the first content queue, put them into the third intermediate queue, and update the element extraction position using the element extraction quantity.

5. The personalized content recommendation cold start method according to claim 4, characterized in that: According to the behavior weight score and the personalized ratio, elements are selected from the basic content queue and re-ordered and combined to generate a content recommendation queue, further comprising: For the second content tag in the database, determining whether the behavior weight score corresponding to the second content tag of the user is greater than a preset weight score; If the behavior weight score corresponding to the second content tag is greater than the preset weight score, taking a third preset number of content identifiers and evaluation results from the head of the corresponding second content queue and placing them into a third intermediate queue; If the behavior weight score corresponding to the second content tag is less than or equal to the preset weight score, the content identifier and the evaluation result are not extracted from the second content queue and are placed in the third intermediate queue; The elements in the third intermediate queue are reordered according to the evaluation results of the corresponding contents to obtain a content recommendation queue.

6. A personalized content recommendation cold start system, characterized by: The system comprises: The content quality assessment module is used to perform quality assessment on each content in the database and obtain assessment results; A basic content queue generation module is used to select target content from the database and generate a basic content queue using the content identifier of the target content and the corresponding evaluation result; A personalized building module is used to count the number of behaviors of each user with the corresponding tag in the relevant content from the database to obtain the behavior weight score and personalized ratio of the user; a content recommendation queue generation module, configured to select elements from the basic content queue and re-order and combine them according to the behavior weight score and the personalization ratio to generate a content recommendation queue; A content recommendation module, configured to recommend content data in the content recommendation queue to users; Selecting target content from the database and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result includes: Initializing the basic content queue according to a second preset period; Defining a first preset number of first content queues according to platform user attributes, each of the first content queues corresponding to at least one first content tag; For each first content tag in the database, define a corresponding tag content set and a single extraction quantity of elements corresponding to each tag content set; Query the database for content under all content tags; Determining whether the number of views of the content is greater than a preset number of views; If the page views of the content are greater than a preset page views, the content is used as the first content; According to the first content tag corresponding to the first content, storing the content identifier of the first content and the corresponding evaluation result as elements in a corresponding tag content set; Reordering the elements in the tag content set according to the evaluation results of the corresponding contents; For the first content tag corresponding to the first content queue, extract the corresponding single extraction quantity of elements from the head of the corresponding sorted tag content set, and place them into the first intermediate queue; Deleting the extracted elements from the sorted tag content set to obtain an updated tag content set; Reorder the elements in the first intermediate queue according to the evaluation results of the corresponding contents; Determine whether there is a sorted tag content set without element extraction; If there is a sorted tag content set that has not been subjected to element extraction, then extract elements from the next sorted tag content set; If all sorted tag content sets have been subjected to element extraction, the data in the sorted first intermediate queue is placed into the second intermediate queue, and the first intermediate queue is cleared; Determine whether the number of elements in the current second intermediate queue reaches a second preset number; If the number of elements in the current second intermediate queue does not reach the second preset number, then re-extracting the corresponding single extraction number of elements from the head of each updated tag content set; If the number of elements in the second intermediate queue currently reaches a second preset number, using the second intermediate queue as the first content queue; Selecting target content from the database, and generating a basic content queue using the content identifier of the target content and the corresponding evaluation result, further comprising: Using each content tag in the database as a second content tag, defining a corresponding second content queue; If the page views of the content are less than or equal to the preset page views, the content is used as the second content; storing the content identifier and the evaluation result of the second content in a corresponding second content queue according to the second content tag corresponding to the second content; The basic content queue is obtained by using the first content queue and / or the second content queue.

7. A personalized content recommendation cold start device, characterized in that: The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to execute the steps of the personalized content recommendation cold start method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the personalized content recommendation cold start method according to any one of claims 1 to 5.

Images