Method for determining content of information to be recommended

Abstract

The present application relates to a method, device, computer equipment and storage medium for determining information content to be recommended. The method comprises: obtaining a first predetermined number of first information content objects that are most recent; for each of the first information content objects, obtaining the click rate score of each of the first information content objects within the most recent first predetermined time; performing missing value smoothing on each of the click rate scores so that information content objects without click rate scores also obtain corresponding click rate scores, and obtaining click rate scores after missing value smoothing; determining the information content objects to be recommended based on the click rate scores after missing value smoothing on each of the first information content objects. The use of this method can improve the quality of information content recommendation.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer device and storage medium for determining information content to be recommended. Background Art

[0002] With the development of information technology, the recommendation of objects related to information content, such as user-generated content (UGC), has become an important part of network technology applications. When recommending information, the common method is to recommend related information based on the user's click, watch or read related operation information, but the current information content recommendation method has the problem of poor recommendation quality. Summary of the invention

[0003] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for determining the content of information to be recommended that can improve the quality of advancement in response to the above technical problems.

[0004] A method for determining content of information to be recommended, the method comprising:

[0005] Acquire a first predetermined number of first information content objects that are most recent in time;

[0006] For each of the first information content objects, obtaining a click rate score of each of the first information content objects within a recent first predetermined time period;

[0007] Performing missing value smoothing processing on each of the click rate scores so that the information content objects without click rate scores also obtain corresponding click rate scores, and obtaining click rate scores after the missing value smoothing processing;

[0008] The information content object to be recommended is determined based on the click rate score after the missing value smoothing processing of each of the first information content objects.

[0009] Based on the scheme of the embodiment described above, it obtains the click-through rate scores within the most recent first predetermined time for the most recent first predetermined number of first information content objects, and performs missing value processing on this basis, so as to combine the click-through rate of a certain number of information content objects that have been recently released, and make the first information content objects without click-through rate scores have certain click-through rate scores and have the probability of being recommended, thereby solving the cold start problem of newly released information content due to the lack of historical data, being able to effectively recommend new content so that it can enter the recommendation process, and improving the quality of information content recommendation.

[0010] A method for determining content of information to be recommended, the method comprising:

[0011] Acquire a second information content object within the most recent second predetermined time;

[0012] Performing frequency statistics on the second information content object to obtain a first frequency of each first participle in the second information content object;

[0013] Acquire a second frequency of each of the first participles in the information content object within a recent third predetermined time, wherein the third predetermined time is greater than the second predetermined time;

[0014] Determining a content hotspot score of each of the second information content objects based on a first frequency and a second frequency of each of the first participles contained in the second information content object;

[0015] Based on the content hotspot score of each of the second information content objects, determine the information content object to be recommended in the second information content objects.

[0016] Based on the scheme of the embodiment described above, for each second information content object within the second scheduled time that is most recent, by counting the first frequency of the first participle contained therein within the second scheduled time and the second frequency within the third scheduled time, a content hotspot score of each second information content object is calculated, thereby being able to discover hot information content objects from the second information content objects released within the second scheduled time that is most recent, and being able to discover hot content in newly released information content, thereby improving the quality of information content recommendation.

[0017] A method for determining content of information to be recommended, the method comprising:

[0018] Acquire a first predetermined number of first information content objects that are most recent in time, and a second information content object that is most recent in a second predetermined time;

[0019] For a first predetermined number of the first information content objects, obtaining a click rate score of each of the first information content objects within a recent first predetermined time; and performing missing value smoothing processing on each of the click rate scores to obtain a click rate score of each of the first information content objects after the missing value smoothing processing;

[0020] Perform frequency statistics on the second information content object to obtain a first frequency of each first participle in the second information content object within the second predetermined time and a second frequency within a latest third predetermined time; determine a content hotspot score of each second information content object based on the first frequency and the second frequency of each first participle contained in the second information content object, wherein the third predetermined time is greater than the second predetermined time;

[0021] The information content object to be recommended is determined based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects.

[0022] Based on the scheme of the embodiment described above, for the first predetermined number of first information content objects that are most recent, the click-through rate scores within the most recent first predetermined time are obtained, and missing value processing is performed on this basis, so that the click-through rate of a certain number of information content objects that are recently released can be combined, and the first information content objects without click-through rate scores can also have a certain click-through rate score and can also have a probability of being recommended, thereby solving the cold start problem of the newly released information content due to the lack of historical data. At the same time, for each second information content object within the most recent second predetermined time, by counting the first frequency of the first participle contained therein within the second predetermined time and the second frequency within the third predetermined time, the content hotspot score of each second information content object is calculated, so that hot information content objects can be discovered from the second information content objects that are released within the most recent second predetermined time, and hot content in the newly released information content can be discovered, so that on the basis of solving the cold start problem and being able to effectively recommend new content so that it can enter the recommendation process, real-time hotspots can be discovered, thereby improving the overall recommendation quality.

[0023] In one embodiment, after obtaining the most recent first predetermined number of first information content objects, and before obtaining the click rate score of each of the first information content objects within the most recent first predetermined time, it also includes: removing the first information content objects containing sensitive words from the first predetermined number of first information content objects.

[0024] In one embodiment, after obtaining the most recent first predetermined number of first information content objects, and before obtaining the click rate score of each of the first information content objects within the most recent first predetermined time, it also includes: removing the first information content objects whose content length is less than a predetermined length threshold among the first predetermined number of first information content objects.

[0025] In one embodiment, before obtaining the click rate score of each of the first information content objects within the most recent first predetermined time, the method further includes:

[0026] Each of the first information content objects is written into a time-reversed sequence.

[0027] In one embodiment, obtaining the click rate score of each of the first information content objects within the most recent first predetermined time includes:

[0028] Obtaining the number of clicks and exposures of each of the first information content objects within a recent first predetermined time period;

[0029] Based on the number of clicks and the number of exposures, a click rate score of each of the first information content objects within a most recent first predetermined time is calculated.

[0030] In one embodiment, the CTR score is confidence-smoothed to obtain the CTR score after confidence-smoothing, including: confidence-smoothing the CTR score according to the exposure, wherein the greater the exposure, the smaller the penalty for the corresponding CTR score.

[0031] In one embodiment, performing frequency statistics on the second information content object to obtain a first frequency of each first word in the second information content object includes:

[0032] A unigram language model and a bigram language model are used to perform frequency statistics on each of the second information content objects to obtain a first frequency of each first word in the second information content object.

[0033] In one embodiment, the method further comprises the steps of: obtaining a third information content object within the latest third predetermined time;

[0034] Perform frequency statistics on the third information content object to obtain a second frequency of each second participle in the third information content object, each second participle including the first participle.

[0035] In one embodiment, obtaining the second frequency of each of the first participles in the information content object within the latest third predetermined time includes:

[0036] The unigram language model and the bigram language model are used to perform frequency statistics on the contents of each third information object within a recent third predetermined time to obtain a second frequency of each second participle in the third information content object, wherein each second participle includes the first participle.

[0037] In one embodiment, determining the content hotspot score of each of the second information content objects based on the word hotspot score of each of the first words contained in the second information content object includes:

[0038] The word hotspot scores of the first word segments contained in the second information content object are summed to obtain the content hotspot score of the information content object.

[0039] In one embodiment, determining the first information content object to be recommended based on the recommendation probability value of each first information content object after the normalization process includes:

[0040] Determining a recommendation probability interval of each of the first information content objects based on the recommendation probability value of each of the first information content objects;

[0041] A randomly selected first probability value is obtained, and a first information content object corresponding to a recommendation probability interval corresponding to the first probability value is determined as the first information content object to be recommended.

[0042] In one embodiment, determining the second information content object to be recommended based on the recommendation probability value of each second information content object after the normalization process includes:

[0043] Determining a recommendation probability interval of each of the second information content objects based on the recommendation probability value of each of the second information content objects;

[0044] A randomly selected second probability value is obtained, and a second information content object corresponding to a recommendation probability interval corresponding to the second probability value is determined as the second information content object to be recommended.

[0045] In one embodiment, determining the information content object to be recommended based on the recommendation probability value of each second information content object after the normalization process includes:

[0046] Determining a recommendation probability interval of each of the second information content objects based on the recommendation probability value of each of the second information content objects;

[0047] A randomly selected probability value is obtained, and a second information content object corresponding to a recommendation probability interval corresponding to the probability value is determined as the information content to be recommended.

[0048] A device for determining content of information to be recommended, comprising:

[0049] A first object acquisition module, used to acquire a first predetermined number of first information content objects that are most recent in time;

[0050] A click rate score determination module, configured to obtain, for each of the first information content objects, a click rate score of each of the first information content objects within a recent first predetermined time period;

[0051] A missing value smoothing processing module is used to perform missing value smoothing processing on each of the click rate scores, so that the information content objects without click rate scores also obtain the corresponding click rate scores, and obtain the click rate scores after the missing value smoothing processing;

[0052] The first to-be-recommended object determination module is used to determine the to-be-recommended information content object based on the click-through rate score after the missing value smoothing processing of each of the first information content objects.

[0053] A device for determining content of information to be recommended, comprising:

[0054] A second object acquisition module, used to acquire a second information content object within a recent second predetermined time;

[0055] a frequency determination module, configured to perform frequency statistics on the second information content object to obtain a first frequency of each first participle in the second information content object; and obtain a second frequency of each first participle in the information content object within a recent third predetermined time, wherein the third predetermined time is greater than the second predetermined time;

[0056] a content hotspot score determination module, configured to determine a content hotspot score of each of the second information content objects based on a first frequency and a second frequency of each of the first participles contained in the second information content object;

[0057] The second to-be-recommended object determination module determines the to-be-recommended information content object in the second information content objects based on the content hotspot score of each of the second information content objects.

[0058] A device for determining content of information to be recommended, comprising:

[0059] An object acquisition module, which acquires a first predetermined number of first information content objects that are most recent, and a second information content object that is most recent within a second predetermined time;

[0060] A click score determination module is used to obtain, for a first predetermined number of the first information content objects, a click rate score of each of the first information content objects within a first predetermined time period; and to perform missing value smoothing processing on each of the click rate scores to obtain a click rate score of each of the first information content objects after the missing value smoothing processing;

[0061] a hotspot score determination module, configured to perform frequency statistics on the second information content object, obtain a first frequency of each first participle in the second information content object within the second predetermined time and a second frequency within a latest third predetermined time; determine a content hotspot score of each second information content object based on the first frequency and the second frequency of each first participle contained in the second information content object, wherein the third predetermined time is greater than the second predetermined time;

[0062] The recommended object determination module is used to determine the information content object to be recommended based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects.

[0063] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps included in any of the above methods when executing the computer program.

[0064] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps included in any of the above methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A diagram of an application environment of a method for determining information content to be recommended in an embodiment;

[0066] Figure 2 A schematic diagram of a flow chart of a method for determining content of information to be recommended in an embodiment;

[0067] Figure 3 is a flow chart of a method for determining content of information to be recommended in another embodiment;

[0068] Figure 4 is a flow chart of a method for determining content of information to be recommended in another embodiment;

[0069] FIG5 is a schematic diagram showing a comparison between the method of the embodiment of the present application and the traditional method in a specific example;

[0070] FIG6 is a schematic diagram showing a comparison between the method of the embodiment of the present application and the traditional method for recommendation in a specific example;

[0071] Figure 7 It is a structural block diagram of a device for determining content of information to be recommended in one embodiment;

[0072] Figure 8 It is a structural block diagram of a device for determining information content to be recommended in another embodiment;

[0073] Fig. 9 It is a structural block diagram of a device for determining information content to be recommended in another embodiment;

[0074] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0075] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0076] The method for determining the content of information to be recommended provided in this application can be applied to Figure 1In the application environment shown. In some embodiments, the application environment may only involve the terminal 102 and the server 104. The terminal 102 communicates with the server 104 through a network, and the terminal user can access or obtain the relevant information content published on the server 104 through the terminal 102, such as posts, articles, and other user-generated content. The server 104 recommends relevant information content to the terminal 102 when it visits, or actively recommends relevant information content to the terminal 102, for example, the recommended relevant information content can be displayed on the homepage provided to the terminal when it visits, or displayed on other relevant pages. The server 104 can determine the recommended information content to be recommended to the terminal 102 in combination with various strategies.

[0077] In some embodiments, the application environment may involve the terminal 102, the server 104 and the server 106 at the same time. The terminal 102 communicates with the server 104 through the network, and the terminal user can access or obtain the relevant information content published on the server 104 through the terminal 102. The server 104 communicates with the server 106 through the network, and the server 106 can determine the information content to be recommended based on the relevant information of the server 104, such as log data, and provide the determined information content to be recommended to the server 104. When the terminal 102 accesses the server 104, the server 104 determines the information content to be recommended to the terminal 102 in combination with the information content to be recommended provided by the server 106. The information content determined by the server 104 to be recommended to the terminal 102 can be directly the information content to be recommended provided by the server 106, or it can be combined with the information content to be recommended provided by the server 106 to perform relevant recommendation configuration, such as recommending the information to be recommended on multiple pages such as the homepage and other pages, or combining with the results of other recommendation strategies, etc. The server 104 and the server 106 can be independent servers, or they can refer to the same server. The terminal 102 may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0078] In one embodiment, Figure 2 As shown, a method for determining the content of information to be recommended is provided, and the method is applied to Figure 1 The server 104 or the server 106 is taken as an example to illustrate, including the following steps S201 to S204.

[0079] Step S201: Acquire a first predetermined number of first information content objects that are most recent in time.

[0080] The time here may refer to the release time of the information content object, such as the release time of the information content object of user-generated content such as posts and articles. When obtaining the first predetermined number of first information content objects that are most recently released, the first predetermined number of information content objects that are most recently released may be obtained in reverse order of time.

[0081] Among them, the first predetermined number can be set in combination with actual technical needs. In some embodiments, it can be determined in combination with the number of information content objects stored in each cycle, and can also be determined in combination with the current business situation. For example, assuming that the number of information content objects stored in each hour is N with a cycle of 1 hour, a value between 4N and 6N can be used as the first predetermined number. If the number of N is large, it can be appropriately reduced, for example, a value between 3N and 4N can be used as the first predetermined number. If the number of N is small, it can be appropriately increased, for example, a value between 5N and 7N can be used as the first predetermined number. It can be understood that in other embodiments, other methods can also be used to determine the first predetermined number.

[0082] In some embodiments, after obtaining the first predetermined number of first information content objects with the most recent time, the first information content objects containing sensitive words in the first predetermined number of first information content objects may be further removed. Sensitive words generally refer to terms that violate relevant laws and regulations and affect the health of the network environment. In some scenarios, they also involve words that are not suitable to appear on the relevant pages of the current server, etc. By removing the first information content objects containing sensitive words, the recommendation and dissemination of the first information content objects containing sensitive words can be avoided, which is conducive to optimizing the network environment.

[0083] In some embodiments, after obtaining the first predetermined number of first information content objects with the latest time, first information content objects with content lengths less than a predetermined length threshold may be further removed from the first predetermined number of first information content objects. For first information content objects with content lengths less than the predetermined length threshold, it means that they do not actually contain enough information. Therefore, removing these first information content objects can reduce the number of first information content objects involved in subsequent processing, which helps to improve processing efficiency.

[0084] It can be understood that in actual processing, after obtaining the first predetermined number of first information content objects with the most recent time, the first information content objects containing sensitive words and the first information content objects with a content length less than a predetermined length threshold can be removed from the first predetermined number of first information content objects at the same time.

[0085] After the above processing, the first information content objects finally obtained can be generated into a time-reversed list and written into a time-reversed sequence, thereby obtaining each first information content object in time-reversed order.

[0086] Step S202: for each of the first information content objects, obtain a click rate score of each of the first information content objects within a recent first predetermined time period.

[0087] Among them, it should be understood that when there are sensitive words or information content objects with content length less than a predetermined length threshold among the above-mentioned first preset number of first information content objects, after the above-mentioned removal operation, the number of each of the first information content objects involved in the time reversal will be less than the above-mentioned first predetermined number.

[0088] In one embodiment, obtaining the click rate score of each of the first information content objects within the most recent first predetermined time may specifically include step S2021 and step S2022.

[0089] Step S2021: Obtain the number of clicks and exposures of each of the first information content objects within the most recent first predetermined time.

[0090] The duration of the first scheduled time can be set in combination with actual technical needs. Generally, the first scheduled time should be set relatively short to achieve the recommendation of newly released information content objects, but the first scheduled time should also have a certain length of time to balance the clicks, likes, comments and other reading-related situations of the newly released information content objects within this period. For example, in one embodiment, the first scheduled time can be set to 5 minutes. It is understandable that in other embodiments, the first scheduled time can also be set to other lengths.

[0091] Among them, the number of clicks and exposures of each first information content object in the most recent first predetermined time can be obtained from relevant log data, such as the MTA log of the server, which records all log data related to the user's behavior of the information content object, such as the exposure action and exposure time of the relevant information of the information content object displayed on the relevant display page of the user, the click action and click time of the user clicking the information content object, the like action and like time of the user liking the information content object, and the comment action, comment content and comment time of the user commenting on the information content object. It can be understood that these recorded relevant information can also include the user ID of the corresponding user at the same time. Therefore, by accessing the MTA log, through the identifier of the information content object and the time of the user action (such as the click time), etc., the number of clicks and exposures of each information content object in the first predetermined time can be screened and statistically obtained.

[0092] Step S2022: Based on the number of clicks and the number of exposures, a click rate score of each of the first information content objects within a most recent first predetermined time is calculated.

[0093] In one embodiment of the present application, the click rate of the first information content object within the first predetermined time may be obtained by calculating the ratio of the above-mentioned number of clicks and the number of exposures, and then the corresponding click rate score within the most recent first predetermined time may be determined in combination with the click rate. In some embodiments, the obtained click rate may be directly used as the click rate score. In other embodiments, the click rate score may be obtained by further calculation based on the click rate. For example, assuming that the number of clicks of a first information content object D within the most recent first predetermined time T1 is d1, and the number of exposures is b1, then its corresponding click rate is d1 / b1. The click rate d1 / b1 may be directly used as the click rate score of D within T1, or d1 / b1 may be multiplied by a certain fixed value, such as 100, and 100d1 / b1 may be used as the click rate score of D within T1.

[0094] In one embodiment, after obtaining the click rate score as described above, the click rate score obtained may be further confidence smoothed to obtain the click rate score after confidence smoothing. Thus, by confidence smoothing the obtained click rate score, a smooth penalty process is performed so that the click rate score finally obtained can better conform to the actual click situation of the information content object. Among them, when performing the confidence smoothing process, the click rate score may be confidence smoothed according to the exposure amount. Specifically, the greater the exposure amount, the more confident the corresponding click rate score is and the smaller the penalty is. In some specific examples, the confidence smoothing process may be performed based on the Wilson confidence interval algorithm.

[0095] Step S203: performing missing value smoothing processing on each of the click rate scores, so that information content objects without click rate scores also obtain corresponding click rate scores, and obtaining click rate scores after the missing value smoothing processing.

[0096] Through the missing value smoothing process, the information content objects without click-through rate scores in each first information content object can also obtain the corresponding click-through rate scores, so that they can have the probability of being recommended and can enter the recommendation library for recommendation, so that the newly released information content objects without actual click-through rate can also enter the recommendation process, solving the cold start problem of the newly distributed content due to the lack of historical data. And by performing the missing value smoothing process after the confidence smoothing process, each first information content object can have a corresponding click-through rate score, and the click-through rate score is credible, which can ensure that the newly released information content objects can have the probability of being recommended and can enter the recommendation process.

[0097] Step S204: determining the information content object to be recommended based on the click rate score after the missing value smoothing processing of each of the first information content objects.

[0098] When determining the information content object to be recommended based on the click rate score after the missing value smoothing processing of each first information content object, the information content object to be recommended (also referred to as the first information content object to be recommended in the embodiment of the present application) can be determined directly based on the click rate score.

[0099] When determining the first information content object to be recommended based on the click rate score, a determined click rate score interval may be obtained, and the first information content object with a click rate score in the click rate score interval is determined as the first information content object to be recommended.

[0100] In one embodiment, when determining the information content object to be recommended based on the click rate score after the missing value smoothing processing of each of the first information content objects, the following method may be used, which may specifically include step S2041 and step S2042.

[0101] Step S2041: normalizing the click rate scores of each of the first information content objects after the missing value smoothing process to obtain the recommendation probability value of each of the first information content objects after the normalization process.

[0102] Through the normalization operation, each click rate score can be mapped to a decimal in the interval (0, 1). The specific normalization method can be performed using an existing normalization method.

[0103] Step S2042: Determine the first information content object to be recommended based on the recommendation probability value of each first information content object after the normalization process.

[0104] Based on the normalized recommendation probability values ​​of each first information content object, various possible methods may be used to determine the first information content object to be recommended. For example, by obtaining a determined recommendation probability interval, the first information content object whose recommendation probability value is within the recommendation probability interval is determined as the first information content object to be recommended.

[0105] In some embodiments, based on the recommendation probability values ​​of the normalized first information content objects, determining the first information content object to be recommended may be performed in the following manner.

[0106] First, based on the recommendation probability value of each first information content object, determine the recommendation probability interval of each first information content object. Thus, based on the recommendation probability value of each first information content object, each first information content object can be mapped to a corresponding recommendation probability interval. For example, assuming that there are 5 first information content objects D1-D5, and their recommendation probability values ​​are all 0.2, then the recommended probability intervals mapped to them can be (0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), [0.9, 1) respectively.

[0107] Then, a randomly selected first probability value is obtained, and a first information content object corresponding to a recommendation probability interval corresponding to the first probability value is determined as the first information content object to be recommended.

[0108] The first probability value randomly selected may be only one value, or may be obtained by multiple selections to obtain multiple first probability values, and then the first information content object corresponding to the recommended probability interval corresponding to the first probability value is used as the first information content object to be recommended. For example, assuming that the first probability values ​​randomly selected are 0.3 and 0.7, D1 and D4 corresponding to their corresponding intervals [0.2, 0.4) and [0.6, 0.8) are used as the first information content object to be recommended.

[0109] Based on the embodiments of the present application as described above, the click-through rate scores within the most recent first predetermined time are obtained for the most recent first predetermined number of first information content objects, and missing value processing is performed on this basis, so that the click-through rate of a certain number of information content objects that have been recently released can be combined, and the first information content objects that have no click-through rate scores can also have a certain click-through rate score and can also have a probability of being recommended, which solves the cold start problem of newly released information content due to the lack of historical data, can effectively recommend new content so that it can enter the recommendation process, and improves the quality of information content recommendation.

[0110] As shown in reference to FIG5 , by comparing the method for determining the content of information to be recommended in the above-mentioned embodiment with the minimum recommendation method provided by the business side in the traditional recommendation method and the recommendation method provided by the business side in a time-reversed manner, it is found by statistically analyzing the data within 10 days after adopting these three different methods that the method provided in the embodiment of the present application is superior to these two traditional methods in multiple indicators such as PV click-through rate, UV click-through rate, user like rate, user comment rate, and total number of exposed content items. The total number of exposed content items is greatly improved due to these two methods. On the basis of solving the problem of new content exposure, it can basically cover all new content that is released every day. Among them, the PV click-through rate is the ratio of the number of clicks to the number of exposures, the UV click-through rate is the ratio of the number of users who click to the number of exposures, the user like rate is the ratio of the number of users who like to the total number of users who have been exposed to the user, and the user comment rate is the ratio of the number of users who comment to the total number of users who have been exposed to the user.

[0111] In one embodiment, Figure 3 As shown, a method for determining the content of information to be recommended is provided, and the method is applied to Figure 1 Taking the server 104 or the server 106 in the example as an example, the method for determining the content of information to be recommended includes the following steps S301 to S305.

[0112] Step S301: Acquire a second information content object within the most recent second predetermined time.

[0113] The time here may refer to the publishing time of the information content object, such as the publishing time of the information content object of user-generated content such as posts and articles, that is, the publishing time of each second information content object is within the second predetermined time.

[0114] The duration of the second scheduled time can be set in combination with actual technical needs. Generally, the second scheduled time should be set relatively short to achieve the mining of hot content in the newly released information content object and to mine sudden hot events. However, the second scheduled time can also have a certain length of time to balance the newly released information content objects and the user's behavior such as clicks, likes, comments and other reading-related situations during this period. For example, in one embodiment, the second scheduled time can be set to 2 hours. It is understandable that in other embodiments, the second scheduled time can also be set to other durations.

[0115] Step S302: Perform frequency statistics on the second information content object to obtain the first frequency of each first word in the second information content object.

[0116] When performing frequency statistics on the second information content object, the frequency statistics of the word segmentation can be combined with the word segmentation model. In some embodiments, the unigram language model and the bigram language model can be used to perform frequency statistics on each of the second information content objects to obtain the first frequency of each first word segmentation in the second information content object. Among them, the unigram language model is used as a model of unigram word segmentation, and the word segmentation is performed by dividing the second information content object into individual words, and the frequency statistics are performed on this basis. The bigram language model is used as a model of binary word segmentation, and the word segmentation is performed by dividing every two words in the second information content object into one word, and the frequency statistics are performed on this basis. By combining the unigram language model and the bigram language model for frequency statistics, the problem of reduced overall accuracy caused by word segmentation errors can be avoided, and the effect attenuation caused by the absence of word segmentation can be avoided, so that the final frequency statistics can cover more words, so that the coverage of the word segmentation is wider, which helps to improve the accuracy of the final result.

[0117] Step S303: obtaining the second frequency of each of the first participles in the information content object within a recent third predetermined time, wherein the third predetermined time is greater than the second predetermined time.

[0118] The duration of the third scheduled time can be set in combination with actual technical needs. Generally, the duration of the third scheduled time is set much longer than the duration of the second time, but it cannot be too long, so as to reflect that hot content is mined in the newly released information content object and sudden hot events can be mined. For example, in one embodiment, the third scheduled time can be set to 48 hours. It is understandable that in other embodiments, the third scheduled time can also be set to other durations.

[0119] When obtaining the second frequency of each first participle in the information content object within the latest third predetermined time, the second frequency of each second participle in the third information content object can be obtained by performing frequency statistics on each third information content object within the latest third predetermined time. Since the third predetermined time is greater than the second predetermined time, it can be determined that the third information content object contains the second information content object, and the second participle finally obtained will contain the first participle, so that the second frequency of the first participle can be obtained through the consistency relationship between the first participle and the second participle.

[0120] In a specific example, the unigram language model and the bigram language model may be used to perform frequency statistics on the contents of each third information object within a recent third predetermined time to obtain the second frequency of each second word in the third information content object.

[0121] Step S304: Determine the content hotspot score of each of the second information content objects based on the first frequency and the second frequency of each of the first participles contained in the second information content object.

[0122] When determining the content hotspot score of each of the second information content objects based on the first frequency and the second frequency of each of the first participles contained in the second information content object, various possible methods can be used as long as the hotspot change of the second information content object within the second predetermined time relative to the third predetermined time can be reflected.

[0123] In one embodiment, determining the content hotspot score of each of the second information content objects based on the first frequency and the second frequency of each of the first participles contained in the second information content object may include the following steps S3041 and S3042.

[0124] Step S3041: Determine a segmentation hotspot score of each of the first segmentations based on the first frequency and the second frequency of each of the first segmentations.

[0125] In one embodiment, the word segmentation hotspot score can be the sum of the first frequency and the inverse frequency, and the ratio of the frequency difference to the sum of the inverse frequencies, wherein the inverse frequency is the inverse of the sum of the first frequency and the second frequency, and the frequency difference is the difference between the second frequency and the first frequency.

[0126] The first frequency of a first participle is recorded as F rencent , the second frequency is recorded as F old , the word segmentation hotspot score is recorded as score gram , then the formula for determining the word segmentation hotspot score can be expressed as:

[0127] score gram =(F rencent +likelyscore) / (F old -F rencent +likelyscore)

[0128] Among them, the reciprocal of frequency, likelyscore, is likelyscore = 1 / (F rencent +F old ).

[0129] It is understandable that in other embodiments, other methods may be used to determine the word segmentation hotspot score.

[0130] Step S3042: Determine the content hotspot score of each of the second information content objects based on the hotspot score of each of the first words contained in the second information content object.

[0131] In one embodiment, the word segmentation hotspot scores of the first word segments contained in the second information content object may be directly summed up, and the obtained sum value may be used as the content hotspot score of the second information content object.

[0132] Step S305: Based on the content hotspot score of each of the second information content objects, determine the information content object to be recommended in the second information content objects.

[0133] When determining the information content object to be recommended based on the content hotspot score of each of the second information content objects, the information content object to be recommended (also referred to as the second information content object to be recommended in the embodiment of the present application) can be determined directly based on the content hotspot score.

[0134] When determining the second information content object to be recommended based on the content hotspot score, the determined content hotspot score interval may be obtained, and the second information content object with a content hotspot score within the content hotspot score interval may be determined as the second information content object to be recommended.

[0135] In one embodiment, when determining the information content object to be recommended based on the content hotspot score of each of the second information content objects, the following method may be used, which may specifically include step S3051 and step S3052.

[0136] Step S3051: normalizing the content hotspot scores of each of the second information content objects to obtain a normalized recommendation probability value of each of the second information content objects.

[0137] Through the normalization operation, each click rate score can be mapped to a decimal in the interval (0, 1). The specific normalization method can be performed using an existing normalization method.

[0138] Step S3052: Determine the second information content object to be recommended based on the recommendation probability value of each second information content object after the normalization process.

[0139] Based on the normalized recommendation probability values ​​of each second information content object, various possible methods may be used to determine the second information content object to be recommended. For example, by obtaining a determined recommendation probability interval, the second information content object whose recommendation probability value is within the recommendation probability interval is determined as the second information content object to be recommended.

[0140] In some embodiments, based on the recommendation probability values ​​of the normalized second information content objects, determining the second information content object to be recommended may be performed in the following manner.

[0141] First, based on the recommendation probability value of each second information content object, determine the recommendation probability interval of each second information content object. Thus, based on the recommendation probability value of each second information content object, each second information content object can be mapped to a corresponding recommendation probability interval. For example, assuming that there are 10 second information content objects d1-d10, and their recommendation probability values ​​are all 0.1, the recommended probability intervals mapped to them can be (0, 0.1), [0.1, 0.2), [0.2, 0.3), [0.3, 0.4), [0.4, 0.5), [0.5, 0.6), [0.6, 0.7), [0.7, 0.8), [0.8, 0.9), [0.9, 1).

[0142] Then, a randomly selected second probability value is obtained, and the second information content object corresponding to the recommendation probability interval corresponding to the second probability value is determined as the second information content object to be recommended.

[0143] The randomly selected second probability value may be only one value, or may be obtained by multiple selections to obtain multiple second probability values, and then the second information content object corresponding to the recommended probability interval corresponding to the second probability value is used as the second information content object to be recommended. For example, assuming that the randomly selected second probability values ​​are 0.3, 0.6, and 0.7, then d3, d6, and d7 corresponding to their corresponding intervals [0.3, 0.4), [0.6, 0.7), and [0.7, 0.8) are used as the second information content objects to be recommended.

[0144] Based on the scheme of the embodiment described above, for each second information content object within the second scheduled time that is most recent, by counting the first frequency of the first participle contained therein within the second scheduled time and the second frequency within the third scheduled time, a content hotspot score of each second information content object is calculated, thereby being able to discover hot information content objects from the second information content objects released within the second scheduled time that is most recent, and being able to discover hot content in newly released information content, thereby improving the quality of information content recommendation.

[0145] In one embodiment, Figure 4 As shown, a method for determining the content of information to be recommended is provided, and the method is applied to Figure 1 Taking the server 104 or the server 106 in the example, the method for determining the information content to be recommended combines the click rate score based on the information content object involved in the above-mentioned embodiment and the method for combining the content hotspot score of the information content object to determine the information content to be recommended, including the following steps S401 to S405.

[0146] Step S401: Acquire a first predetermined number of first information content objects that are most recent in time, and a second information content object that is most recent in a second predetermined time.

[0147] The time here may refer to the publishing time of the information content object, such as the publishing time of the information content object of user-generated content such as posts and articles, that is, the publishing time of each second information content object is within the second predetermined time. When obtaining the first predetermined number of first information content objects that are most recently published, the first predetermined number of information content objects that are most recently published may be obtained in reverse order of time.

[0148] Among them, the first predetermined number can be set in combination with actual technical needs. In some embodiments, it can be determined in combination with the number of information content objects stored in each statistical period, and can also be determined in combination with the current business situation. In some embodiments, after obtaining the first predetermined number of first information content objects with the most recent time, the first information content objects containing sensitive words in the first predetermined number of first information content objects can be further removed. Among them, sensitive words generally refer to terms that violate relevant laws and regulations and affect the health of the network environment. In some scenarios, they also involve words that are not suitable to appear on the relevant pages of the current server, etc. By removing the first information content objects containing sensitive words, the recommendation and dissemination of the first information content objects containing sensitive words can be avoided, which is conducive to optimizing the network environment.

[0149] In some embodiments, after obtaining the first predetermined number of first information content objects with the latest time, first information content objects with content lengths less than a predetermined length threshold may be further removed from the first predetermined number of first information content objects. For first information content objects with content lengths less than the predetermined length threshold, it means that they do not actually contain enough information. Therefore, removing these first information content objects can reduce the number of first information content objects involved in subsequent processing, which helps to improve processing efficiency.

[0150] It can be understood that in actual processing, after obtaining the first predetermined number of first information content objects with the most recent time, the first information content objects containing sensitive words and the first information content objects with a content length less than a predetermined length threshold can be removed from the first predetermined number of first information content objects at the same time.

[0151] The duration of the second scheduled time can be set in combination with actual technical needs. Generally, the second scheduled time should be set relatively short to achieve the mining of hot content in the newly released information content object and to mine sudden hot events. However, the second scheduled time can also have a certain length of time to balance the newly released information content objects and the user's behavior such as clicks, likes, comments and other reading-related situations during this period. For example, in one embodiment, the second scheduled time can be set to 2 hours. It is understandable that in other embodiments, the second scheduled time can also be set to other durations.

[0152] Step S402: for a first predetermined number of the first information content objects, obtain the click rate score of each of the first information content objects within a recent first predetermined time; and perform missing value smoothing on each of the click rate scores to obtain the click rate score of each of the first information content objects after the missing value smoothing.

[0153] In one embodiment, the click rate of the first information content object within the first predetermined time may be obtained by obtaining the number of clicks and exposures of each of the first information content objects within the most recent first predetermined time and calculating the ratio of the above number of clicks to the number of exposures, and then the corresponding click rate score within the most recent first predetermined time may be determined in combination with the click rate.

[0154] The duration of the first predetermined time can be set in combination with actual technical needs. For example, in one embodiment, the first predetermined time can be set to 5 minutes. It is understandable that in other embodiments, the first predetermined time can also be set to other durations.

[0155] Among them, the number of clicks and exposures of each first information content object in the most recent first predetermined time can be obtained from relevant log data, such as the MTA log of the server, which records all log data related to the user's behavior of the information content object, such as the exposure action and exposure time of the relevant information of the information content object displayed on the relevant display page of the user, the click action and click time of the user clicking the information content object, the like action and like time of the user liking the information content object, and the comment action, comment content and comment time of the user commenting on the information content object. It can be understood that these recorded relevant information can also include the user ID of the corresponding user at the same time. Therefore, by accessing the MTA log, through the identifier of the information content object and the time of the user action (such as the click time), etc., the number of clicks and exposures of each information content object in the first predetermined time can be screened and statistically obtained.

[0156] In some embodiments, the obtained click rate may be directly used as the click rate score. In other embodiments, the click rate score may be obtained by further calculation based on the click rate.

[0157] In one embodiment, after obtaining the click rate scores of each of the first information content objects within the most recent first predetermined time for the first predetermined number of the first information content objects, the obtained click rate scores may be further subjected to confidence smoothing processing to obtain the click rate scores after confidence smoothing processing, and then the click rate scores after confidence smoothing processing may be subjected to missing value smoothing processing. Thus, by performing confidence smoothing processing on the obtained click rate scores, a smoothing penalty processing is performed so that the click rate scores finally obtained are more consistent with the actual click situation of the information content object. Among them, when performing confidence smoothing processing, the click rate scores may be subjected to confidence smoothing processing according to the exposure amount. Specifically, the greater the exposure amount, the more confident the corresponding click rate score is and the smaller the penalty is. In some specific examples, confidence smoothing processing may be performed based on the Wilson confidence interval algorithm.

[0158] Through the missing value smoothing process, the information content objects without click-through rate scores in each first information content object can also obtain the corresponding click-through rate scores, so that they can have the probability of being recommended and can enter the recommendation library for recommendation, so that the newly released information content objects without actual click-through rate can also enter the recommendation process, solving the cold start problem of the newly distributed content due to the lack of historical data. And by performing the missing value smoothing process after the confidence smoothing process, each first information content object can have a corresponding click-through rate score, and the click-through rate score is credible, which can ensure that the newly released information content objects can have the probability of being recommended and can enter the recommendation process.

[0159] Step S404: Perform frequency statistics on the second information content object to obtain the first frequency of each first participle in the second information content object within the second predetermined time and the second frequency within the most recent third predetermined time; determine the content hotspot score of each second information content object based on the first frequency and the second frequency of each first participle contained in the second information content object, and the third predetermined time is greater than the second predetermined time.

[0160] When performing frequency statistics on the second information content object, the word segmentation model can be combined to perform frequency statistics on the word segmentation. In some embodiments, a unigram language model and a bigram language model can be used to perform frequency statistics on each of the second information content objects to obtain the first frequency of each first word segmentation in the second information content object. Among them, the unigram language model is used as a model for unary word segmentation. It performs word segmentation by dividing the second information content object into individual characters, and performs frequency statistics on this basis. The bigram language model is used as a model for binary word segmentation. It performs word segmentation by dividing every two characters in the second information content object into one word, and performs frequency statistics on this basis. By combining the unigram language model and the bigram language model for frequency statistics, the final frequency statistics can cover more words, so that the coverage of the word segmentation is wider, which helps to improve the accuracy of the final result.

[0161] The duration of the third scheduled time can be set in combination with actual technical needs. Generally, the duration of the third scheduled time is set much longer than the duration of the second time, but it cannot be too long, so as to reflect that hot content is mined in the newly released information content object and sudden hot events can be mined. For example, in one embodiment, the third scheduled time can be set to 48 hours. It is understandable that in other embodiments, the third scheduled time can also be set to other durations.

[0162] When obtaining the second frequency of each first participle in the information content object within the latest third predetermined time, the second frequency of each second participle in the third information content object can be obtained by performing frequency statistics on each third information content object within the latest third predetermined time. Since the third predetermined time is greater than the second predetermined time, it can be determined that the third information content object contains the second information content object, and the second participle finally obtained will contain the first participle, so that the second frequency of the first participle can be obtained through the consistency relationship between the first participle and the second participle.

[0163] In a specific example, the unigram language model and the bigram language model may be used to perform frequency statistics on the contents of each third information object within a recent third predetermined time to obtain the second frequency of each second word in the third information content object.

[0164] When determining the content hotspot score of each of the second information content objects based on the first frequency and the second frequency of each of the first participles contained in the second information content object, various possible methods can be used as long as the hotspot change of the second information content object within the second predetermined time relative to the third predetermined time can be reflected.

[0165] In one embodiment, based on the first frequency and the second frequency of each of the first participles contained in the second information content object, the content hotspot score of each of the second information content objects is determined in the following manner.

[0166] First, based on the first frequency and the second frequency of each of the first participles, a participle hotspot score of each of the first participles is determined.

[0167] In one embodiment, the word segmentation hotspot score can be the sum of the first frequency and the inverse frequency, and the ratio of the frequency difference to the sum of the inverse frequencies, wherein the inverse frequency is the inverse of the sum of the first frequency and the second frequency, and the frequency difference is the difference between the second frequency and the first frequency.

[0168] The first frequency of a first participle is recorded as F rencent , the second frequency is recorded as F old , the word segmentation hotspot score is recorded as score gram , then the formula for determining the word segmentation hotspot score can be expressed as:

[0169] score gram =(F rencent +likelyscore) / (F old -F rencent +likelyscore)

[0170] Among them, the reciprocal of frequency, likelyscore, is likelyscore = 1 / (F rencent +F old ).

[0171] It is understandable that in other embodiments, other methods may be used to determine the word segmentation hotspot score.

[0172] Secondly, based on the word segmentation hotspot scores of each of the first word segments contained in the second information content object, the content hotspot scores of each of the second information content objects are determined.

[0173] In one embodiment, the word segmentation hotspot scores of the first word segments contained in the second information content object may be directly summed up, and the obtained sum value may be used as the content hotspot score of the second information content object.

[0174] Step S405: Determine the information content object to be recommended based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects.

[0175] In some embodiments, when determining the information content object to be recommended based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects, the following method can be used.

[0176] Normalizing the click rate scores of each of the first information content objects after the missing value smoothing process to obtain the normalized recommendation probability values ​​of each of the first information content objects; determining the first information content objects to be recommended based on the normalized recommendation probability values ​​of each of the first information content objects;

[0177] The content hotspot scores of each second information content object are normalized to obtain the normalized recommendation probability values ​​of each second information content object; and the second information content object to be recommended is determined based on the normalized recommendation probability values ​​of each second information content object.

[0178] The determined information content object to be recommended includes the first information content object to be recommended and the second information content object to be recommended.

[0179] By normalizing, each click rate score or content hotspot score can be mapped to a corresponding decimal in the interval (0, 1). The specific normalization method can be performed using an existing normalization method.

[0180] Based on the normalized recommendation probability values ​​of each first information content object, various possible methods may be used to determine the first information content object to be recommended. For example, by obtaining a determined recommendation probability interval, the first information content object whose recommendation probability value is within the recommendation probability interval is determined as the first information content object to be recommended.

[0181] In some embodiments, based on the recommendation probability values ​​of the normalized first information content objects, determining the first information content object to be recommended may be performed in the following manner.

[0182] First, based on the recommendation probability value of each first information content object, determine the recommendation probability interval of each first information content object. Thus, based on the recommendation probability value of each first information content object, each first information content object can be mapped to a corresponding recommendation probability interval. For example, assuming that there are 5 first information content objects D1-D5, and their recommendation probability values ​​are all 0.2, then the recommended probability intervals mapped to them can be (0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), [0.9, 1) respectively.

[0183] Then, a randomly selected first probability value is obtained, and a first information content object corresponding to a recommendation probability interval corresponding to the first probability value is determined as the first information content object to be recommended.

[0184] The first probability value randomly selected may be only one value, or may be obtained by multiple selections to obtain multiple first probability values, and then the first information content object corresponding to the recommended probability interval corresponding to the first probability value is used as the first information content object to be recommended. For example, assuming that the first probability values ​​randomly selected are 0.3 and 0.7, D1 and D4 corresponding to their corresponding intervals [0.2, 0.4) and [0.6, 0.8) are used as the first information content object to be recommended.

[0185] Based on the normalized recommendation probability values ​​of each second information content object, various possible methods may be used to determine the second information content object to be recommended. For example, by obtaining a determined recommendation probability interval, the second information content object whose recommendation probability value is within the recommendation probability interval is determined as the second information content object to be recommended.

[0186] In some embodiments, based on the recommendation probability values ​​of the normalized second information content objects, determining the second information content object to be recommended may be performed in the following manner.

[0187] First, based on the recommendation probability value of each second information content object, determine the recommendation probability interval of each second information content object. Thus, based on the recommendation probability value of each second information content object, each second information content object can be mapped to a corresponding recommendation probability interval. For example, assuming that there are 10 second information content objects d1-d10, and their recommendation probability values ​​are all 0.1, the recommended probability intervals mapped to them can be (0, 0.1), [0.1, 0.2), [0.2, 0.3), [0.3, 0.4), [0.4, 0.5), [0.5, 0.6), [0.6, 0.7), [0.7, 0.8), [0.8, 0.9), [0.9, 1).

[0188] Then, a randomly selected second probability value is obtained, and the second information content object corresponding to the recommendation probability interval corresponding to the second probability value is determined as the second information content object to be recommended.

[0189] The randomly selected second probability value may be only one value, or may be obtained by multiple selections to obtain multiple second probability values, and then the second information content object corresponding to the recommended probability interval corresponding to the second probability value is used as the second information content object to be recommended. For example, assuming that the randomly selected second probability values ​​are 0.3, 0.6, and 0.7, then d3, d6, and d7 corresponding to their corresponding intervals [0.3, 0.4), [0.6, 0.7), and [0.7, 0.8) are used as the second information content objects to be recommended.

[0190] Among them, since in the above step S401, the first information content objects obtained are the first preset number of objects with the most recent release time, and the second information content objects are the objects within the most recent second predetermined time, the first information content objects and the second information content objects are derived from the same information content object pool, so there may be duplication between the first information content objects and the second information content objects, that is, ultimately a certain information content object may be included in both the first information content objects to be recommended and the second information content objects to be recommended.

[0191] At this time, in some embodiments, the first information content object to be recommended and the second information content object to be recommended may be deduplicated and then used as the final information content object to be recommended.

[0192] In some embodiments, since the first information content object to be recommended is mainly used to recommend new information content objects, and the second information content object to be recommended is mainly used to explore hot spots in new information content objects, the first information content object to be recommended and the second information content object to be recommended can be recommended at the same time to achieve the recommendation of different types of information content objects, for example, they can be recommended in different areas or in different ways.

[0193] Based on the scheme of the embodiment described above, for the first predetermined number of first information content objects that are most recent, the click-through rate scores within the most recent first predetermined time are obtained, and missing value processing is performed on this basis, so that the click-through rate of a certain number of information content objects that are recently released can be combined, and the first information content objects without click-through rate scores can also have a certain click-through rate score and can also have a probability of being recommended, thereby solving the cold start problem of the newly released information content due to the lack of historical data. At the same time, for each second information content object within the most recent second predetermined time, by counting the first frequency of the first participle contained therein within the second predetermined time and the second frequency within the third predetermined time, the content hotspot score of each second information content object is calculated, so that hot information content objects can be discovered from the second information content objects that are released within the most recent second predetermined time, and hot content in the newly released information content can be discovered, so that on the basis of solving the cold start problem and being able to effectively recommend new content so that it can enter the recommendation process, real-time hotspots can be discovered, thereby improving the overall recommendation quality.

[0194] Based on the scheme of the embodiment described above, for each second information content object within the second scheduled time that is most recent, by counting the first frequency of the first participle contained therein within the second scheduled time and the second frequency within the third scheduled time, a content hotspot score of each second information content object is calculated, thereby being able to discover hot information content objects from the second information content objects released within the second scheduled time that is most recent, and being able to discover hot content in newly released information content, thereby improving the quality of information content recommendation.

[0195] As shown in reference to FIG6 , by comparing the method for determining the information content to be recommended in the above embodiment that combines the click-through rate score and the content hotspot score with the method for determining the information content to be recommended based only on the click-through rate score, it is found through statistics of the data within 11 days after the two different methods are adopted that both methods can discover and recommend content with good performance from new information content objects. The method that combines the click-through rate score and the content hotspot score adds hotspot screening to the pool of new information content objects, and by comparing the indicators of the period when the number of posts doubles every day during the statistical activity, there has been a significant improvement in the user like rate, user comment rate, UV click rate, and PV click rate.

[0196] It should be understood that, although the various steps in the various flow charts that the above-described embodiments are related to are displayed in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in these flow charts may include a plurality of steps or a plurality of stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps.

[0197] In one embodiment, Figure 7 As shown, a device for determining content of information to be recommended is provided, comprising:

[0198] The first object acquisition module 701 is used to acquire a first predetermined number of first information content objects that are most recent in time;

[0199] A click rate score determination module 702 is used to obtain, for each of the first information content objects, a click rate score of each of the first information content objects within a recent first predetermined time period;

[0200] A missing value smoothing processing module 703 is used to perform missing value smoothing processing on each of the click rate scores, so that the information content objects without click rate scores also obtain the corresponding click rate scores, and obtain the click rate scores after the missing value smoothing processing;

[0201] The first to-be-recommended object determination module 704 is configured to determine the to-be-recommended information content object based on the click-through rate score after the missing value smoothing processing of each of the first information content objects.

[0202] In one embodiment, it further includes a filtering module for removing the first information content objects containing sensitive words from the first predetermined number of first information content objects.

[0203] In one embodiment, the filtering module is further configured to remove first information content objects whose content length is less than a predetermined length threshold from the first predetermined number of first information content objects.

[0204] In one embodiment, the click rate score determination module 702 obtains the number of clicks and exposures of each of the first information content objects within the most recent first predetermined time, and calculates the click rate score of each of the first information content objects within the most recent first predetermined time based on the number of clicks and the number of exposures.

[0205] In one embodiment, it further includes: a confidence smoothing processing module, which is used to perform confidence smoothing processing on the click rate score obtained by the click rate score determination module 702 to obtain the click rate score after the confidence smoothing processing. At this time, the missing value smoothing processing module 703 performs missing value smoothing processing on the click rate score after the confidence smoothing processing by the confidence smoothing processing module.

[0206] In one embodiment, the first to-be-recommended object determination module 704 includes:

[0207] The normalization processing module is used to normalize the click rate scores of each of the first information content objects after the missing value smoothing processing, so as to obtain the recommendation probability value of each of the first information content objects after the normalization processing.

[0208] The first determination module is used to determine the first information content object to be recommended based on the recommendation probability value of each first information content object after the normalization process.

[0209] In one embodiment, the first determination module is used to determine the recommendation probability interval of each of the first information content objects based on the recommendation probability value of each of the first information content objects, and obtain a randomly selected first probability value, and determine the first information content object corresponding to the recommendation probability interval corresponding to the first probability value as the first information content object to be recommended.

[0210] In one embodiment, Figure 8 As shown, a device for determining content of information to be recommended is provided, comprising:

[0211] The second object acquisition module 801 is used to acquire a second information content object within the latest second predetermined time;

[0212] The frequency determination module 802 is used to perform frequency statistics on the second information content object to obtain a first frequency of each first participle in the second information content object; and obtain a second frequency of each first participle in the information content object within a recent third predetermined time, wherein the third predetermined time is greater than the second predetermined time;

[0213] A content hotspot score determination module 803, configured to determine a content hotspot score of each of the second information content objects based on a first frequency and a second frequency of each of the first segmented words contained in the second information content object;

[0214] The second to-be-recommended object determining module 804 determines the to-be-recommended information content object in the second information content objects based on the content hotspot score of each of the second information content objects.

[0215] In one embodiment, the frequency determination module 802 uses a unigram language model and a bigram language model to perform frequency statistics on each of the second information content objects to obtain a first frequency of each first word in the second information content object.

[0216] In one embodiment, the second object acquisition module 801 further acquires each third information content object within a third predetermined time period closest to the publishing time, and the frequency determination module 802 uses a unigram language model and a bigram language model to perform frequency statistics on each of the third information content objects to obtain a second frequency of each second participle in the third information content object, wherein the second participle includes the first participle.

[0217] In one embodiment, the content hotspot score determination module 803 determines the segmentation hotspot score of each of the first segmentations based on the first frequency and the second frequency of each of the first segmentations, and determines the content hotspot score of each of the second information content objects based on the segmentation hotspot score of each of the first segmentations contained in the second information content object.

[0218] In one embodiment, the word segmentation hotspot score can be the sum of the first frequency and the inverse frequency, and the ratio of the frequency difference to the sum of the inverse frequencies, wherein the inverse frequency is the inverse of the sum of the first frequency and the second frequency, and the frequency difference is the difference between the second frequency and the first frequency.

[0219] In one embodiment, the content hotspot score of the second information content object is the sum of the hotspot scores of each first word contained in the second information content object.

[0220] In one embodiment, the second to-be-recommended object determination module 804 includes:

[0221] The normalization processing module is used to normalize the content hotspot score of each second information content object to obtain the recommendation probability value of each second information content object after normalization.

[0222] The second determining module is used to determine the second information content object to be recommended based on the recommendation probability value of each second information content object after the normalization process.

[0223] In one embodiment, the second determination module determines the recommendation probability interval of each second information content object based on the recommendation probability value of each second information content object, obtains a randomly selected second probability value, and determines the second information content object corresponding to the recommendation probability interval corresponding to the second probability value as the second information content object to be recommended.

[0224] In one embodiment, Fig. 9As shown, a device for determining content of information to be recommended is provided, comprising:

[0225] The object acquisition module 901 may specifically include the first object acquisition module 701 and the second object acquisition module 801 as described above;

[0226] A click score determination module 902, which may specifically include the click rate score determination module 702 and the missing value smoothing processing module 703 as described above;

[0227] The hotspot score determination module 903 may specifically include the frequency determination module 802 and the content hotspot score determination module 803 as described above;

[0228] The recommendation object determination module 904 may specifically include the first to-be-recommended object determination module 704 and the second to-be-recommended object determination module 804 as described above;

[0229] In one embodiment, the click score determination module 902 further includes a filtering module for removing the first information content objects containing sensitive words from the first predetermined number of first information content objects.

[0230] In one embodiment, the filtering module is further configured to remove first information content objects whose content length is less than a predetermined length threshold from the first predetermined number of first information content objects.

[0231] In one embodiment, the click score determination module 902 further includes a confidence smoothing processing module, which is used to perform confidence smoothing processing on the click rate score obtained by the click rate score determination module 702 to obtain a click rate score after the confidence smoothing processing. At this time, the missing value smoothing processing module 703 performs missing value smoothing processing on the click rate score after the confidence smoothing processing by the confidence smoothing processing module.

[0232] For the specific definition of the device for determining the content of information to be recommended, please refer to the definition of the method for determining the content of information to be recommended above, which will not be repeated here. Each module in the above-mentioned device for determining the content of information to be recommended can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0233] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.10As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store log data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for determining the content of information to be recommended is implemented.

[0234] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0235] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the method for determining the content of information to be recommended in any one of the above embodiments is implemented.

[0236] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0237] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0238] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0239] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for determining the content of information to be recommended, characterized in that: The method comprises: Acquire a first predetermined number of first information content objects that are most recent in time, and a second information content object that is most recent in a second predetermined time; For a first predetermined number of the first information content objects, obtaining a click rate score of each of the first information content objects within a recent first predetermined time; and performing missing value smoothing processing on each of the click rate scores to obtain a click rate score of each of the first information content objects after the missing value smoothing processing; Perform frequency statistics on the second information content object to obtain a first frequency of each first participle in the second information content object within the second predetermined time and a second frequency within a latest third predetermined time; determine a content hotspot score of each second information content object based on the first frequency and the second frequency of each first participle contained in the second information content object, wherein the third predetermined time is greater than the second predetermined time; The information content object to be recommended is determined based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects.

2. The method according to claim 1, characterized in that Determining the information content object to be recommended based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects includes: Normalizing the click rate scores of each of the first information content objects after the missing value smoothing process to obtain the normalized recommendation probability values ​​of each of the first information content objects; determining the first information content objects to be recommended based on the normalized recommendation probability values ​​of each of the first information content objects; Normalizing the content hotspot scores of each second information content object to obtain a normalized recommendation probability value of each second information content object; determining the second information content object to be recommended based on the normalized recommendation probability value of each second information content object; The information content object to be recommended includes the first information content object to be recommended and the second information content object to be recommended.

3. The method according to claim 1, characterized in that Before performing the missing value smoothing process on each of the click rate scores, the method further includes: The click rate score is subjected to confidence smoothing processing to obtain a click rate score after the confidence smoothing processing.

4. The method according to claim 2, characterized in that: The determining of the first information content object to be recommended based on the recommendation probability value of each first information content object after the normalization process includes: Determining a recommendation probability interval of each of the first information content objects based on the recommendation probability value of each of the first information content objects; A randomly selected probability value is obtained, and a first information content object corresponding to a recommendation probability interval corresponding to the probability value is determined as the first information content object to be recommended.

5. The method according to claim 1, characterized in that The determining of the content hotspot score of each of the second information content objects based on the first frequency and the second frequency of each of the first participles contained in the second information content object includes: Determine a hotspot score of each of the first participles based on the first frequency and the second frequency of each of the first participles; Based on the word segmentation hotspot scores of each of the first word segments contained in the second information content object, the content hotspot scores of each of the second information content objects are determined.

6. The method according to claim 5, characterized in that The word segmentation hotspot score is the sum of the first frequency and the reciprocal of the frequency, and the ratio of the frequency difference to the sum of the reciprocal of the frequency, the reciprocal of the frequency is the reciprocal of the sum of the first frequency and the second frequency, and the frequency difference is the difference between the second frequency and the first frequency.

7. A device for determining the content of information to be recommended, characterized in that: The device comprises: An object acquisition module, used to acquire a first predetermined number of first information content objects that are most recent, and a second information content object that is most recent within a second predetermined time; A click score determination module is used to obtain, for a first predetermined number of the first information content objects, a click rate score of each of the first information content objects within a first predetermined time period; and to perform missing value smoothing processing on each of the click rate scores to obtain a click rate score of each of the first information content objects after the missing value smoothing processing; a hotspot score determination module, configured to perform frequency statistics on the second information content object, obtain a first frequency of each first participle in the second information content object within the second predetermined time and a second frequency within a latest third predetermined time; determine a content hotspot score of each second information content object based on the first frequency and the second frequency of each first participle contained in the second information content object, wherein the third predetermined time is greater than the second predetermined time; The recommended object determination module is used to determine the information content object to be recommended based on the click rate score of each of the first information content objects after the missing value smoothing processing and the content hotspot score of each of the second information content objects.

8. The device according to claim 7, characterized in that The recommended object determination module is further used to normalize the click rate scores of each of the first information content objects after the missing value smoothing process, so as to obtain the recommendation probability value of each of the first information content objects after the normalization process; Based on the recommendation probability values ​​of the normalized first information content objects, determining the first information content objects to be recommended; normalizing the content hotspot scores of the second information content objects to obtain the recommendation probability values ​​of the normalized second information content objects; Based on the recommendation probability values ​​of the normalized second information content objects, the second information content objects to be recommended are determined; the information content objects to be recommended include the first information content objects to be recommended and the second information content objects to be recommended.

9. The device according to claim 7, characterized in that The click score determination module is further used to perform confidence smoothing processing on the click rate scores before performing missing value smoothing processing on each click rate score to obtain the click rate scores after confidence smoothing processing.

10. The device according to claim 8, characterized in that The recommended object determination module is also used to determine the recommended probability interval of each of the first information content objects based on the recommended probability value of each of the first information content objects; obtain a randomly selected probability value, and determine the first information content object corresponding to the recommended probability interval corresponding to the probability value as the first information content object to be recommended.

11. The device according to claim 7, characterized in that The hotspot score determination module is further used to determine the segmentation hotspot score of each of the first segmentations based on the first frequency and the second frequency of each of the first segmentations; and to determine the content hotspot score of each of the second information content objects based on the segmentation hotspot score of each of the first segmentations contained in the second information content object.

12. The device according to claim 11, characterized in that The word segmentation hotspot score is the sum of the first frequency and the reciprocal of the frequency, and the ratio of the frequency difference to the sum of the reciprocal of the frequency, the reciprocal of the frequency is the reciprocal of the sum of the first frequency and the second frequency, and the frequency difference is the difference between the second frequency and the first frequency.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

14. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 6.

15. A computer program product, comprising computer instructions, which implement the method according to any one of claims 1 to 6 when executed by a processor.

Images