Method and device for sorting comments

Abstract

This application belongs to the field of Internet technology, and discloses a method and device for comment sorting. The method includes: obtaining the basic comment data of each comment; for any comment, determining the first dynamic quality score of the comment based on the number of likes, number of replies, user level, proportion of historical high-quality comments, content information, number of reports and click-through rate; for any comment, determining the attenuation coefficient of the comment based on the release time and content information, and determining the second dynamic quality score as the product of the attenuation coefficient and the first dynamic quality score; for any comment, when it is detected that the number of consecutive days of dominating the list of the comment exceeds the preset number of days, determining the dominating penalty coefficient based on the consecutive dominating days, and determining the third dynamic quality score as the product of the dominating penalty coefficient and the second dynamic quality score; sorting the comments according to the third dynamic quality score. This application can be used to sort the comments based on the quality of the comments.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method and device for sorting comments. Background Art

[0002] Currently, ticketing apps rely on fixed ranking rules for review ranking, such as descending order by number of likes or posting date. This results in older, more engaged reviews consistently occupying the top positions, placing excessive reliance on static weights. Furthermore, because older reviews can't naturally sink to the bottom, higher-quality new reviews aren't visible to other users, making them less likely to gain exposure. Consequently, the information users receive about new reviews is less fresh and timely. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and device for sorting comments to address the above technical issues.

[0004] In a first aspect, a method for sorting comments is provided, the method comprising:

[0005] Obtain basic comment data for each comment, including posting time, number of likes, number of replies, user rating, percentage of historical high-quality comments, content information, number of reports, and click-through rate;

[0006] For any comment, determine a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate;

[0007] For any comment, determine the attenuation coefficient of the comment based on the publishing time and the content information, and multiply the attenuation coefficient and the first dynamic quality score to determine the second dynamic quality score of the comment after attenuation;

[0008] For any comment, when it is detected that the number of consecutive days that the comment has been ranked on the top of the list exceeds a preset threshold, a dominance penalty coefficient is determined based on the consecutive dominance period, and the product of the dominance penalty coefficient and the second dynamic quality score is determined as the third dynamic quality score after the penalty; the consecutive dominance period is the number of consecutive days that the comment is ranked before the preset ranking threshold;

[0009] The comments are sorted according to the third dynamic quality score, and the comments are displayed to the user in order.

[0010] As an optional implementation manner, determining the first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate includes:

[0011] Determine an interaction score for the comment based on the number of likes, the number of replies, and the click-through rate;

[0012] Determine the user authority score of the review based on the user level and the proportion of historical high-quality reviews;

[0013] Determine a content value score of the review based on the content information;

[0014] Determine the reporting penalty score for the comment based on the number of reports;

[0015] The difference between the sum of the product of the interaction score and the preset interaction score weight, the product of the user authority score and the preset user authority score weight, the product of the content value score and the preset content value score weight, and the product of the report penalty score and the preset report penalty score weight is determined as the first dynamic quality score of the comment.

[0016] As an optional implementation, the formula for determining the interaction score of the comment based on the number of likes, the number of replies, and the click-through rate is:

[0017]

[0018] Among them, InteractionScore represents the interaction score, L represents the number of likes, R represents the number of replies, and CTR represents the click-through rate.

[0019] As an optional implementation, the formula for determining the user authority score of the review based on the user level and the proportion of historical high-quality reviews is:

[0020]

[0021] Among them, UserAuthority represents the user authority rating, VIPLevel represents the user level, and HistoryQuality represents the proportion of historical high-quality reviews.

[0022] As an optional implementation, the content information includes text length, content format, and multimedia tags corresponding to the content format. The formula for determining the content value score of the review based on the content information is:

[0023] ContentValue=min(10, 2×MediaFlag+0.1×Length)

[0024] Among them, ContentValue represents the content value score, MediaFlag represents the multimedia tag, and Length represents the text length.

[0025] As an optional implementation, the formula for determining the reporting penalty score of the comment based on the number of reports is:

[0026]

[0027] Among them, ReportPenalty represents the reporting penalty score, and ReporCount represents the number of reports.

[0028] As an optional implementation, the method further includes:

[0029] For any comment, a pre-stored first exposure rate threshold is assigned to the comment; if the publishing duration of the comment reaches the first preset duration threshold, determining whether the click-through rate is less than the first preset click-through rate threshold; if so, determining the pre-stored first exposure rate threshold as the exposure rate of the comment;

[0030] If the click rate is greater than or equal to the first preset click rate threshold and less than a pre-stored second click rate threshold, the difference between the product of the first preset exposure coefficient and the click rate and the second preset exposure coefficient is determined as the exposure rate of the comment;

[0031] If the click rate is greater than or equal to the second click rate threshold, the pre-stored second exposure rate threshold is determined as the exposure of the comment, and the first exposure rate threshold is less than the second exposure rate threshold.

[0032] As an optional implementation, the content information includes a content format. When the content format does not include a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is:

[0033]

[0034] Where Decay represents the decay coefficient, and T represents the release duration;

[0035] When the content format includes a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is:

[0036]

[0037] Here, Decay represents the attenuation coefficient, and T represents the release duration.

[0038] As an optional implementation, the formula for determining the dominance penalty coefficient based on the consecutive days of dominance is:

[0039] P = max(0.5, 1 − 0.1 × TopDays)

[0040] Among them, P represents the penalty coefficient for dominating the chart, and TopDays represents the number of consecutive days of dominating the chart.

[0041] In a second aspect, a device for sorting comments is provided, the device comprising:

[0042] An acquisition module is used to obtain the basic comment data of each comment, including the posting time, number of likes, number of replies, user level, proportion of historical high-quality comments, content information, number of reports, and click-through rate;

[0043] A first determination module is configured to determine, for any comment, a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate;

[0044] a second determination module configured to determine, for any comment, a decay coefficient of the comment based on the publishing time and the content information, and determine a product of the decay coefficient and the first dynamic quality score as a second dynamic quality score of the comment after decay;

[0045] A third determination module is configured to determine, for any comment, a ranking dominance penalty coefficient based on the number of consecutive ranking days detected to have exceeded a preset ranking threshold, and to multiply the ranking dominance penalty coefficient by the second dynamic quality score as a third dynamic quality score after the penalty; the consecutive ranking days being the number of consecutive days the comment was ranked above the preset ranking threshold;

[0046] The sorting module is used to sort the comments according to the third dynamic quality score and display the comments to the user in order.

[0047] As an optional implementation manner, the first determining module is specifically configured to:

[0048] Determine an interaction score for the comment based on the number of likes, the number of replies, and the click-through rate;

[0049] Determine the user authority score of the review based on the user level and the proportion of historical high-quality reviews;

[0050] Determine a content value score of the review based on the content information;

[0051] Determine the reporting penalty score for the comment based on the number of reports;

[0052] The difference between the sum of the product of the interaction score and the preset interaction score weight, the product of the user authority score and the preset user authority score weight, the product of the content value score and the preset content value score weight, and the product of the report penalty score and the preset report penalty score weight is determined as the first dynamic quality score of the comment.

[0053] As an optional implementation, the device further includes:

[0054] a judgment module configured to assign a pre-stored first exposure rate threshold to any comment, and if the publishing duration of the comment reaches the first preset duration threshold, determine whether the click-through rate is less than the first preset click-through rate threshold; if so, determine the pre-stored first exposure rate threshold as the exposure rate of the comment;

[0055] a fourth determining module, configured to determine, if the click-through rate is greater than or equal to the first preset click-through rate threshold and less than a pre-stored second click-through rate threshold, the difference between the product of the first preset exposure coefficient and the click-through rate and the second preset exposure coefficient as the exposure rate of the comment;

[0056] The fifth determination module is configured to determine the exposure of the comment using a pre-stored second exposure rate threshold if the click rate is greater than or equal to the second click rate threshold, and the first exposure rate threshold is less than the second exposure rate threshold.

[0057] In a third aspect, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the method steps described in the first aspect are implemented.

[0058] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method steps described in the first aspect are implemented.

[0059] The present application provides a method for sorting comments. The technical solution provided by the embodiments of the present application brings at least the following beneficial effects: the quality of each comment is dynamically scored through multiple dimensions such as the interaction of each comment, the user authority of each comment, the value of the comment content, and whether it has been reported by other users, integrating user interaction, content value and user authority to quantify the quality of the comments. In order to avoid the rigidity of the top comments, an adaptive time decay mechanism is set, and the differentiated decay of different time windows is achieved through a piecewise function. To prevent comments from dominating the list, an anti-dominance coefficient is set to prevent the rigidity of the sorting. Thus, the comments are sorted according to their scores, and the quality of the sorting of comments is greatly improved, the user interaction rate is increased, the user's negative feedback is reduced, and the retention rate of users of the comment module is also improved.

[0060] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0062] Figure 1 A flowchart of a method for sorting comments provided in an embodiment of the present application;

[0063] Figure 2 A schematic diagram of the structure of a comment sorting device provided in an embodiment of the present application;

[0064] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0066] The following will describe in detail a method for sorting comments provided by an embodiment of the present application in conjunction with a specific implementation method. Figure 1 A flowchart of a method for sorting comments provided in an embodiment of the present application is shown as follows: Figure 1 The specific steps are as follows:

[0067] Step 101, obtain the basic comment data of each comment, which includes the publishing time, number of likes, number of replies, user level, proportion of historical high-quality comments, content information, number of reports and click-through rate.

[0068] In practice, existing techniques rank reviews in ticketing software based on interaction data or posting time, resulting in older, more interactive reviews permanently occupying the top positions. Therefore, to allow older comments and content to naturally decline while allowing high-quality new comments to be seen by other users, the review ranking method for ticketing software needs to be redefined. Each review's quality score can be determined based on its interaction level, user authority, the value of the review content, and whether it has been reported by other users. Then, a decay function can be implemented based on the comment's age, allowing it to decline as the comment's posting time increases. Comments that have consistently topped the review rankings can also be penalized to prevent dominance, forcing them to decline. First, the quality score of each review needs to be determined. Comment interaction can be determined based on the number of likes, replies, and click-through rate, so the number of likes, replies, and click-through rate for each comment needs to be obtained. User authority can be determined based on the user's rank and the percentage of historically high-quality reviews, so the user rank and percentage of historically high-quality reviews for each comment need to be obtained. The content value score of a review can also be determined based on its content information, so the content information for each comment needs to be obtained. Comments can also be penalized based on the number of reports, so the number of reports for each comment is also required. Therefore, if you want to sort the comments, you need to first obtain the basic comment data of each comment, which includes the time it was published, the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, content information, the number of reports, and the click-through rate.

[0069] Step 102: For any comment, determine the first dynamic quality score of the comment based on the number of likes, number of replies, user level, proportion of historical high-quality comments, content information, number of reports and click-through rate.

[0070] In practice, if you want to sort the comments, you need to first obtain the first dynamic quality score. You can determine the first dynamic quality score of each comment based on the interaction of each comment, the user authority of each comment, the value of the comment content, and whether it has been reported by other users. The number of likes, number of replies, and click-through rate are parameters of a comment, so the interaction score of a comment can be determined based on the number of likes, number of replies, and click-through rate. The user authority score of a comment can be determined based on the user level and the proportion of historical high-quality comments. The content value score of a comment can be determined based on the content information of the comment. The report penalty score of the comment can be determined based on the number of reports. Therefore, for any comment, the first dynamic quality score of the comment can be determined based on the number of likes, number of replies, user level, proportion of historical high-quality comments, content information, number of reports, and click-through rate.

[0071] Furthermore, the specific steps of executing step 102 are as follows:

[0072] Step 1: Determine the interaction score of the comment based on the number of likes, replies, and click-through rate.

[0073] In implementation, after obtaining the number of likes, replies, and click-through rate, the interaction score of the comment can be determined based on the number of likes, replies, and click-through rate. A subsequent step can determine the first dynamic quality score based on the interaction score.

[0074] As an optional implementation, the formula for determining the interaction score of the comment based on the number of likes, number of replies, and click-through rate is:

[0075]

[0076] Among them, InteractionScore represents the interaction score, L represents the number of likes, R represents the number of replies, and CTR represents the click-through rate.

[0077] For example, if the number of likes (L) of a comment is 10, the number of replies (R) is 5, and the click-through rate (CTR) is 0.18, then the interaction score is calculated as InteractionScore = +0.5×5 0.6 +0.2×0.18=3.09+1.31+0.04=4.44.

[0078] Step 2: Determine the user authority score of the review based on the user level and the proportion of historical high-quality reviews.

[0079] In implementation, after obtaining the user level and historical percentage of high-quality reviews for a review, a user authority score for the review can be determined based on the user level and the percentage of high-quality reviews. User levels can include regular users and VIP users. Regular users are assigned a score of 1, while VIP users are assigned a score of 2. The percentage of high-quality reviews can be scored on a scale of 0-10. Subsequent steps can determine a first dynamic quality score based on the user authority score.

[0080] As an optional implementation, the formula for determining the user authority score of a review based on the user level and the proportion of historical high-quality reviews is:

[0081]

[0082] Among them, UserAuthority represents the user authority rating, VIPLevel represents the user level, and HistoryQuality represents the proportion of historical high-quality reviews.

[0083] For example, the commenting user is a VIP user with a user level (VIPLevel) of 2. The proportion of historical high-quality comments (HistoryQuality) is 8 / 10. Then, to calculate the user authority score, UserAuthority = 2×1.5 + 8 / 10 = 3.0 + 0.8 = 3.8.

[0084] Step three: Determine the content value score of the review based on the content information.

[0085] In implementation, after obtaining the content information of a review, a content value score for the review can be determined based on the content information. The content information includes text length, content format, and multimedia tags corresponding to the content format. The content format can be None, Image, or Video. If the content format is None, the corresponding multimedia tag is 0; if the content format is Image, the corresponding multimedia tag is 2; and if the content format is Video, the corresponding multimedia tag is 3. Subsequent steps can determine a first dynamic quality score based on the content value score.

[0086] As an optional implementation, the formula for determining the content value score of the review based on the content information is:

[0087] ContentValue=min(10, 2×MediaFlag+0.1×Length)

[0088] Among them, ContentValue represents the content value score, MediaFlag represents the multimedia tag, and Length represents the text length.

[0089] For example, if the content of a review is 120 words long and includes an image, the multimedia flag is 2. Then, to calculate the content value score, ContentValue = min(10, 2×MediaFlag+0.1×Length) = min(10, 4+12) = 10.

[0090] Step 4: Determine the reporting penalty score for the comment based on the number of reports.

[0091] In implementation, after obtaining the number of reports of a comment, a report penalty score for the comment can be determined based on the number of reports. In subsequent steps, the first dynamic quality score can be determined based on the report penalty score.

[0092] As an optional implementation, the formula for determining the report penalty score for the comment based on the number of reports is:

[0093]

[0094] Among them, ReportPenalty represents the reporting penalty score, and ReporCount represents the number of reports.

[0095] For example, if the number of reports for a comment is 0, the report penalty score is ReportPenalty= =0.

[0096] Step 5: Determine the first dynamic quality score of the comment as the difference between the sum of the product of the interaction score and the preset interaction score weight, the product of the user authority score and the preset user authority score weight, the content value score and the preset content value score weight, and the product of the report penalty score and the preset report penalty score weight.

[0097] In implementation, after determining a review's interaction score, user authority score, content value score, and report penalty score, a first dynamic quality score for the review can be determined based on these scores. The interaction score, user authority score, and content value score combine user interaction, content value, and user authority to quantify review quality. However, the report penalty score can lower the quality of the review. Therefore, the difference between the sum of the product of the interaction score and a preset interaction score weight, the product of the user authority score and a preset user authority score weight, and the product of the content value score and a preset content value score weight, and the product of the report penalty score and a preset report penalty score weight, is used to determine the first dynamic quality score for the review. The preset interaction score weight can be 0.5, the preset user authority score weight can be 0.3, the preset content value score weight can be 0.2, and the preset report penalty score weight can be 0.1. Since interaction is still a key factor in ranking reviews, the preset interaction score weight is relatively large. Secondly, user authority and the value of a review's content are crucial for other users to use as a reference. Therefore, the preset user authority and content value scores are weighted more heavily, while the preset report penalty score is weighted least. In practice, the preset interaction score weights, user authority score weights, content value score weights, and report penalty score weights can also be other values, and are not limited here. This allows for a multi-dimensional review quality assessment.

[0098] For example, if a comment's interaction score is 4.4, its user authority score is 3.8, its content value score is 10, and its report penalty score is 0, the preset interaction score weight can be 0.5, the preset user authority score weight can be 0.3, the preset content value score weight can be 0.2, and the preset report penalty score weight can be 0.1. Therefore, the first dynamic quality score Q = 0.5 × 4.44 + 0.3 × 3.8 + 0.2 × 10 − 0.1 × 0 = 1.78 + 1.14 + 2.0 = 4.92.

[0099] Step 103 : For any comment, determine the attenuation coefficient of the comment based on the publishing time and content information, and multiply the attenuation coefficient and the first dynamic quality score as the second dynamic quality score of the comment after attenuation.

[0100] In practice, to prevent old comments from occupying the top of the ranking for a long time, a dynamic decay mechanism can be introduced. As the age of a comment increases, older comments can be adaptively decayed and pushed down, thus avoiding the rigidity of the existing comment ranking. The degree of comment decay can vary depending on the age of the comment. When comments are young, new comments face a cold start dilemma and are more likely to sink to the bottom due to insufficient initial interaction data. Alternatively, high-quality new comments may have more subsequent interaction data. Therefore, a large decay rate is not required for new comments with a short history. Furthermore, to ensure that other users can obtain the latest and most timely information through comments, excessive decay is not required in the early stages of comment publishing. Therefore, when the age of a comment is short, the decay rate is low. In the middle of the comment's lifespan, the decay rate can increase with the increase in age. When comments are long, their timeliness is low, and to avoid rigid ranking, the comments will continue to decay at a higher rate. If a comment is marked as "essence", the decay rate of the comment will be reduced by 30%, that is, when the decay coefficient of the essence comment is determined based on the release time, the release time will be reduced by 30% before the decay coefficient is determined. Furthermore, comments containing videos can be considered to have high content value and can decay in the middle of the release time, and the decay stage can be extended by a preset time. The extended preset time can be 24 hours. Therefore, the decay of comments can be achieved in the form of a piecewise function, realizing differentiated decay in different time windows, so that the comments can adaptively decay in time. The product value of the decay coefficient and the first dynamic quality score is determined as the second dynamic quality score of the comment after decay.

[0101] As an optional implementation, the content information includes the content format. When the content format does not include a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is:

[0102]

[0103] Here, Decay represents the attenuation coefficient, and T represents the release duration.

[0104] When the content format includes video, the formula for determining the comment attenuation coefficient based on the release time and content information is:

[0105]

[0106] Here, Decay represents the attenuation coefficient, and T represents the release duration.

[0107] Step 104: For any comment, when it is detected that the number of consecutive days that the comment has been on the top of the list exceeds the preset number of days, a dominance penalty coefficient is determined based on the consecutive dominance days, and the product of the dominance penalty coefficient and the second dynamic quality score is determined as the third dynamic quality score after the penalty; the consecutive dominance days are the number of days that the comment is continuously ranked before the preset ranking threshold.

[0108] In practice, in order to avoid rigid ranking, in addition to adaptive time decay of comments, an anti-dominance mechanism can also be set up to prevent rigid dominance of comments. A dominance penalty coefficient can be determined for a comment that has been on the top of the list for a long time, and the product of the dominance penalty coefficient and the second dynamic quality score can be determined as the third dynamic quality score after the penalty. In this way, after integrating multi-dimensional quality assessment and adaptive time decay, multiple protections against dominance can be carried out. Therefore, for any comment, when it is detected that the number of consecutive days of dominance of the comment exceeds the preset number of days, a dominance penalty coefficient will be determined based on the number of consecutive days of dominance, and the product of the dominance penalty coefficient and the second dynamic quality score can be determined as the third dynamic quality score after the penalty; the number of consecutive days of dominance is the number of days that the comment is continuously ranked before the preset ranking threshold.

[0109] As an optional implementation, the formula for determining the penalty coefficient for dominating the charts based on the number of consecutive days of dominating the charts is:

[0110] P = max(0.5, 1 − 0.1 × TopDays)

[0111] Among them, P represents the penalty coefficient for dominating the chart, and TopDays represents the number of consecutive days of dominating the chart.

[0112] For example, comment C_003 has been in the top 3 for four consecutive days and requires dynamic demotion. The number of consecutive days the comment has been in the top 3 is checked. The dominance coefficient is calculated as P = max(0.5, 1 − 0.1 × 4) = 0.6. Therefore, the second dynamic quality score is 25, and the third dynamic quality score after the penalty is 25 × 0.6 = 15. This causes the comment to drop from the top 3 to the top 25.

[0113] Step 105 : sort the comments according to the third dynamic quality score, and display the comments to the user in order.

[0114] In implementation, after determining the third dynamic quality score, the comments may be sorted according to the third dynamic quality score, in descending order of the score, and the comments may be displayed to the user in that order.

[0115] Furthermore, when new comments are first published, they have a high probability of sinking to the bottom of the forums (80%) due to insufficient initial interaction data. To address this issue, we perform cold start optimization for new comments. The specific process is as follows:

[0116] Step A: For any comment, a pre-stored first exposure rate threshold is assigned to the comment. If the publishing time of the comment reaches the first preset time threshold, it is determined whether the click-through rate is less than the first preset click-through rate threshold. If so, the pre-stored first exposure rate threshold is determined as the exposure rate of the comment.

[0117] In practice, to increase the exposure of new comments, a pre-stored first exposure threshold is randomly assigned to each new comment. Then, when the comment's publishing time reaches the first preset exposure threshold, the comment's click-through rate (CTR) is obtained. The pre-stored first exposure threshold can be 5%. The CTR is calculated as the number of clicks divided by the number of impressions. A higher CTR indicates good interaction with the comment, indicating high quality and user popularity. To prevent the comment from sinking to the bottom of the list, the comment's exposure can be increased. A very low CTR indicates poor interaction with the comment. Consequently, no further exposure will be allocated to the comment. Therefore, after obtaining the CTR, the CTR is compared with the first preset CTR threshold. If the CTR is determined to be less than the first preset CTR threshold, the comment is further assigned the exposure of the first exposure threshold. The preset exposure threshold can be 12 hours, and the pre-stored first preset CTR threshold can be 0.03. For any comment, if the publishing time of the comment reaches the first preset time threshold, it is determined whether the click-through rate is greater than the pre-stored first preset click-through rate threshold. If so, the pre-stored first exposure rate threshold is determined as the exposure rate of the comment.

[0118] Step B: If the click rate is greater than or equal to the first preset click rate threshold and less than the pre-stored second click rate threshold, the difference between the product of the first preset exposure coefficient and the click rate and the second preset exposure coefficient is determined as the exposure rate of the comment.

[0119] In implementation, when the comment's publishing duration reaches a first preset duration threshold, the click-through rate (CTR) is compared to the first preset CTR threshold and a pre-stored second CTR threshold. If the CTR is greater than or equal to the first preset CTR threshold and less than the pre-stored second CTR threshold, the comment's interaction data is good and it can be given more exposure. The second CTR threshold can be 0.15. The comment's exposure rate is then determined as the difference between the product of the first preset exposure coefficient and the CTR and the second preset exposure coefficient. The first and second preset exposure coefficients can be coefficients of a proportional relationship determined based on the first preset CTR threshold, the first exposure threshold, and the second CTR threshold and the second exposure threshold, as linear endpoints. The first preset exposure coefficient can be 95 / 12, and the second preset exposure coefficient can be 3 / 16. In practical applications, the first and second preset exposure coefficients can also take other values, which are not limited here.

[0120] Step C: If the click rate is greater than or equal to the second click rate threshold, the pre-stored second exposure rate threshold is determined as the exposure rate of the comment, and the first exposure rate threshold is less than the second exposure rate threshold.

[0121] During implementation, when the publishing time of the comment reaches the first preset time threshold, the click-through rate and the second click-through rate threshold are compared. If the click-through rate is greater than or equal to the second click-through rate threshold, it means that the comment has a good interaction and is very popular with users. In order to allow more users to see the comment and further prevent old comments from dominating the list or ranking in the front row, it is necessary to increase the exposure of the comment. The pre-stored second exposure rate threshold is determined as the exposure of the comment, and the first exposure rate threshold is less than the second exposure rate threshold. Among them, the second click-through rate threshold can be 0.15, and the second exposure rate threshold can be 100%. Therefore, if the click-through rate is greater than or equal to the second click-through rate threshold, the comment will be fully exposed. After full exposure, the interaction data of the comment will increase, and then the ranking of the comment will increase.

[0122] Furthermore, when the time period of the new comment being published reaches a second preset time period, the exposure rate of the new comment is fully exposed. The second preset time period may be 24 hours.

[0123] As an optional implementation, a 5% exposure rate is first allocated to new comments. Then, when the new comments are published for 12 hours, the exposure rate is allocated to the new comments according to the following formula:

[0124]

[0125] Among them, Exposure means exposure rate and CTR means click-through rate.

[0126] In this way, when CTR > 0.15, all new comments are exposed. When 0.03 < CTR < 0.15, the exposure rate is calculated according to the CTR value, and then all comments are exposed after 24 hours. When CTR < 0.03, the comments are transferred to a low-priority queue, and no new exposure is added within 24 hours. All comments are exposed again after 24 hours.

[0127] Furthermore, during the cold start phase, it is also possible to push based on interest matching: based on the user's historical behavior (such as the regular train number G123), relevant new comments are pushed定向.

[0128] The embodiment of this application provides a method for comment ranking. The quality of each comment is dynamically scored through multiple dimensions such as the interaction situation of each comment, the user authority of each comment, the value of the comment content, and whether it has been reported by other users. By integrating user interaction, content value, and user authority, the quality of the comment is quantified. To avoid the rigidity of the top-ranked comments, an adaptive time decay mechanism is set up, and different attenuation in different time windows is achieved through a piecewise function. To improve the exposure rate of new comments, cold start optimization is carried out, and the exposure rate of new comments is increased through small-flow testing and initial weighting. To prevent comments from dominating the rankings, an anti-dominance coefficient is set to prevent ranking rigidity. Thus, the comments are ranked according to the comment scores from high to low, greatly improving the quality of comment ranking, increasing the user interaction rate, reducing negative feedback from users, and also improving the retention rate of users in the comment module. Through cold start optimization, the exposure rate of new comments is greatly increased.

[0129] It should be understood that although Figure 1 each step in the flowchart of is displayed in sequence according to the arrow indication, these steps are not necessarily executed in the order indicated by the arrow. Unless there is a clear indication in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 at least a part of the steps in may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps in other steps.

[0130] It can be understood that the same / similar parts among the various embodiments of the above methods in this specification can be referred to each other. Each embodiment focuses on the differences from other embodiments. For the relevant parts, refer to the descriptions of other method embodiments.

[0131] The embodiment of this application also provides a comment ranking device, as Figure 2 shown. This device includes:

[0132] Acquisition module 201 is used to obtain basic comment data for each comment, including the posting time, number of likes, number of replies, user level, percentage of historical high-quality comments, content information, number of reports, and click-through rate;

[0133] A first determination module 202 is configured to determine, for any comment, a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the percentage of historical high-quality comments, the content information, the number of reports, and the click-through rate;

[0134] A second determining module 203 is configured to determine, for each comment, a decay coefficient of the comment based on the publishing time and the content information, and determine a product of the decay coefficient and the first dynamic quality score as a second dynamic quality score of the comment after decay;

[0135] The third determination module 204 is configured to determine, for any comment, a ranking dominance penalty coefficient based on the number of consecutive ranking days detected to exceed a preset threshold, and to multiply the ranking dominance penalty coefficient by the second dynamic quality score as the third dynamic quality score after the penalty; the consecutive ranking days are the number of consecutive days the comment is ranked above the preset threshold;

[0136] The sorting module 205 is configured to sort the comments according to the third dynamic quality score and display the comments to the user in order.

[0137] As an optional implementation manner, the first determining module 202 is specifically configured to:

[0138] Determine an interaction score for the comment based on the number of likes, the number of replies, and the click-through rate;

[0139] Determine the user authority score of the review based on the user level and the proportion of historical high-quality reviews;

[0140] Determine a content value score of the review based on the content information;

[0141] Determine the reporting penalty score for the comment based on the number of reports;

[0142] The difference between the sum of the product of the interaction score and the preset interaction score weight, the product of the user authority score and the preset user authority score weight, the product of the content value score and the preset content value score weight, and the product of the report penalty score and the preset report penalty score weight is determined as the first dynamic quality score of the comment.

[0143] As an optional implementation, the device further includes:

[0144] a judgment module configured to assign a pre-stored first exposure rate threshold to any comment, and if the publishing duration of the comment reaches the first preset duration threshold, determine whether the click-through rate is less than the first preset click-through rate threshold; if so, determine the pre-stored first exposure rate threshold as the exposure rate of the comment;

[0145] a fourth determining module, configured to determine, if the click-through rate is greater than or equal to the first preset click-through rate threshold and less than a pre-stored second click-through rate threshold, the difference between the product of the first preset exposure coefficient and the click-through rate and the second preset exposure coefficient as the exposure rate of the comment;

[0146] The fifth determination module is configured to determine the exposure of the comment using a pre-stored second exposure rate threshold if the click rate is greater than or equal to the second click rate threshold, and the first exposure rate threshold is less than the second exposure rate threshold.

[0147] The embodiment of the present application provides a device for sorting comments, which dynamically scores the quality of each comment through multiple dimensions such as the interaction of each comment, the user authority of each comment, the value of the comment content, and whether it has been reported by other users. It integrates user interaction, content value and user authority to quantify the quality of the comments. In order to avoid the rigidity of the top comments, an adaptive time decay mechanism is set, and differentiated decay of different time windows is achieved through piecewise functions. In order to increase the exposure rate of new comments, cold start optimization is performed, and the exposure rate of new comments is improved through small traffic tests and initial weighting. To prevent comments from dominating the list, an anti-dominance coefficient is set to prevent the rigidity of sorting. Thus, the comments are sorted according to their scores, and the quality of comment sorting is greatly improved, the user interaction rate is increased, the user's negative feedback is reduced, and the retention rate of users of the comment module is also improved. Through cold start optimization, the exposure rate of new comments is greatly increased.

[0148] For the specific definition of the comment sorting device, please refer to the definition of the comment sorting method above, which will not be repeated here. The various modules in the above comment sorting device can be implemented in whole or in part by software, hardware, or a combination thereof. The above 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 the above modules.

[0149] In one embodiment, a computer device is provided, such as Figure 3 As shown, it includes a memory and a processor, the memory stores a computer program that can be run on the processor, and the processor implements the above-mentioned method steps for sorting comments when executing the computer program.

[0150] In one embodiment, a computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned comment ranking method when executed by a processor.

[0151] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0152] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0153] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0154] Each embodiment in this specification is described in a related manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiment is generally similar to the method embodiment, so the description is relatively simple. For related parts, refer to the description of the method embodiment.

[0155] The technical features of the above embodiments can 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.

[0156] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for sorting comments, characterized in that: The method comprises: Obtain basic comment data for each comment, including posting time, number of likes, number of replies, user rating, percentage of historical high-quality comments, content information, number of reports, and click-through rate; For any comment, determine a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate; For any comment, the attenuation coefficient of the comment is determined based on the publishing time and the content information, and the product of the attenuation coefficient and the first dynamic quality score is determined as the second dynamic quality score of the comment after attenuation; the content information includes the content format. When the content format does not include a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is: ； Where Decay represents the decay coefficient, and T represents the release duration; When the content format includes a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is: ； Where Decay represents the decay coefficient, and T represents the release duration; For any comment, when it is detected that the number of consecutive days that the comment has been ranked on the top of the list exceeds a preset threshold, a dominance penalty coefficient is determined based on the consecutive dominance period, and the product of the dominance penalty coefficient and the second dynamic quality score is determined as the third dynamic quality score after the penalty; the consecutive dominance period is the number of consecutive days that the comment is ranked before the preset ranking threshold; The comments are sorted according to the third dynamic quality score, and the comments are displayed to the user in order.

2. The method according to claim 1, characterized in that Determining a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate includes: Determine an interaction score for the comment based on the number of likes, the number of replies, and the click-through rate; Determine the user authority score of the review based on the user level and the proportion of historical high-quality reviews; Determine a content value score of the review based on the content information; Determine the reporting penalty score for the comment based on the number of reports; The difference between the sum of the product of the interaction score and the preset interaction score weight, the product of the user authority score and the preset user authority score weight, the product of the content value score and the preset content value score weight, and the product of the report penalty score and the preset report penalty score weight is determined as the first dynamic quality score of the comment.

3. The method according to claim 2, characterized in that The formula for determining the interaction score of the comment based on the number of likes, the number of replies, and the click-through rate is: ； Among them, InteractionScore represents the interaction score, L represents the number of likes, R represents the number of replies, and CTR represents the click-through rate.

4. The method according to claim 2, characterized in that The content information includes text length, content format, and multimedia tags corresponding to the content format. The formula for determining the content value score of the review based on the content information is: ContentValue=min(10, 2×MediaFlag+0.1×Length); Among them, ContentValue represents the content value score, MediaFlag represents the multimedia tag, and Length represents the text length.

5. The method according to claim 2, characterized in that The formula for determining the reporting penalty score for the comment based on the number of reports is: ； Among them, ReportPenalty represents the reporting penalty score, and ReporCount represents the number of reports.

6. The method according to claim 1, wherein The method further comprises: For any comment, a pre-stored first exposure rate threshold is assigned to the comment; if the publishing duration of the comment reaches the first preset duration threshold, determining whether the click-through rate is less than the first preset click-through rate threshold; if so, determining the pre-stored first exposure rate threshold as the exposure rate of the comment; If the click rate is greater than or equal to the first preset click rate threshold and less than a pre-stored second click rate threshold, the difference between the product of the first preset exposure coefficient and the click rate and the second preset exposure coefficient is determined as the exposure rate of the comment; If the click rate is greater than or equal to the second click rate threshold, the pre-stored second exposure rate threshold is determined as the exposure of the comment, and the first exposure rate threshold is less than the second exposure rate threshold.

7. The method according to claim 1, characterized in that The formula for determining the penalty coefficient for dominating the charts based on the consecutive days of dominating the charts is: P=max(0.5, 1−0.1×TopDays); Among them, P represents the penalty coefficient for dominating the chart, and TopDays represents the number of consecutive days of dominating the chart.

8. A device for sorting comments, characterized in that: The device comprises: An acquisition module is used to obtain basic comment data for each comment, including the posting time, number of likes, number of replies, user level, percentage of historical high-quality comments, content information, number of reports, and click-through rate; A first determination module is configured to determine, for any comment, a first dynamic quality score of the comment based on the number of likes, the number of replies, the user level, the proportion of historical high-quality comments, the content information, the number of reports, and the click-through rate; The second determination module is configured to determine, for any comment, an attenuation coefficient of the comment based on the publishing time and the content information, and determine the product of the attenuation coefficient and the first dynamic quality score as the second dynamic quality score of the comment after attenuation; the content information includes the content format, and when the content format does not include a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is: ； Where Decay represents the decay coefficient, and T represents the release duration; When the content format includes a video, the formula for determining the attenuation coefficient of the comment based on the publishing time and the content information is: ； Where Decay represents the decay coefficient, and T represents the release duration; A third determination module is configured to determine, for any comment, a ranking dominance penalty coefficient based on the number of consecutive ranking days detected to have exceeded a preset ranking threshold, and to multiply the ranking dominance penalty coefficient by the second dynamic quality score as a third dynamic quality score after the penalty; the consecutive ranking days being the number of consecutive days the comment was ranked above the preset ranking threshold; The sorting module is used to sort the comments according to the third dynamic quality score and display the comments to the user in order.

Images