Content recommendation methods and devices

Abstract

This disclosure relates to a content recommendation method and apparatus. The method includes: obtaining a user preference score for each tag based on the user's positive behaviors for the day, the weight scores corresponding to each positive behavior, and the implicit negative feedback coefficients corresponding to each positive behavior; obtaining a historical preference score for each tag requiring attribution; obtaining a post-attribution score for each tag requiring attribution based on the unattributed preference score, the attribution coefficient, the historical preference score, and the non-attribution preference score of tags identical to the tags requiring attribution; obtaining a user preference score for each tag based on the post-attribution score and the user preference score of each non-attribution tag different from the tags requiring attribution; and determining the content to be recommended based on the user's preference scores for each tag.

Description

Technical Field

[0001] This disclosure relates to the field of big data technology, and in particular to content recommendation methods and apparatus. Background Technology

[0002] In internet apps such as Xiaohongshu, Taobao, and Weidian, user profiles can help businesses better understand their target users, thereby developing more targeted marketing strategies and product designs, improving user experience and satisfaction, increasing user stickiness and loyalty, and ultimately enhancing the company's market competitiveness and profitability.

[0003] A common method for calculating user profiles is to score users based on the number and weight of their positive behaviors. For example, user preference score = (number of clicks under each tag) * click score + number of likes * like score + number of shares * share score.

[0004] However, this method does not perform preference attribution. For example, if a user clicks on content with multiple tags, it is not known which tag the user clicked because of, thus making the user profile less accurate.

[0005] Therefore, improving the accuracy of user profile calculation is an urgent problem to be solved. Summary of the Invention

[0006] To overcome the problems existing in related technologies, this disclosure provides a content recommendation method and apparatus. The technical solution is as follows:

[0007] According to a first aspect of the present disclosure, a content recommendation method is provided, which obtains a user preference score corresponding to each tag based on each positive behavior of a user on a given day, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior; wherein, the tags include tags that need to be attributed and tags that do not need to be attributed; the user preference score corresponding to the tags that need to be attributed is an unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is an unattributed preference score;

[0008] Obtain the historical preference score corresponding to each of the labels that need to be attributed;

[0009] The attribution score for each label requiring attribution is obtained based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that is the same as the label requiring attribution but does not require attribution.

[0010] The user's preference score for each of the tags is obtained based on the preference attribution score corresponding to each tag that needs attribution and the user preference score for each tag that does not need attribution that is different from the tag that needs attribution.

[0011] The content to be recommended is determined based on the preference scores corresponding to each of the user's tags.

[0012] In one embodiment, obtaining the historical preference score corresponding to each of the labels requiring attribution includes:

[0013] Based on the preference attribution scores for each of the aforementioned attributable tags in each day of the preset number of days prior to the current day, the historical preference scores for each of the aforementioned attributable tags on the current day at the preset number of days from the current day, and the decay weight coefficients for each date, the historical preference scores for each of the aforementioned attributable tags are obtained.

[0014] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0015] Attenuation weighting coefficient = 1 / e a*days ;

[0016] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0017] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0018] Attenuation weighting coefficient = a / days;

[0019] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0020] In one embodiment, obtaining the post-attribution score for each label requiring attribution based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that does not require attribution and is identical to the label requiring attribution, includes:

[0021] The preference attribution score for each of the labels requiring attribution is obtained using the following formula:

[0022]

[0023] Where m represents the degree of attribution, and n represents...

[0024] In one embodiment, the implicit negative feedback coefficient is obtained according to the following formula:

[0025] max(0, user positive behavior statistics - the threshold corresponding to the user positive behavior statistics).

[0026] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects:

[0027] According to a second aspect of the present disclosure, a content recommendation apparatus is provided, comprising:

[0028] The first acquisition module is used to acquire the user preference score corresponding to each tag based on the user's positive behaviors on the same day under each tag, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior; wherein, the tags include tags that need to be attributed and tags that do not need to be attributed; the user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attribution preference score;

[0029] The second acquisition module is used to acquire the historical preference score corresponding to each of the labels that need to be attributed;

[0030] The third acquisition module is used to acquire the post-attribution score for each label that needs to be attributed based on the non-attribution preference score corresponding to the label that needs to be attributed, the attribution coefficient corresponding to the label that needs to be attributed, the historical preference score corresponding to the label that needs to be attributed, and the non-attribution preference score corresponding to the label that is the same as the label that needs to be attributed and does not need to be attributed.

[0031] The fourth acquisition module is used to acquire the user's preference score for each of the tags based on the preference attribution score corresponding to each of the tags that need to be attributed and the user preference score for each of the tags that do not need to be attributed that is different from the tags that need to be attributed.

[0032] The determination module is used to determine the content to be recommended based on the user's preference scores corresponding to each of the user's tags.

[0033] In one embodiment, the second acquisition module includes:

[0034] The first acquisition submodule is used to acquire the historical preference score corresponding to each of the tags to be attributed based on the preference attribution score corresponding to each of the tags to be attributed in each day of a preset number of days prior to the current day, the historical preference score corresponding to each of the tags to be attributed on the current day of a preset number of days prior to the current day, and the decay weight coefficient corresponding to each date.

[0035] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0036] Attenuation weighting coefficient = 1 / e a*days ;

[0037] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0038] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0039] Attenuation weighting coefficient = a / days;

[0040] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0041] In one embodiment, the third acquisition module includes:

[0042] The second acquisition submodule is used to obtain the preference attribution score corresponding to each of the labels that need to be attributed according to the following formula:

[0043]

[0044] Where m represents the degree of attribution, and n represents...

[0045] In one embodiment, the implicit negative feedback coefficient is obtained according to the following formula:

[0046] max(0, user positive behavior statistics - the threshold corresponding to the user positive behavior statistics).

[0047] According to a third aspect of the present disclosure, a content recommendation apparatus is provided, comprising:

[0048] processor;

[0049] Memory used to store processor-executable instructions;

[0050] The processor is configured as follows:

[0051] The user preference score for each tag is obtained based on the user's positive behaviors on the same day, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior. The tags include tags that need to be attributed and tags that do not need to be attributed. The user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attribution preference score.

[0052] Obtain the historical preference score corresponding to each of the labels that need to be attributed;

[0053] The attribution score for each label requiring attribution is obtained based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that is the same as the label requiring attribution but does not require attribution.

[0054] The user's preference score for each of the tags is obtained based on the preference attribution score corresponding to each tag that needs attribution and the user preference score for each tag that does not need attribution that is different from the tag that needs attribution.

[0055] The content to be recommended is determined based on the preference scores corresponding to each of the user's tags.

[0056] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method described in any of the first aspects.

[0057] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0059] Figure 1 This is a flowchart illustrating a content recommendation method according to an exemplary embodiment.

[0060] Figure 2 This is a block diagram illustrating a content recommendation apparatus according to an exemplary embodiment.

[0061] Figure 3 This is a block diagram illustrating a content recommendation device 80 according to an exemplary embodiment. Detailed Implementation

[0062] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0063] Figure 1 This is a flowchart illustrating a content recommendation method according to an exemplary embodiment, such as... Figure 1As shown, the method includes the following steps S101-S105:

[0064] In step S101, the user preference score for each tag is obtained based on the user's positive behaviors for the day, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior. The tags include tags that need to be attributed and tags that do not need to be attributed. The user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attributed preference score.

[0065] The relevant technologies do not consider implicit negative feedback: for example, a user might be exposed to "movie" 10,000 times but only click on it 3 times, while being exposed to "idol worship" 2 times and clicking on it 2 times. In the scoring calculation based on these technologies, "movie" receives a higher user preference score, which is unreasonable. The user's 9,998 exposures without clicks can actually be seen as a form of implicit negative feedback.

[0066] To reduce bias calculation errors, this disclosure considers the case of implicit negative feedback.

[0067] The implicit negative feedback coefficient can be obtained using the following formula:

[0068] max(0, user positive behavior statistics - threshold corresponding to user positive behavior statistics).

[0069] If the user's positive behavior statistics are click-through rate (CTR), then the threshold corresponding to the user's positive behavior statistics is the same as the threshold corresponding to the CTR; if the user's positive behavior statistics are like rate (FOLLOW), then the threshold corresponding to the user's positive behavior statistics is the same as the threshold corresponding to the like rate.

[0070] This disclosure first calculates user preference scores, taking into account implicit negative feedback coefficients such as h, i, and j, and uses positive user behaviors such as clicking, liking, and sharing as examples for illustration.

[0071] The calculation formula at this time is:

[0072] User preference score = (Number of clicks for each tag on that day) * Click score * h + Number of likes * Like score * i + Number of shares * Share score * j.

[0073] Among them, clicks are weighted by the click behavior, likes are weighted by the like behavior, and shares are weighted by the share behavior.

[0074] Here is the general calculation logic for h, i, and c (using h and i as an example; others are similar and will be omitted):

[0075] h = max(0, click rate – 1 / 100);

[0076] i = max(0, likes rate – 1 / 1000);

[0077] This means that, assuming the click-through rate threshold is 1 / 100, if a user is exposed to the content under the "movie" tag 100 times but only clicks it once or less, it can be assumed that the user does not like "movie," and this tag will not be included in subsequent calculations (the value of max in the formula is 0). The threshold value of 1 / 1000 for the number of likes in the formula is because the act of liking is much less frequent than clicking, hence the lower threshold.

[0078] Note that attribution is only necessary when the user's positive behavior content has multiple tags; attribution is not required if it has a single tag. In this case, the scores for single-tag content and multi-tag content need to be calculated separately.

[0079] The final calculation results are as follows; the next step is to perform attribution expansion calculations.

[0080] User preferences before attribution for the day: {Movies (unattributed): 0.3, Idols (unattributed): 0.1, Movies (single label, no attribution required): 0.2, Idols: 0.1, Food: 0.3}.

[0081] If the user exhibits other positive behaviors, then the formula for calculating the user preference score can be expressed as:

[0082] User preference score = (Number of clicks for each tag on that day) * Click score * h + Number of likes * Like score * i + Number of shares * Share score * j + Other positive behaviors...

[0083] In step S102, the historical preference score corresponding to each label that needs to be attributed is obtained.

[0084] In one embodiment, obtaining the historical preference score corresponding to each label that needs to be attributed includes:

[0085] Based on the preference attribution scores for each attributable label in each day of the preset number of days prior to the current day, the historical preference scores for each attributable label on the current day of the preset number of days prior to the current day, and the decay weight coefficient for each date, the historical preference scores for each attributable label are obtained.

[0086] For example, if the current day is June 4th and the preset number of days is 3, then the preset number of days since the current day is 0 days (6.4 days), 1 day (6.3 days), 2 days (6.2 days), and 3 days (6.1 days). In other words, the historical preference scores for each attributed label on the current day are the same as the historical preference scores for each attributed label on the current day (6.1 days).

[0087] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0088] Attenuation weighting coefficient = 1 / e a*days ;

[0089] Where days represents the number of days since today; a represents the exponent of how fast time decays.

[0090] Specifically, 'a' can also be considered a hyperparameter, determined based on specific business scenarios, used to control the rate of time decay. Assuming a = 0.5, then the decay weight coefficient for day T-1 (T is today, T-1 is 1 day from today) = 1 / e^(0.5*1) = 0.607, the decay weight coefficient for day T-2 (T is today, T-2 is 2 days from today) = 1 / e^(0.5*2) = 0.368, and so on...

[0091] In another embodiment, the expression for the attenuation weighting coefficient corresponding to each date is:

[0092] Attenuation weighting coefficient = a / days;

[0093] Where days represents the number of days since today; a represents the exponent of how fast time decays.

[0094] In this disclosure, the time decay considered is based on the assumption that the closer a user's positive action is to the present, the higher the user's action weight score should be. For example:

[0095] Attenuation weighting coefficient = 1 / e a*days ;

[0096] Here, `days` refers to the number of days since the positive action occurred; `e` is the natural logarithm = 2.718...; `a` is a hyperparameter used to control the rate of time decay. A simpler formula, such as `a / days`, can also be used.

[0097] Because related technologies do not consider the decay of user behavior over time, for example, if a user clicked "movie" 10 times a year ago and "idol" 10 times yesterday, theoretically the user should currently have a higher preference for "idol". Therefore, the scoring results are not accurate. This invention, however, takes into account the decay of user behavior over time, which not only makes the calculation results more reasonable, but also more accurately describes the process of changes in user preferences.

[0098] The historical preference score for each user's attributable label is the cumulative score calculated by applying time decay to the daily preference scores. Note that performing time decay calculations on the preferences for each historical day every day would be computationally and storage-intensive, and unnecessary.

[0099] Here, a threshold of N days can be set (i.e., the preset number of days mentioned above). Before the threshold, the cumulative score only needs to be calculated once. Therefore, data from N days ago and earlier only needs to be saved as a single value. Data from N-1 days ago needs to be decayed daily.

[0100] The actual calculation period N can be chosen as one week or one month, but for the sake of simplicity, let's assume N=3 and the attenuation weighting coefficient is 1 / day. Then the attenuation coefficients for today, yesterday, the day before yesterday, and previous days are 1, 0.5, and 0.33 respectively. Using the specific figures in Table 1 as an example: we only need to retain the data from the most recent N days for calculation, and then add the cumulative data to the attribution preference attenuation for N-1 days. The scores calculated here are all after the preference attribution for the current day.

[0101] Table 1

[0102]

[0103]

[0104] Assume the user's preference attribution score for the movie tag on June 1st is 0.6; the user's preference attribution score for the idol tag on June 2nd is 0.2; the user's preference attribution score for the movie tag on June 3rd is 0.8; the user's preference attribution score for the idol tag on June 4th is 0.5; and the user did not take any action on June 5th.

[0105] Therefore, when calculating the historical preference score for each tag on June 1st, we need to use the attribution scores for each tag on June 1st and May 31st, as well as the historical preference score for each tag on May 30th. Specifically, the attribution weighting coefficient for the attribution scores for each tag on June 1st is 1, the attribution weighting coefficient for the attribution scores for each tag on May 31st is 0.5, and the attribution weighting coefficient for the historical preference score for each tag on May 30th is 0.33. Since we only have data for June 1st, the historical preference score for the user on June 1st is: Movies: 0.6 × 1 = 0.6.

[0106] Similarly, when calculating the historical preference scores for each tag on June 4th, we need to use the attribution scores for each tag on June 4th and June 3rd, and the historical preference scores for each tag on June 2nd. The attribution weighting coefficient for each tag's preference score on June 4th is 1, for June 3rd it is 0.5, and for June 2nd it is 0.33. Therefore, on June 4th, the attribution score for idol worship is 0.5 × 1 = 0.5; on June 3rd, the attribution score for movies is 0.8 × 0.5 = 0.4; and on June 2nd, the historical preference score for idol worship is 0.2 × 0.33 = 0.066, and the historical preference score for movies is 0.3 × 0.33 = 0.099.

[0107] User's historical preference score on June 4th = 1 * the preference attribution score on June 4th + 0.5 * the preference attribution score on June 3rd + 0.33 * the historical preference score on June 2nd;

[0108] Therefore, the historical preferences corresponding to the labels that need to be attributed for June 4th are as follows: Movies: 0.4 + 0.099 = 0.499; Idol worship: 0.5 + 0.066 = 0.566.

[0109] User's historical preference score over 6.5 days = 1 * the preference attribution score on 6.5 days + 0.5 * the preference attribution score on 6.4 days + 0.33 * the historical preference score on 6.3 days.

[0110] The calculation method for other dates is the same as described above, and will not be repeated here.

[0111] In step S103, the post-attribution score for each label requiring attribution is obtained by taking the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that does not require attribution and is the same as the label requiring attribution.

[0112] The preference attribution score for each label that needs to be attributed can be obtained using the following formula:

[0113]

[0114] Where m represents the degree of attribution, and n is used to indicate the stability of the data magnitude.

[0115] This disclosure uses user history to attribute preferences to multi-tag content, resulting in more accurate user preferences.

[0116] The attribution in this disclosure is based on the following assumption: if a user's positive behavior today involves multi-label content such as "movie" + "idol worship", then based on the user's historical preference score, if the movie score is high, then the user's behavior today is more likely to be attributed to the movie; if it is a single label, then attribution is not necessary.

[0117] Note that a user's historical cumulative preference score is an accumulation based on the daily preference calculation results after time decay (see step S102 for details), and needs to be distinguished from the user's preference on the current day. These are referred to as historical preference and current day preference.

[0118] Suppose a user's historical preference score on June 3rd is: {Movies: 0.8; Idol worship: 0.2}.

[0119] Prior to attribution preference on June 4th: {Movies (unattributed): 0.3, Idol worship (unattributed): 0.1, Movies (single label, no attribution required): 0.2, Idol worship: 0.1, Food: 0.3}.

[0120] Based on the above assumptions, this disclosure provides a general formula:

[0121]

[0122] Here, 'm' is a hyperparameter greater than 0, representing the degree of attribution, and ensuring that the score is not zero when the same tag has not been clicked in the past. 'n' is another hyperparameter greater than 0, used for data standardization to ensure data stability. Specific values ​​and formulas can be adjusted according to the business scenario.

[0123] Assume m = 0.1; n = 1. Then:

[0124] The preference attribution score for the movie is 0.3*(0.1+0.8) / 1+0.2=0.47;

[0125] The post-attribution score for the preference related to idol worship = 0.1*(0.1+0.2) / 1+0.1 = 0.13;

[0126] Since food is a single-label food, no attribution is needed.

[0127] In step S104, the user's preference score for each label is obtained based on the preference attribution score corresponding to each label that needs attribution and the user preference score for each label that does not need attribution that is different from the label that needs attribution.

[0128] Based on the example above, the preference scores for June 4th are: {Movies: 0.47, Idol worship: 0.13, Food: 0.3}.

[0129] In step S105, the content to be recommended is determined based on the preference scores corresponding to each user's tags.

[0130] Based on the above preference scores, corresponding content can be recommended to users. For example, since the preference score for movies is 0.47, which is higher than the other two, content related to movies will be recommended to users.

[0131] This disclosure first presents a daily preference calculation considering implicit negative feedback coefficients; then, it performs attribution calculations on daily preferences based on historical preferences; finally, it provides the calculation logic for user preference time decay. Once user preferences are obtained, they can be used for subsequent precise operations, recommendation, and other business scenarios.

[0132] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0133] Figure 2 This is a block diagram illustrating a content recommendation device according to an exemplary embodiment. This device can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 2 As shown, the content recommendation device includes:

[0134] The first acquisition module 11 is used to acquire the user preference score corresponding to each tag based on the user's positive behaviors on the same day under each tag, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior; wherein, the tags include tags that need to be attributed and tags that do not need to be attributed; the user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attribution preference score;

[0135] The second acquisition module 12 is used to acquire the historical preference score corresponding to each of the labels that need to be attributed;

[0136] The third acquisition module 13 is used to acquire the post-attribution score of each label that needs to be attributed based on the non-attribution preference score corresponding to the label that needs to be attributed, the attribution coefficient corresponding to the label that needs to be attributed, the historical preference score corresponding to the label that needs to be attributed, and the non-attribution preference score corresponding to the label that is the same as the label that needs to be attributed.

[0137] The fourth acquisition module 14 is used to acquire the user's preference score for each of the tags based on the preference attribution score corresponding to each of the tags that need to be attributed and the user preference score for each of the tags that do not need to be attributed that is different from the tags that need to be attributed.

[0138] The determination module 15 is used to determine the content to be recommended based on the user's preference scores corresponding to each of the tags.

[0139] In one embodiment, the second acquisition module includes:

[0140] The first acquisition submodule is used to acquire the historical preference score corresponding to each of the tags to be attributed based on the preference attribution score corresponding to each of the tags to be attributed in each day of a preset number of days prior to the current day, the historical preference score corresponding to each of the tags to be attributed on the current day of a preset number of days prior to the current day, and the decay weight coefficient corresponding to each date.

[0141] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0142] Attenuation weighting coefficient = 1 / e a*days ;

[0143] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0144] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0145] Attenuation weighting coefficient = a / days;

[0146] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0147] In one embodiment, the third acquisition module includes:

[0148] The second acquisition submodule is used to obtain the preference attribution score corresponding to each of the labels that need to be attributed according to the following formula:

[0149]

[0150] Where m represents the degree of attribution, and n represents...

[0151] In one embodiment, the implicit negative feedback coefficient is obtained according to the following formula:

[0152] max(0, user positive behavior statistics - the threshold corresponding to the user positive behavior statistics).

[0153] According to a third aspect of the present disclosure, a content recommendation apparatus is provided, comprising:

[0154] processor;

[0155] Memory used to store processor-executable instructions;

[0156] The processor is configured as follows:

[0157] The user preference score for each tag is obtained based on the user's positive behaviors on the same day, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior. The tags include tags that need to be attributed and tags that do not need to be attributed. The user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attribution preference score.

[0158] Obtain the historical preference score corresponding to each of the labels that need to be attributed;

[0159] The attribution score for each label requiring attribution is obtained based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that is the same as the label requiring attribution but does not require attribution.

[0160] The user's preference score for each of the tags is obtained based on the preference attribution score corresponding to each tag that needs attribution and the user preference score for each tag that does not need attribution that is different from the tag that needs attribution.

[0161] The content to be recommended is determined based on the preference scores corresponding to each of the user's tags.

[0162] The processor described above can also be configured as:

[0163] In one embodiment, obtaining the historical preference score corresponding to each of the labels requiring attribution includes:

[0164] Based on the preference attribution scores for each of the aforementioned attributable tags in each day of the preset number of days prior to the current day, the historical preference scores for each of the aforementioned attributable tags on the current day at the preset number of days from the current day, and the decay weight coefficients for each date, the historical preference scores for each of the aforementioned attributable tags are obtained.

[0165] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0166] Attenuation weighting coefficient = 1 / e a*days ;

[0167] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0168] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0169] Attenuation weighting coefficient = a / days;

[0170] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0171] In one embodiment, obtaining the post-attribution score for each label requiring attribution based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that does not require attribution and is identical to the label requiring attribution, includes:

[0172] The preference attribution score for each of the labels requiring attribution is obtained using the following formula:

[0173]

[0174] Where m represents the degree of attribution, and n represents...

[0175] In one embodiment, the implicit negative feedback coefficient is obtained according to the following formula:

[0176] max(0, user positive behavior statistics - the threshold corresponding to the user positive behavior statistics).

[0177] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0178] Figure 3 This is a block diagram illustrating a content recommendation device 80 according to an exemplary embodiment, which is suitable for terminal devices. For example, device 80 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0179] The device 80 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0180] Processing component 802 typically controls the overall operation of device 80, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 802 may include one or more modules to facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

[0181] Memory 804 is configured to store various types of data to support the operation of device 80. Examples of such data include instructions for any application or method operating on device 80, contact data, phonebook data, messages, pictures, videos, etc. Memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0182] Power supply component 806 provides power to various components of device 80. Power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to device 80.

[0183] Multimedia component 808 includes a screen that provides an output interface between the device 80 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 808 includes a front-facing camera and / or a rear-facing camera. When the device 80 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0184] Audio component 810 is configured to output and / or input audio signals. For example, audio component 810 includes a microphone (MIC) configured to receive external audio signals when device 80 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 804 or transmitted via communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

[0185] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0186] Sensor assembly 814 includes one or more sensors for providing status assessments of various aspects of device 80. For example, sensor assembly 814 may detect the on / off state of device 80, the relative positioning of components such as the display and keypad of device 80, changes in the position of device 80 or a component of device 80, the presence or absence of user contact with device 80, the orientation or acceleration / deceleration of device 80, and temperature changes of device 80. Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 814 may also include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, or a temperature sensor.

[0187] Communication component 816 is configured to facilitate wired or wireless communication between device 80 and other devices. Device 80 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 816 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0188] In an exemplary embodiment, device 80 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0189] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, which can be executed by a processor 820 of the device 80 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0190] A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of a device 80, enables the device 80 to perform the aforementioned content recommendation method, the method comprising:

[0191] The user preference score for each tag is obtained based on the user's positive behaviors on the same day, the weight score corresponding to each positive behavior, and the implicit negative feedback coefficient corresponding to each positive behavior. The tags include tags that need to be attributed and tags that do not need to be attributed. The user preference score corresponding to the tags that need to be attributed is the unattributed preference score, and the user preference score corresponding to the tags that do not need to be attributed is the no-attribution preference score.

[0192] Obtain the historical preference score corresponding to each of the labels that need to be attributed;

[0193] The attribution score for each label requiring attribution is obtained based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that is the same as the label requiring attribution but does not require attribution.

[0194] The user's preference score for each of the tags is obtained based on the preference attribution score corresponding to each tag that needs attribution and the user preference score for each tag that does not need attribution that is different from the tag that needs attribution.

[0195] The content to be recommended is determined based on the preference scores corresponding to each of the user's tags.

[0196] In one embodiment, obtaining the historical preference score corresponding to each of the labels requiring attribution includes:

[0197] Based on the preference attribution scores for each of the aforementioned attributable tags in each day of the preset number of days prior to the current day, the historical preference scores for each of the aforementioned attributable tags on the current day at the preset number of days from the current day, and the decay weight coefficients for each date, the historical preference scores for each of the aforementioned attributable tags are obtained.

[0198] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0199] Attenuation weighting coefficient = 1 / e a*days ;

[0200] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0201] In one embodiment, the expression for the attenuation weight coefficient corresponding to each date is:

[0202] Attenuation weighting coefficient = a / days;

[0203] Wherein, days represents the number of days since today; and a represents the exponent of the rate of time decay.

[0204] In one embodiment, obtaining the post-attribution score for each label requiring attribution based on the non-attribution preference score corresponding to the label requiring attribution, the attribution coefficient corresponding to the label requiring attribution, the historical preference score corresponding to the label requiring attribution, and the non-attribution preference score corresponding to the label that does not require attribution and is identical to the label requiring attribution, includes:

[0205] The preference attribution score for each of the labels requiring attribution is obtained using the following formula:

[0206]

[0207] Where m represents the degree of attribution, and n represents...

[0208] In one embodiment, the implicit negative feedback coefficient is obtained according to the following formula:

[0209] max(0, user positive behavior statistics - the threshold corresponding to the user positive behavior statistics).

[0210] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0211] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A content recommendation method characterized by, The method comprises the following steps: obtaining a user preference score corresponding to each label according to each positive behavior of the user on the day, a weight score corresponding to each positive behavior and an implicit negative feedback coefficient corresponding to each positive behavior; wherein the labels include labels that need to be attributed and labels that do not need to be attributed; the user preference score corresponding to the label that needs to be attributed is an unattributed preference score, and the user preference score corresponding to the label that does not need to be attributed is an unattributed preference score; obtaining a historical preference score corresponding to each label that needs to be attributed, comprising: obtaining a historical preference score corresponding to each label that needs to be attributed according to a preference attribution score after attribution of each label that needs to be attributed on each day within a preset number of days before the day, a historical preference score corresponding to each label that needs to be attributed on the day that is a preset number of days away from the day, and a decay weight coefficient corresponding to each date; obtaining a preference attribution score after attribution of each label that needs to be attributed according to the unattributed preference score corresponding to the label that needs to be attributed, an attribution coefficient corresponding to the label that needs to be attributed, the historical preference score corresponding to the label that needs to be attributed and the unattributed preference score corresponding to the label that does not need to be attributed and is the same as the label that needs to be attributed, comprising: obtaining a preference attribution score after attribution of each label that needs to be attributed according to the following formula: wherein m represents the degree of attribution and n is used for the order of magnitude stabilization of the data; obtaining a preference score corresponding to each label of the user according to the preference attribution score after attribution of each label that needs to be attributed and the user preference score of each label that does not need to be attributed and is different from the label that needs to be attributed; determining the content to be recommended according to the preference score corresponding to each label of the user.

2. The method of claim 1, wherein, The expression of the decay weight coefficient corresponding to each date is: decay weight coefficient = 1 / ea*days; wherein the days represent the number of days away from today, and the a represents a time decay speed index.

3. The method of claim 1, wherein, The expression of the decay weight coefficient corresponding to each date is: decay weight coefficient = a / days; wherein the days represent the number of days away from today, and the a represents a time decay speed index.

4. The method of claim 1, wherein, The implicit negative feedback coefficient is obtained according to the following formula: max(0, user positive behavior statistical data - threshold value corresponding to the user positive behavior statistical data).

5. A content recommendation apparatus characterized by comprising: The method comprises the following steps: The first obtaining module is configured to obtain a user preference score corresponding to each label according to each positive behavior of the user on the day, a weight score corresponding to each positive behavior and an implicit negative feedback coefficient corresponding to each positive behavior; wherein the labels include labels that need to be attributed and labels that do not need to be attributed; the user preference score corresponding to the label that needs to be attributed is an unattributed preference score, and the user preference score corresponding to the label that does not need to be attributed is an unattributed preference score; The second obtaining module is configured to obtain a historical preference score corresponding to each label that needs to be attributed; The second obtaining module comprises a first obtaining submodule configured to obtain a historical preference score corresponding to each label that needs to be attributed according to a preference attribution score after attribution of each label that needs to be attributed on each day within a preset number of days before the day, a historical preference score corresponding to each label that needs to be attributed on the day that is a preset number of days away from the day, and a decay weight coefficient corresponding to each date; ​ ​ The historical preference score corresponding to each of the labels requiring attribution is obtained according to a post-attribution preference score corresponding to each of the labels requiring attribution, a historical preference score corresponding to each of the labels requiring attribution on a day that is a preset number of days away from the current day, and a decay weight coefficient corresponding to each date. The third obtaining module is configured to obtain the post-attribution preference score corresponding to each of the labels requiring attribution according to the un-attributed preference score corresponding to each of the labels requiring attribution, the attribution coefficient corresponding to each of the labels requiring attribution, the historical preference score corresponding to each of the labels requiring attribution, and the un-attributed preference score corresponding to each of the labels not requiring attribution that are the same as the labels requiring attribution, and specifically configured to obtain the post-attribution preference score corresponding to each of the labels requiring attribution according to the following formula: wherein m indicates the degree of attribution and n is used for the magnitude stabilization of data; The fourth obtaining module is configured to obtain the preference score corresponding to each of the labels of the user according to the post-attribution preference score corresponding to each of the labels requiring attribution and the user preference score of each of the labels not requiring attribution that are different from the labels requiring attribution. The determining module is configured to determine the content to be recommended according to the preference score corresponding to each of the labels of the user.

6. A content recommendation apparatus characterized by comprising: The processor is configured to: obtain the user preference score corresponding to each of the labels according to each positive behavior of the user on the current day, a weight score corresponding to each of the positive behaviors, and an implicit negative feedback coefficient corresponding to each of the positive behaviors; the labels include labels requiring attribution and labels not requiring attribution; the user preference score corresponding to the labels requiring attribution is an un-attributed preference score, and the user preference score corresponding to the labels not requiring attribution is an un-attributed preference score; obtain the historical preference score corresponding to each of the labels requiring attribution, including: obtaining the historical preference score corresponding to each of the labels requiring attribution according to a post-attribution preference score corresponding to each of the labels requiring attribution, a historical preference score corresponding to each of the labels requiring attribution on a day that is a preset number of days away from the current day, and a decay weight coefficient corresponding to each date; obtain the post-attribution preference score corresponding to each of the labels requiring attribution according to the un-attributed preference score corresponding to each of the labels requiring attribution, the attribution coefficient corresponding to each of the labels requiring attribution, the historical preference score corresponding to each of the labels requiring attribution, and the un-attributed preference score corresponding to each of the labels not requiring attribution that are the same as the labels requiring attribution, including: obtaining the post-attribution preference score corresponding to each of the labels requiring attribution according to the following formula: obtain the preference score corresponding to each of the labels of the user according to the post-attribution preference score corresponding to each of the labels requiring attribution and the user preference score of each of the labels not requiring attribution that are different from the labels requiring attribution; determine the content to be recommended according to the preference score corresponding to each of the labels of the user. The instructions are executed by the processor to implement the steps of the method of any one of claims 1 to 4. The instructions are executed by the processor to implement the steps of the method of any one of claims 1 to 4. wherein m represents the degree of attribution and n is used for the order of magnitude stabilization of the data; ​ ​ 7. A computer readable storage medium having stored thereon computer instructions, wherein, ​

Images