A content recommendation optimization method and server

Abstract

The present application provides a content recommendation optimization method and server. The provided content recommendation optimization method is applied to the server, including: receiving a media asset recommendation request sent by a terminal device, determining a media asset recommendation list corresponding to the terminal device based on uploaded historical media asset playback records, wherein the media asset recommendation list includes multiple media assets, obtaining historical media asset playback records that have not been uploaded in the terminal device, and if the historical media asset playback records that have not been uploaded exist, re-ordering the media assets in the media asset recommendation list based on the content attributes of the media assets in the historical media asset playback records that have not been uploaded and the media asset recommendation list, and returning the re-ordered media asset recommendation list to the terminal device. The present application solves the problem that the existing media asset recommendation algorithm does not add the most recently generated historical media asset playback records to the offline calculation during the generation of the media asset recommendation list, resulting in the media asset sorting in the pushed media asset recommendation list not meeting the user's interest characteristics, thereby reducing the user's experience.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a content recommendation optimization method and server. Background Art

[0002] Smart TV is an important display device for people to watch TV series, movies and other media. In order to enable people to watch their favorite media, the server of smart TV usually analyzes the media that people may be interested in through some media recommendation algorithms, and then displays the media that people may be interested in on some media recommendation pages of smart TV, so that users can watch them on demand.

[0003] Different people are interested in different media resources. In related technologies, in order to improve the accuracy of media resource recommendations, media resource recommendation algorithms can recommend media resources based on historical media resource playback records on smart TVs over a period of time. Smart devices upload historical media resource playback records over a period of time to the server, and the server determines recommended media resources based on the media resource playback records and sends the recommended media resources to smart TVs for recommendation to users.

[0004] When there are multiple historical media asset playback records within a period of time, the server will often determine multiple different recommended media assets based on the media asset playback records. When there are many recommended media assets, the user needs to further search for content of interest. Summary of the invention

[0005] The present application provides a content recommendation method and server to solve the problem that the existing media recommendation algorithm does not perform offline calculations on the most recently generated historical media playback records during the generation of the media recommendation list, resulting in the media sorting in the pushed media recommendation list not meeting the user's interest characteristics, thereby reducing the user's experience.

[0006] In a first aspect, the present application provides a content recommendation optimization method, which is applied to a server and includes:

[0007] Receiving a media asset recommendation request sent by a terminal device, and determining a media asset recommendation list corresponding to the terminal device according to uploaded historical media asset playback records, wherein the media asset recommendation list includes multiple media assets;

[0008] Obtaining the playback records of the historical media resources not uploaded in the terminal device;

[0009] If the unuploaded historical media asset playback record exists, the media assets in the media asset recommendation list are reordered according to the unuploaded historical media asset playback record and content attributes of the media assets in the media asset recommendation list, and the reordered media asset recommendation list is returned to the terminal device.

[0010] In a second aspect, the present application provides a content recommendation optimization method, which is applied to a terminal device, comprising:

[0011] Send a media resource recommendation request to the server;

[0012] Send the playback records of historical media assets that have not been uploaded to the server;

[0013] The media asset recommendation list returned by the server is received and displayed, wherein the arrangement order of the media assets in the media asset recommendation list is used to determine the display order of the media assets on the terminal device.

[0014] In a third aspect, the present application provides a server, including a memory and a processor, wherein the memory stores program instructions, and the processor executes the following steps by running the program instructions:

[0015] Receiving a media asset recommendation request sent by a terminal device, and determining a media asset recommendation list corresponding to the terminal device according to uploaded historical media asset playback records, wherein the media asset recommendation list includes multiple media assets;

[0016] Obtaining the historical media resource playback records that have not been uploaded in the terminal device;

[0017] If the unuploaded historical media asset playback record exists, the media assets in the media asset recommendation list are reordered according to the unuploaded historical media asset playback record and content attributes of the media assets in the media asset recommendation list, and the reordered media asset recommendation list is returned to the terminal device.

[0018] It can be seen from the above technical solutions that based on the content recommendation method provided by the embodiment of the present application, when the terminal device sends a media asset recommendation request to the server, if the historical media asset playback record that has not been uploaded exists, the media assets in the media asset recommendation list are reordered according to the historical media asset playback record that has not been uploaded and the content attributes of the media assets in the media asset recommendation list, and the reordered media asset recommendation list is returned to the terminal device. Thus, by obtaining the historical media asset playback records that are not included in the offline calculation in the terminal device, the user characteristics are calculated, and the original order of the media assets in the media asset recommendation list issued by the server is reordered. At the same time, according to the real-time media asset on-demand record of the current user and the similarity between the media assets, the arrangement order of the media assets is adjusted, which is conducive to enabling the current user to see the media assets that he is interested in at the front of the arrangement, thereby achieving accurate media asset recommendations and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 The following is an exemplary illustration of the use scenario of the display device provided in the embodiment of the present application;

[0021] Figure 2 An exemplary diagram of the interface of the system homepage in the display device is shown;

[0022] Figure 3 The schematic diagram of the interaction process between the display device and the server in the prior art is exemplarily shown;

[0023] Figure 4 A schematic diagram of the interaction process between a display device and a server according to some embodiments is exemplarily shown;

[0024] Figure 5 A schematic diagram of a home page of a display device according to some embodiments is exemplarily shown. DETAILED DESCRIPTION

[0025] In order to make the purpose and implementation method of the present application clearer, the exemplary implementation method of the present application will be clearly and completely described below in conjunction with the drawings in the exemplary embodiments of the present application. Obviously, the described exemplary embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0026] It should be noted that the brief description of terms in this application is only for the convenience of understanding the embodiments described below, and is not intended to limit the embodiments of this application. Unless otherwise specified, these terms should be understood according to their common and usual meanings.

[0027] The terms "first", "second", "third", etc. in the specification and claims of this application and the above drawings are used to distinguish similar or similar objects or entities, and do not necessarily mean to limit a specific order or sequence, unless otherwise noted. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances.

[0028] The terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover but not exclude inclusion, for example, a product or device comprising a list of components is not necessarily limited to all the components expressly listed but may include other components not expressly listed or inherent to such product or device.

[0029] The term "module" refers to any known or later developed hardware, software, firmware, artificial intelligence, fuzzy logic, or combination of hardware and / or software code that is capable of performing the functions associated with that element.

[0030] The terminal device in the embodiment of the present application may include any one of a mobile phone, a tablet computer, a wearable device, a laptop computer, and a display device. The present application will be described below using a display device as an example.

[0031] Figure 1 Schematic diagram of a usage scenario of a display device according to an embodiment. Figure 1 As shown in FIG. 1 , a user can operate the display device 200 through the smart device 300 and the control apparatus 100 .

[0032] In some embodiments, the control device 100 may be a remote controller, and the communication between the remote controller and the display device includes infrared protocol communication or Bluetooth protocol communication, and other short-range communication methods, etc., to control the display device 200 wirelessly or in a wired manner. Among them, the wireless method may be a direct connection or an indirect connection, and may be routed or not. The user may input user commands through buttons on the remote controller, voice input, control panel input, etc. to control the display device 200. For example, the user may input corresponding control commands through the volume plus and minus keys, channel control keys, up / down / left / right movement keys, voice input keys, menu keys, power on / off keys, etc. on the remote controller to realize the function of controlling the display device 200.

[0033] In some embodiments, the smart device 300 may also be used to control the display device 200. For example, the smart device 300 may be used to control an application running on the display device 200, or the application running on the smart device 300 may be used to control the display device 200. The application may be configured to provide the user with various controls in an intuitive user interface (UI) on a screen associated with the smart device.

[0034] In some embodiments, the smart device 300 and the display device 200 may also be used to communicate data.

[0035] In some embodiments, the display device 200 can also be controlled in a manner other than the control device 100 and the smart device 300. For example, the user's voice command control can be directly received through a module for obtaining voice commands configured inside the display device 200, or the user's voice command control can be received through a voice control device set outside the display device 200.

[0036] In some embodiments, the display device 200 also communicates data with the server 400. The display device 200 may be allowed to communicate via a local area network (LAN), a wireless local area network (WLAN), and other networks. The server 400 may provide various content and interactions to the display device 200. The server 400 may be one cluster or multiple clusters, and may include one or more types of servers.

[0037] In some embodiments, the software steps executed by a step execution subject can be migrated to another step execution subject in data communication with it as needed for execution. Exemplarily, the software steps executed by a server can be migrated to a display device in data communication with it as needed for execution, and vice versa.

[0038] In some embodiments, see Figure 2 The navigation bar can be set with multiple controls, such as "My", "Channel", "Movies and TV", "Fitness", "VIP", "Education", "Mall", "Games" and Applications. Different navigation bar controls correspond to different interfaces, and different interfaces can display different media resources in the content display area. Figure 2 This is the interface corresponding to the "Channel" control, such as Figure 2 As shown in the figure, the content display area under the "Channel" control can display multiple media asset controls, corresponding to media assets 1 to 10 respectively. After the user clicks a media asset control, he can enter the details page corresponding to the media asset control, or enter the playback interface corresponding to the media asset control. The user can also Figure 2 Press the down button on the remote control to scroll down to view other media asset controls.

[0039] In some embodiments, the media data corresponding to the media asset control is sent from the server to the display device. The server can generate a media asset recommendation list with an arrangement order according to some media asset recommendation algorithms, and then send the media asset data corresponding to the media asset recommendation list to the display device in the arrangement order, so that the display device can display the corresponding media assets. The arrangement order can be represented by the user's interest in the media asset calculated according to the media asset recommendation algorithm. The higher the arrangement order, the more likely the user is to be interested in the media asset, and the display device will display the media asset in a relatively forward position. For example, Figure 2 In the example, media asset 1 is arranged in a front order than media asset 2, and accordingly the display device displays media asset 1 in a relatively front position and displays media asset 2 in a relatively back position.

[0040] For example, see Figure 3 , Figure 3The schematic diagram of the interaction process between the display device and the server in the prior art. Before the user's display device displays the content display area, the media recommendation algorithm needs to train the recommendation sorting model in advance, which needs to be done through media information, user information, and context information, wherein the user information includes historical media playback records, and the similarity of the media to be recommended is calculated in sequence according to the historical media playback records to obtain the corresponding similarity, and the user features corresponding to the user information are calculated at the same time, and the similarity and user features are stored in the server. When a media recommendation request from the display device is received, the corresponding context features are obtained according to the similarity and user features, and the similarity, user features and context features are used as model data for training to obtain a recommendation sorting model. Specifically, the media to be recommended is scored and sorted according to the recommendation sorting model to obtain a media recommendation list, and the media in the media recommendation list is sorted according to the original order, and the media recommendation list is returned to the display device.

[0041] During the process of online real-time reasoning of the model, we need to obtain the features corresponding to the above three types of information of the items to be sorted. Among them, media information is relatively fixed and can be calculated offline in advance. Context information is relatively simple to generate. However, it takes a long time to generate user features in real time based on all historical records of the user, which affects the user experience. The existing technology adopts the method of calculating user features offline in advance, specifically obtaining historical media playback records before a preset time point in advance to calculate user features. For example, the preset time point can be 24 hours ago. When the user needs to display the media in the content display area of ​​the display device, the calculated user features are directly used during online model reasoning, which can shorten the online reasoning time. However, the user features used in the reasoning process of the recommendation ranking model are only obtained by obtaining the historical media playback records before the preset time point. For example, only the historical media playback records before 24 hours are used, and the historical media playback records generated by users clicking on media in the last 24 hours are not included in the offline calculation of user features. It is well known that user interests are constantly changing, and the media that users have watched recently can better reflect the user's interests. The historical media playback records that users have recently generated are discarded, and the accuracy of the media ranking in the calculated media recommendation list is greatly affected and cannot accurately reflect the user's interests. When a user does not find the media that he or she is interested in at the head of the media recommendation list, the user needs to Figure 2 It takes multiple pages to find the media resources that you are interested in, which seriously affects the user experience.

[0042] To solve the above technical problems, an embodiment of the present application provides a content recommendation optimization method, which obtains historical media playback records in the terminal device that are not included in offline calculations, and reorders the media in the media recommendation list sent by the server, so that when the user views the media recommendation list sent by the server, he can find the media he is interested in at the front of the sort, thereby improving the user experience.

[0043] When implementing it specifically, Figure 4 FIG. 1 is a schematic diagram of the interaction process between a display device and a server according to some embodiments of the present application. Figure 4 As shown, the server receives a media asset recommendation request sent by the display device, and the media asset recommendation request is used to request the server to return a media asset recommendation list, which includes multiple media assets with original order, and the original order is calculated offline in advance by the media asset recommendation algorithm based on the historical media asset playback records uploaded to the server by the display device. At this time, the server obtains the historical media asset playback records that have not been uploaded to the server from the display device. If the historical media asset playback records that have not been uploaded exist, the server reorders the media assets in the media asset recommendation list according to the historical media asset playback records that have not been uploaded and the original order of the media assets in the media asset recommendation list, and returns the reordered media asset recommendation list to the terminal device. It should be noted that since the media recommendation algorithm is relatively complex, it is usually necessary to obtain the historical media playback records of the current display device at a preset time in advance, and calculate the original order of the media in the corresponding media recommendation list according to the media recommendation algorithm. When the user chooses to watch the media after obtaining the historical media playback records, the historical media playback records generated at this time have not been uploaded to the server. Therefore, there will be unuploaded historical media playback records in the display device, and these unuploaded historical media playback records are not included in the media recommendation algorithm for calculation.

[0044] For example, the original order of the media assets in the current media asset recommendation list is media asset 1, media asset 2, media asset 3, and media asset 4. When there is no historical media asset playback record uploaded, the server re-sorts the media assets in the media asset recommendation list to media asset 3, media asset 2, media asset 4, and media asset 1. The server sends the re-sorted media asset recommendation list to the display device, which displays it to the user. The user can click on the media assets in the media asset recommendation list to select the media assets they want to play. Since the original order of the media assets in the media asset recommendation list is optimized by the media assets that the user has recently clicked, it can better match the user's current interests. The user can click on the media of interest at the head of the media asset recommendation list to play it.

[0045] The above solution solves the problem that in the calculation process of the existing media recommendation algorithm, the media playback records that the user has recently watched are not included in the offline calculation due to the need to obtain the historical media playback records in advance for offline calculation. By obtaining the historical media playback records that are not included in the offline calculation in the display device, the media in the media recommendation list sent by the server are optimized and sorted. At the same time, according to the real-time media on-demand records of the current user and the similarity between the media, the arrangement order of the media is adjusted in time, which is conducive to the current user seeing the media he is interested in at the front of the arrangement, realizing accurate media recommendation and improving user experience.

[0046] In some embodiments, the server calculates the media assets to be sorted and the user features in advance and stores them in the storage server. The user features are calculated by obtaining the historical media asset playback records uploaded to the server by the display device, and the feature values ​​corresponding to the media assets and the feature values ​​corresponding to the user are stored in the server. In the process of online reasoning, the feature values ​​corresponding to the media assets and the current user are obtained, and then the context features are obtained online and input into the recommendation sorting model for training to obtain the score S corresponding to the media assets to be sorted. n(n＝1,2,3.n) , by treating the score S of the sorted media n Sort from high to low, and finally get the original order of the media in the media recommendation list. For example, the similar media and scores corresponding to media 1 can be stored in the form of a map, {item2: 0.92, item3: 0.88, item4: 0.75...}, where S2 = 0.92, S3 = 0.88, S4 = 0.75.

[0047] In some embodiments, historical media asset playback records that have not been uploaded to the server within a preset time range in the display device are obtained. For example, the preset time range may be 24 hours. Since these historical media asset playback records are recently generated by the user, these historical media asset playback records truly reflect the user's current interests and are not included in the user feature calculation. Therefore, the media in the media asset recommendation list that is more similar to the media in the historical media asset playback records is more likely to be accepted by the user. The similarity order of the media in the media asset recommendation list is obtained by calculating the similarity of the media in the media asset recommendation list corresponding to each historical media asset playback record.

[0048] In some embodiments, similarity is used to reflect the degree of similarity between two media assets. The more users who like two media assets at the same time, the higher the value of the similarity between the two media assets. More specifically, the user set data that likes the current media asset is a corresponding user set obtained from the server according to the content attribute of the media asset. When the server receives a media asset recommendation request sent by a display device, it obtains the historical media asset playback records in the display device that have not been uploaded to the server, calculates the similarity corresponding to the media assets in the media asset recommendation list and each historical media asset playback record in turn, and obtains the similarity score of the media assets in the media asset recommendation list after summing and averaging.

[0049] The similarity calculation formula is:

[0050]

[0051] Among them, R n is the similarity value, |N(i)| is the first set of users who like item i, |N(j)| is the second set of users who like item j, and |N(i)|∩|N(j)| is the set of users who like both item i and item j.

[0052] In some embodiments, the similarity is normalized and the value is mapped to [0, 1]. This is to facilitate the comparison of the similarities of different media assets in the media asset recommendation list, sort them from large to small according to the similarity, and save the top N media assets in the media asset recommendation list with the highest similarity corresponding to each media asset in each historical media asset playback record to the storage server. The normalization formula for the similarity of media assets is as follows:

[0053]

[0054] Among them, R n Represents the similarity score, R' n is the normalized score value.

[0055] In some embodiments, when the interface server obtains historical media asset playback records that have not been uploaded from the terminal device, some of the historical media asset playback records may be generated due to erroneous clicks by the user. The historical media asset playback records can also be effectively deleted to determine whether the playback time of the media asset corresponding to the current media asset playback record exceeds a preset time. If it exceeds the preset time, it is used as a valid record to calculate the similarity order. If it does not exceed the preset time, it is deleted from the historical media asset playback records that have not been uploaded as an invalid record. For example, the preset time can be set to 20 minutes. When the user plays media asset 6, the playback time is 1 minute. Although the user clicks on media asset 6 to play, the user only watches it for a short time and then stops playing. Therefore, media asset 6 is not the media asset that the user is interested in. If the historical playback record of media asset 6 is added to the calculation of the re-ordering of the media asset recommendation list, it cannot truly reflect the user's interest, resulting in a decrease in the accuracy of the media asset sorting in the media asset recommendation list, and cannot match the user's viewing interest, resulting in the user having to turn pages multiple times to find the media asset that he is interested in, which seriously affects the user's usage experience.

[0056] In some embodiments, when the historical playback time of media asset 7 is 50 minutes, the preset time can be set to 20 minutes. Since the historical playback time of media asset 7 exceeds the preset time, it means that media asset 7 is content that the user is interested in. Therefore, the historical playback record of media asset 7 is added to the re-ordering calculation of the media asset recommendation list to obtain the corresponding similarity score, thereby improving the accuracy of the sorting of the media asset recommendation list and improving the user experience.

[0057] In some embodiments, after re-sorting, each media asset in the media asset recommendation list has two ranking orders, one of which is the sequence number in the ranking order in the media asset recommendation list generated by the recommendation algorithm. The ranking order can be recorded as the original order, which is the score S calculated according to the media asset recommendation algorithm. n One is to sort based on the similarity order calculated between the historical media playback records that have not been uploaded and the media in the media recommendation list. The similarity order is calculated by the similarity R n Sort the values.

[0058] In some embodiments, the original order is used as the first weight, and the similarity order is used as the second weight. After weighting, a weighted score is obtained. The weighted score is sorted to obtain a re-ordered media asset recommendation list, and then the re-optimized and sorted media asset recommendation list is stored in a storage server, and the media assets in the media asset recommendation list are sent to the display device in the re-ordered order. The server can send the media asset data of the media assets in the media asset recommendation list in the order of arrangement. The calculation formula of the weighted score S can be:

[0059] S=w1*S n +w2*Rn

[0060] Among them, S represents the weighted score, w1 is the weight of the original order of the media resources in the media resource recommendation list, which is the first weight, w2 is the weight of the similarity order, which is the second weight, the sum of w1 and w2 is 1, and w1 and w2 are both greater than or equal to 0.

[0061] In some embodiments, the media asset recommendation list can be obtained based on the user behavior of the current user's corresponding media asset playback or media asset collection, and the user behavior can be obtained through data such as historical media asset playback records and historical media asset collection records. The recommendation algorithm server generates a media asset recommendation list containing multiple media assets that the user is interested in based on the user behavior, wherein each media asset has a corresponding media asset id, and then sorts them from high to low according to the scores corresponding to the media asset id. The higher the score corresponding to the media asset id, the more the media asset fits the user's interest, and the higher the original order of the media asset id corresponding to the media asset. Exemplarily, the media asset recommendation list may include the media asset ids of 20 media assets, and the 20 media assets are sorted in the original order.

[0062] In some embodiments, after generating the media asset recommendation list, the recommendation algorithm server may send the media asset recommendation list to the interface server and the storage server. The interface server may obtain the media asset IDs of the first number of media assets in the media asset recommendation list, and then obtain the media asset data corresponding to these media asset IDs from the storage server, and then obtain the historical media asset playback records that have not been uploaded from the display device, reorder these media assets according to the historical media asset playback records, and send the reordered media asset recommendation list to the display device, wherein, in addition to the media asset data, the homepage data may also include some arrangement data, such as the location and size of the recommendation control and other data.

[0063] In some embodiments, after generating the media asset recommendation data, the recommendation algorithm server may also only send the media asset recommendation data to the storage server. The interface server may obtain the media asset recommendation list from the storage server, and then obtain the unuploaded historical media asset playback records from the display device, reorder the media assets in the media asset recommendation list according to the historical media asset playback records, and send the reordered media asset recommendation list to the display device.

[0064] In some embodiments, after receiving the media asset recommendation list, the display device may display the media asset recommendation list on the homepage, the number of media assets in the media asset recommendation list is a first number, the value of the first number may be N, N is greater than or equal to 1, such as Figure 2 As shown, the first number may be 10, and the display device may display 10 media assets, namely, media asset 1 to media asset 10, on the home page. Of course, the first number may also be other values, such as 15. If the first number is 15, the display device may display media asset 1 to media asset 15 on the home page.

[0065] In some embodiments, after the display device displays a first number of media assets, there may be media assets that the user is interested in, and the user may select one or more media assets from these media assets to watch. Each time the user orders a media asset, the display device may send a piece of media asset click information to the interface server, including click information indicating that the user clicked on the media asset, as well as information such as the order time and exit time.

[0066] In some embodiments, the user Figure 2 After playing a few media resources, the user may not be interested in the remaining media resources on the homepage. The user can press the down button of the remote control to turn the page to refresh the media resource recommendation page. The display device regenerates the media resource recommendation request based on the trigger signal received from the down button and the current display interface is the homepage interface, and then sends the media resource recommendation request to the interface server again. The media resource recommendation request is used to request the server to return a media resource recommendation list, which includes multiple media resources.

[0067] In some embodiments, after the user has played all the first number of media assets displayed on the homepage, the display device may also generate a media asset recommendation request based on the generated click information to request a media asset recommendation list from the interface server again. The media asset recommendation list includes multiple media assets, and the arrangement order of the media assets is obtained by optimizing the original order of historical media asset playback records within a preset time range, and the original order is calculated offline in advance by a media asset recommendation algorithm based on historical media asset playback records uploaded to the server by the display device.

[0068] In some embodiments, after the display device displays the media assets in the media asset recommendation list, the user can select one or more media assets from these media assets to watch, wherein each time the user clicks on a media asset, the display device can send a piece of media asset click information to the interface server, indicating the user click information after the server sends the media asset recommendation list. The server adjusts the first weight and the second weight by obtaining the media asset click information within a preset time, and the media asset click information includes click information indicating that the user clicks on the media asset, and also includes information such as on-demand time and exit time.

[0069] In some embodiments, the media asset click information is also screened, and it is determined in turn whether the media asset playback time corresponding to each piece of media asset click information exceeds a preset threshold. If it exceeds the preset threshold, the piece of media asset click information is retained and uploaded to the server; if it does not exceed the preset threshold, the piece of media asset click information is deleted.

[0070] In some embodiments, after receiving a certain number of media asset click information within a preset time range, the server can adjust the weight value according to the original sequence number and the reordered sequence number of the media asset id corresponding to the currently clicked media asset in turn. If the reordered sequence number of the media asset id corresponding to the current media asset click information is less than the original sequence number, then this click is used as a corrected positive feedback. If the reordered sequence number is greater than the original sequence number, then this click is used as a corrected negative feedback. According to multiple media asset click information within a preset time range, multiple user feedbacks can be obtained. Finally, the corrected positive feedback within the preset time range minus the number of corrected negative feedbacks is used as the evaluation parameter c of the correction effect. If c>0, the first weight is correspondingly reduced and the second weight is increased. If c<0, the first weight is correspondingly increased and the second weight is reduced, and the adjusted values ​​of the first weight and the second weight are stored in the server. For example, when a user clicks on media asset 14 to play, media asset click information is generated according to the reordered sequence number of media asset 14, and the media asset click information is sent to the interface server. If the order of media asset 14 in the original order is 8 and the order number in the reorder is 3, the user's behavior of clicking on media asset 14 is regarded as a corrected positive feedback. If the order of media asset 14 in the original order is 3 and the order number in the reorder is 8, the user's behavior of clicking on media asset 14 is regarded as a corrected negative feedback.

[0071] For example, within a preset time range, a user clicks to play 10 media assets, generating 10 media asset click information, of which 8 are corrected positive feedbacks and 2 are corrected negative feedbacks. If the evaluation parameter c obtained by subtracting the number of corrected negative feedbacks from the number of corrected positive feedbacks is greater than 0, then the first weight is correspondingly reduced and the second weight is increased. By effectively counting the number of corrected positive feedbacks and corrected positive feedbacks, the weights can be adjusted to facilitate users to find the media assets they are interested in at the top of the ranking, thereby improving the user experience.

[0072] In some embodiments, if the reordered sequence number of the media asset id corresponding to the current media asset click information is equal to the sequence number of the original sequence, the click is regarded as invalid feedback and is not included in the calculation of the evaluation parameter c of the modified positive feedback and the modified negative feedback.

[0073] In some embodiments, when a media resource recommendation request is received from a display device, the media resources in the media resource recommendation list are reordered according to the adjusted values ​​of the first weight and the second weight, a reordered media resource recommendation list is generated, and the media resource recommendation list reordered according to the adjusted first weight and the second weight is sent to the display device.

[0074] In some embodiments, the server reorders the media assets in the media asset recommendation list according to the adjusted values ​​of the first weight and the second weight, generates a reordered media asset recommendation list, and stores the reordered media asset recommendation list in the server. When a media asset recommendation request is received from a display device, the server sends the reordered media asset recommendation list to the display device.

[0075] See also Figure 5 , is a schematic diagram of a home page interface of a display device according to some embodiments, such as Figure 5 As shown, the server reorders the media assets in the media asset recommendation list according to the adjusted first weight and second weight, and the media assets in the resulting media asset recommendation list are ranked as media asset 12, media asset 16, media asset 18, media asset 11, media asset 17, media asset 15, media asset 14, media asset 13, media asset 19 and media asset 20, respectively. Before the reordering, the media asset recommendation list is sorted using the unadjusted first weight and second weight, and the media assets in the resulting media asset recommendation list are ranked as media asset 11-media asset 20, respectively.

[0076] In some embodiments, when the server receives a media resource recommendation request sent by the terminal device again, the server sends the media resource recommendation list reordered according to the adjusted first weight and second weight to the display device. Figure 5 After playing several media resources, you can press the down button of the remote control to turn the page to update the media resources in the media resource recommendation page. The display device generates a media resource recommendation request based on the trigger signal received from the down button and the current display interface is the home page interface, and then sends a media resource refresh request to the interface server. The interface server returns a media resource recommendation list re-sorted according to the adjusted first weight and second weight to the display device based on the media resource recommendation request.

[0077] In some embodiments, after receiving the media asset recommendation list reordered according to the adjusted first weight and second weight, the display device may refresh the homepage of the display device to display the media assets in the media asset recommendation list reordered according to the adjusted first weight and second weight.

[0078] In some embodiments, when there are no historical media asset playback records that have not been uploaded to the server in the display device, the server returns a media asset recommendation list to the display device, and the media asset recommendation list includes multiple media assets in an original order, and the original order is calculated by the media asset recommendation algorithm based on the historical media asset playback records that have been uploaded to the server by the display device. For example, within a preset time range, the user has not watched any media asset, so there is no historical media asset playback record that has not been added to the offline calculation. At this time, the server cannot re-order according to the unuploaded media asset playback records, and the server returns the media asset playback list in the original order to the display device.

[0079] Furthermore, the present application provides a content recommendation method, applied to a terminal device, comprising: sending a media asset recommendation request to a server, sending a historical media asset playback record that has not been uploaded to the server, receiving and displaying a media asset recommendation list returned by the server, wherein the arrangement order of the media assets in the media asset recommendation list is used to determine the display order of the media assets on the terminal device.

[0080] Furthermore, the present application provides a server including a memory and a processor, wherein program instructions are stored in the memory, and the processor executes other program steps including those that the processor in each embodiment of the above-mentioned terminal device is used to execute or is configured to execute by running the program instructions, which are not repeated here.

[0081] It can be seen from the above technical solutions that based on the content recommendation method provided by the embodiment of the present application, when the terminal device sends a media asset recommendation request to the server, if the historical media asset playback record that has not been uploaded exists, the media assets in the media asset recommendation list are reordered according to the historical media asset playback record that has not been uploaded and the content attributes of the media assets in the media asset recommendation list, and the reordered media asset recommendation list is returned to the terminal device. Thus, by obtaining the historical media asset playback records that are not included in the offline calculation in the terminal device, the user characteristics are calculated, and the original order of the media assets in the media asset recommendation list issued by the server is reordered. At the same time, according to the real-time media asset on-demand record of the current user and the similarity between the media assets, the arrangement order of the media assets in the media asset recommendation list is adjusted, which is conducive to enabling the current user to see the media assets that he is interested in at the front of the arrangement, thereby achieving accurate media asset recommendations and improving user experience.

[0082] Since the above embodiments are all described by citing and combining with other embodiments, different embodiments have the same parts, and the same and similar parts between the embodiments in this specification can be referred to each other. No further detailed description is given here.

[0083] In a specific implementation, the present invention further provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment of the content recommendation method provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0084] It should be noted that, in this specification, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such circuit structure, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the circuit structure, article or device including the element.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0086] For the convenience of explanation, the above description has been made in conjunction with specific embodiments. However, the above exemplary discussion is not intended to be exhaustive or limit the embodiments to the specific forms disclosed above. Based on the above teachings, various modifications and variations can be obtained. The selection and description of the above embodiments are to better explain the principles and practical applications, so that those skilled in the art can better use the embodiments and various different variations of the embodiments suitable for specific use considerations.

Claims

1. A content recommendation optimization method, characterized in that: Applied to a server, the server stores content attributes of media assets, including: Perform offline calculation based on uploaded historical media resource playback records to obtain user features; the uploaded historical media resource playback records are data uploaded by the terminal device before performing offline calculation; Receive a media asset recommendation request sent by a terminal device, and determine a media asset recommendation list; wherein the media asset recommendation list includes multiple media assets, and the multiple media assets are sorted in an original order; the original order is calculated based on feature values ​​corresponding to the multiple media assets and feature values ​​corresponding to the user; Obtaining the historical media playback records of the terminal device that have not been uploaded; Determine a first user set according to content attributes of media assets in the historical media asset playback records that have not been uploaded; the first user set is a set of users who like the media assets in the historical media asset playback records that have not been uploaded; Determine a second user set according to the content attributes of the media in the media resource recommendation list; the second user set is a set of users who like the media in the media resource recommendation list; Calculating the similarity between the media assets in the media asset recommendation list and the historical media asset playback records that have not been uploaded according to the intersection of the first user set and the second user set and the first user set and the second user set; The similarities are summed and averaged to obtain similarity scores corresponding to the media assets in the media asset recommendation list; the similarity scores are sorted to obtain a similarity order of the media assets in the media asset recommendation list; Adjusting the original order of the media assets in the media asset recommendation list according to the similarity order; The reordered media resource recommendation list is returned to the terminal device.

2. The method according to claim 1, characterized in that Before determining the first user set according to the content attributes of the media assets in the historical media asset playback records that have not been uploaded, the method further includes: Determine in sequence whether the media asset playback duration in the historical media asset playback record that has not been uploaded exceeds a preset threshold; If yes, retain the current media asset in the playback record of the historical media asset that has not been uploaded; If not, the current media asset is deleted from the playback record of the historical media asset that has not been uploaded.

3. The method according to claim 2, characterized in that Adjusting the arrangement order of the media assets in the media asset recommendation list according to the similarity order includes: Setting the original order of the media assets in the media asset recommendation list as the first weight; Setting the similarity order to a second weight; The arrangement order of the media assets in the media asset recommendation list is calculated according to the first weight and the second weight.

4. The method according to claim 3, characterized in that Also includes: Returning the reordered media resource recommendation list to the terminal device, and receiving media resource click information within a preset time range; The first weight and the second weight are adjusted according to the original order and the re-ordered order of the clicked media assets, and the adjusted first weight and the second weight are stored.

5. The method according to claim 4, characterized in that According to the original order and the reordered order of the clicked media assets, adjusting the first weight and the second weight includes: Determine in sequence whether the order of the media resources corresponding to the media resource click information after reordering is less than the original order If the clicked media assets are in a lower order after reordering than in the original order, it will be treated as a corrected positive feedback; If the clicked media assets are in a higher order after reordering than in the original order, it will be treated as a corrected negative feedback; Calculate the difference between the number of times the positive feedback is corrected and the number of times the negative feedback is corrected; If the difference is a positive value, increasing the second weight; If the difference is a negative value, the second weight is reduced.

6. A server, characterized in that: The method comprises a memory and a processor, wherein the memory stores program instructions, and the processor executes the following steps by running the program instructions: Perform offline calculation based on uploaded historical media resource playback records to obtain user features; the uploaded historical media resource playback records are data uploaded by the terminal device before performing offline calculation; receiving a media asset recommendation request sent by the terminal device, and determining a media asset recommendation list according to the user characteristics; the media asset recommendation list includes a plurality of media assets, and the plurality of media assets are sorted in an original order; The original order is calculated based on the characteristic values ​​corresponding to the multiple media assets and the characteristic values ​​corresponding to the user; Obtaining the historical media playback records of the terminal device that have not been uploaded; Determine a first user set according to content attributes of media assets in the historical media asset playback records that have not been uploaded; the first user set is a set of users who like the media assets in the historical media asset playback records that have not been uploaded; The content attributes of the media assets in the historical media asset playback records that have not been uploaded are stored in the server; Determine a second user set according to the content attributes of the media in the media resource recommendation list; the second user set is a set of users who like the media in the media resource recommendation list; The content attributes of the media assets in the media asset recommendation list are stored in the server; Calculating the similarity between the media assets in the media asset recommendation list and the historical media asset playback records that have not been uploaded according to the intersection of the first user set and the second user set and the first user set and the second user set; The similarities are summed and averaged to obtain similarity scores corresponding to the media assets in the media asset recommendation list; the similarity scores are sorted to obtain a similarity order of the media assets in the media asset recommendation list; Adjusting the original order of the media assets in the media asset recommendation list according to the similarity order; The reordered media resource recommendation list is returned to the terminal device.

Images