CN108133031A - A kind of method and device of filtered recommendation video candidate result - Google Patents

Abstract

The embodiment of the present invention discloses a method and device for filtering recommended video candidate results, which receives a video recommendation service call request including a user ID sent by a user, and determines from the Redis database that the user ID corresponds to a system history recommendation The first Bloom filter instance of the video and the second Bloom filter instance of the recommended video containing the user's historical consumption; determine the video to be filtered from the recommended video candidate results; query the first Bloom filter instance and the second Bloom filter instance Whether the video to be filtered is included in the Bloom filter instance, if it is included, it is determined that the video to be filtered is the video to be recommended, and the video to be filtered is added to the first Bloom filter instance, if not included, then it is determined that the video to be filtered is not recommended Filter videos. Based on the above method and device, it is possible to improve the accuracy and efficiency of filtering recommended videos in the system history and recommended videos consumed by users in the recommended video candidate results.

Description

A method and device for filtering recommended video candidate results

technical field

The present invention relates to the technical field of video recommendation, in particular to a method and device for filtering recommended video candidate results.

Background technique

In the video recommendation system, it is necessary to obtain the final recommendation to the user from the recommended video candidate results calculated by the recommendation engine. According to the specific demand scenario, every time the video recommendation system user refreshes, it should be a different video recommendation result. The recommended video that the user has already watched, favorited, liked, etc. should not continue to be recommended to the user, so every time the video recommendation service is called It is necessary to efficiently select data from recommended video candidate results that is not repeated with the system's historically recommended videos and user's historically consumed recommended videos.

The video recommendation systems currently implemented in the industry all need to solve the problem of filtering the recommended videos in the system history and the recommended videos consumed by the user in the recommended video candidate results. The current conventional solution to the above problem is to store all recommended videos in system history and recommended videos consumed by users in a data set stored in the Redis database, and then judge whether the candidate recommended videos in the recommended video candidate results already exist In the above data set, if it exists, the candidate recommended video is not recommended.

However, in the prior art, the complete video ID is stored in the above data set. As the user’s time of using the video recommendation system increases, the number of times the user calls the video recommendation system’s recommendation service and the number of times the user consumes recommended video behaviors such as playing favorites It also increases accordingly, and the number of video ID digits is very long, which in turn causes the storage space of the Redis database occupied by the above data collection to become larger and larger. The data structure of the Redis database is based on the hash algorithm, that is to say, the storage structure of the above data collection is based on the hash function, and the hash function will cause hash conflicts when the amount of data is large. Therefore, When the storage space of the Redis database occupied by the data set is too large, the accuracy of filtering the recommended videos in the system history and the recommended videos consumed by the user in the recommended video candidate results is low.

In addition, there are often multiple candidate video recommendations in the recommended video candidate results. Since the above data sets are stored in the Redis database, it is necessary to call Redis multiple times to determine whether the candidate recommended videos in the recommended video candidate results already exist in the above data set. The operation function of the database will result in low efficiency in filtering the recommended videos of the system history and the recommended videos consumed by the user in the recommended video candidate results.

Contents of the invention

In view of this, the embodiments of the present invention provide a method and device for filtering recommended candidate results, which can solve the problem of low accuracy and low accuracy when filtering recommended videos in the system history and recommended videos consumed by users in the prior art among the recommended video candidate results. The problem of low efficiency.

In order to achieve the above purpose, embodiments of the present invention provide the following technical solutions:

A method for filtering recommended video candidate results, comprising:

receiving a video recommendation service invocation request sent by a user, the video recommendation service invocation request including a user identification ID;

From the Redis database, determine the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID;

Determine the video to be filtered from the recommended video candidate results;

Query whether the video to be filtered is contained in the first Bloom filter instance and the second Bloom filter instance, if so, determine that the video to be filtered is a video to be recommended, and set the video to be filtered The video is added to the first Bloom filter instance, if not included, it is determined that the video to be filtered is not recommended.

Optionally, after determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, the method further includes:

Judging whether the number of system history recommended videos included in the first Bloom filter instance reaches a first preset threshold, and if so, resetting the first Bloom filter instance.

Optionally, if the video recommendation service invocation request includes the number of recommended videos requested by the user, then after determining that the video to be filtered is the video to be recommended, and adding the video to be filtered to the first Bloom filter After the implementer instance, the method also includes:

Judging whether the number of recommended videos to the user is greater than the number of recommended videos requested by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and perform sequentially until the video is selected from the recommended video candidate results When it is determined that the number of videos to be filtered reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if it is greater than, a final video recommendation result is directly generated and sent to the user.

Optionally, after generating the final video recommendation result and sending it to the user, the method further includes:

monitoring the user's consumption event of the recommended video in the final video recommendation result;

When the user's consumption event of any recommended video in the final video recommendation result is detected, the recommended video is added to the second Bloom filter instance.

Optionally, determining from the Redis database the first Bloom filter instance containing recommended videos in system history and the second Bloom filter instance containing recommended videos consumed in user history corresponding to the user ID ,include:

generating a storage key value corresponding to the user ID;

Read the Bloom filter serialization string corresponding to the user ID from the Redis database according to the stored key value;

If the reading is successful, obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance;

If the reading fails, create the first Bloom filter instance and the second Bloom filter instance.

A device for filtering recommended video candidate results, comprising:

The receiving module is used to receive the video recommendation service call request sent by the user, and the video recommendation service call request includes the user identification ID;

The Bloom filter instance determination module is used to determine from the Redis database that the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID Long filter instance;

The video to be filtered determination module is used to determine the video to be filtered from the recommended video candidate results;

A processing module, configured to query whether the first Bloom filter instance and the second Bloom filter instance contain the video to be filtered, and if so, determine that the video to be filtered is a video to be recommended, and Add the video to be filtered to the first Bloom filter instance, if not included, determine that the video to be filtered is not recommended.

Optionally, the processing module is also used for:

After determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, determining the system history recommendation included in the first Bloom filter instance Whether the number of videos reaches a first preset threshold, and if so, resets the first Bloom filter instance.

Optionally, the processing module is also used for:

The video recommendation service call request includes the number of recommended videos requested by the user, then after determining that the video to be filtered is the video to be recommended, and adding the video to be filtered to the first Bloom filter instance, it is determined to Whether the number of videos recommended by the user is greater than the number of videos recommended by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and execute in sequence until the video to be filtered is determined from the recommended video candidate results Until the number of times of filtering videos reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if it is greater than, a final video recommendation result is directly generated and sent to the user.

Optionally, the processing module is also used for:

After generating the final video recommendation result and sending it to the user, monitor the user's consumption event of the recommended video in the final video recommendation result; When any recommended video consumption event occurs, the recommended video is added to the second Bloom filter instance.

Optionally, the Bloom filter instance determination module is specifically used for:

generating a storage key value corresponding to the user ID;

Read the Bloom filter serialization string corresponding to the user ID from the Redis database according to the stored key value;

If the reading is successful, obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance;

If the reading fails, create the first Bloom filter instance and the second Bloom filter instance.

Based on the above technical solution, the embodiments of the present invention disclose a method and device for filtering recommended video candidate results, receiving a video recommendation service invocation request sent by a user, and the video recommendation service invocation request includes a user identification ID; from the Redis database Determining the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID; Video; query whether the first Bloom filter instance and the second Bloom filter instance contain the video to be filtered, if so, determine that the video to be filtered is a video to be recommended, and the The video to be filtered is added to the first Bloom filter instance, if not included, it is determined that the video to be filtered is not recommended. Based on the above method and device, it is possible to improve the accuracy and efficiency of filtering recommended videos in the system history and recommended videos consumed by users in the recommended video candidate results.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only It is an embodiment of the present invention, and those skilled in the art can also obtain other drawings according to the provided drawings without creative work.

FIG. 1 is a schematic flowchart of a method for filtering recommended video candidate results provided by an embodiment of the present invention;

FIG. 2 is a schematic flowchart of another method for filtering recommended video candidate results provided by an embodiment of the present invention;

FIG. 3 is a schematic flowchart of another method for filtering recommended video candidate results provided by an embodiment of the present invention;

Fig. 4 is a schematic structural diagram of an apparatus for filtering recommended video candidate results provided by an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In the video recommendation system, it is necessary to obtain the final recommendation to the user from the recommended video candidate results calculated by the recommendation engine. According to the specific demand scenario, every time the video recommendation system user refreshes, it should be a different video recommendation result. The recommended video that the user has already watched, favorited, liked, etc. should not continue to be recommended to the user, so every time the video recommendation service is called It is necessary to efficiently select data from recommended video candidate results that is not repeated with the system's historically recommended videos and user's historically consumed recommended videos.

Redis is an open source, network-enabled, in-memory, optional persistent key-value store database written in ANSI C. Not only is it powerful, but it also has replication features and a unique data model for solving problems. Redis provides five different types of data structures, and various problems can be naturally mapped to these data structures: through features such as replication, persistence, and client-side sharding , users can easily expand Redis into a system that can contain hundreds of gigabytes of data and process millions of requests per second.

The video recommendation systems currently implemented in the industry all need to solve the problem of filtering the recommended videos in the system history and the recommended videos consumed by the user in the recommended video candidate results. The current conventional solution to the above problem is to store all recommended videos in system history and recommended videos consumed by users in a data set stored in the Redis database, and then judge whether the candidate recommended videos in the recommended video candidate results already exist In the above data set, if it exists, the candidate recommended video is not recommended.

The collection storage structure provided by the Redis database is a data structure that does not contain repeated elements. In the Redis database, there are two namespaces for storing system-historical recommended videos and user-historically consumed recommended videos. Each namespace has key-value pairs for each user and its corresponding set of video IDs. That is to say, in the prior art, the complete video ID is stored in the above-mentioned data set. As the user’s time of using the video recommendation system increases, the number of times the user calls the video recommendation system’s recommendation service and consumes recommended video behaviors such as playing favorites The number of times also increases accordingly, and the number of video ID digits is very long, which leads to the storage space of the Redis database occupied by the above data collection will also become larger and larger. The data structure of the Redis database is based on the hash algorithm, that is to say, the storage structure of the above data collection is based on the hash function, and the hash function will cause hash conflicts when the amount of data is large. Therefore, When the storage space of the Redis database occupied by the data set is too large, the accuracy of filtering the recommended videos in the system history and the recommended videos consumed by the user in the recommended video candidate results is low.

In addition, there are often multiple candidate video recommendations in the recommended video candidate results. Since the above data sets are stored in the Redis database, it is necessary to call Redis multiple times to determine whether the candidate recommended videos in the recommended video candidate results already exist in the above data set. The operation function of the database will result in low efficiency in filtering the recommended videos of the system history and the recommended videos consumed by the user in the recommended video candidate results.

Therefore, the present invention provides a method and device for filtering recommended video candidate results, so as to solve the problem of low accuracy and efficiency when filtering system history recommended videos and user historically consumed recommended videos in the prior art. lower question. The method and device for filtering recommended video candidate results provided by the present invention will be described in detail below through the following embodiments.

Please refer to accompanying drawing 1, Fig. 1 is a schematic flow diagram of a method for filtering recommended video candidate results provided by an embodiment of the present invention, the method includes the following steps:

Step S100, receiving a video recommendation service invocation request sent by a user, where the video recommendation service invocation request includes a user identification ID.

Step S110, determining from the Redis database a first Bloom filter instance containing recommended videos in the system history and a second Bloom filter instance containing recommended videos consumed by the user in the history corresponding to the user ID.

In order to ensure the time performance of the video recommendation system, the Redis database is used in this embodiment to store the Bloom filter.

The core implementation of a Bloom filter is a very large bit array and several hash functions. Assuming that the length of the bit array is m and the number of hash functions is k, each bit of the m-bit bit array is set to 0. When there is a piece of data to be written into the Bloom filter, k different hash functions are used. Hash function, calculate the hash value of the piece of data, and ensure that the obtained hash value is in the interval {1, m}. Now k hash functions will get k hash values, and all the bits corresponding to these values are set to 1 in the bit array, then the piece of data has been written into the Bloom filter at this time. When more data is added to the Bloom filter, if it is found that a certain bit is already 1, skip it.

When it is necessary to query whether a piece of data is contained in the Bloom filter, the k hash functions are also used for hash calculation to obtain k hash values, and the k hash values are checked in the bit array for the corresponding bits. Whether the value is 1, if there is a value other than 1 in these k bits, it proves that the piece of data does not exist in the Bloom filter, if the values of the obtained k bits are all 1, it proves that the piece of data may be included in the Bloom filter in the filter.

There is an error in judging whether a piece of data exists. If the Bloom filter judges that there is a piece of data, the data may exist. If it judges that there is no piece of data, it must not exist. For this kind of error, the video recommendation system is tolerable.

Step S120, determining the video to be filtered from the recommended video candidate results.

Step S130, query whether the video to be filtered is included in the first Bloom filter instance and the second Bloom filter instance, if yes, execute step S140, if not, execute step S150.

Step S140, determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance.

Step S150, determining that the video to be filtered is not recommended.

The embodiment of the present invention discloses a method for filtering recommended video candidate results, receiving a video recommendation service invocation request sent by a user, and the video recommendation service invocation request includes a user identification ID; Correspondingly, the first Bloom filter instance containing recommended videos in the system history and the second Bloom filter instance containing recommended videos consumed by users in history; determine the video to be filtered from the recommended video candidate results; query the first Whether the video to be filtered is included in the Bloom filter instance and the second Bloom filter instance, and if so, it is determined that the video to be filtered is a video to be recommended, and the video to be filtered is added to the If the first Bloom filter instance does not contain it, it is determined that the video to be filtered is not recommended. Based on the above method, the accuracy and efficiency of filtering recommended videos in system history and recommended videos consumed by users in history in the recommended video candidate results can be improved.

Please refer to accompanying drawing 2, Fig. 2 is the schematic flow chart of another kind of method for filtering recommended video candidate result provided by the embodiment of the present invention, this method comprises the following steps:

Step S200, receiving a video recommendation service invocation request sent by a user, where the video recommendation service invocation request includes a user identification ID.

Step S210, determining from the Redis database a first Bloom filter instance containing recommended videos in the system history and a second Bloom filter instance containing recommended videos consumed by the user in the history corresponding to the user ID.

This step specifically includes: generating a storage key value corresponding to the user ID; reading the Bloom filter serialization string corresponding to the user ID from the Redis database according to the storage key value; if the read is successful, Then obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance; if the reading fails, create the first Bloom filter filter instance and the second bloom filter instance. When creating or deserializing a Bloom filter instance, you need to specify the estimated capacity and accuracy of the created Bloom filter instance. These parameters are related to the storage space occupied by the generated Bloom filter instance.

Step S220, determining the video to be filtered from the recommended video candidate results.

Step S230, query whether the video to be filtered is included in the first Bloom filter instance and the second Bloom filter instance, if yes, execute step S240 and step S250, if not, execute step S230 S260.

Step S240, determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance.

Step S250, judging whether the number of system history recommended videos included in the first Bloom filter instance reaches a first preset threshold, and if so, resetting the first Bloom filter instance.

The first preset threshold is based on the size of the video recommendation engine's guaranteed data candidate set

In step S260, it is determined that the video to be filtered is not recommended.

In a method for filtering recommended video candidate results disclosed in an embodiment of the present invention, after determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, The method further includes: judging whether the number of system history recommended videos included in the first Bloom filter instance reaches a first preset threshold, and if so, resetting the first Bloom filter instance, which can Further avoiding the problem that too many system history recommended videos included in the first Bloom filter instance lead to a low accuracy rate of filtering the system history recommended videos in the recommended video candidate results.

Please refer to accompanying drawing 3, and Fig. 3 is the schematic flowchart of another kind of method for filtering recommended video candidate results provided by the embodiment of the present invention, and this method comprises the following steps:

Step S300, receiving a video recommendation service invocation request sent by a user, the video recommendation service invocation request including the user ID and the number of recommended videos requested by the user.

Step S310, determining from the Redis database a first Bloom filter instance containing recommended videos in the system history and a second Bloom filter instance containing recommended videos consumed by the user in the history corresponding to the user ID.

This step specifically includes: generating a storage key value corresponding to the user ID; reading the Bloom filter serialization string corresponding to the user ID from the Redis database according to the storage key value; if the read is successful, Then obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance; if the reading fails, create the first Bloom filter filter instance and the second bloom filter instance.

Step S320, determining the video to be filtered from the recommended video candidate results.

Step S330, query whether the video to be filtered is contained in the first Bloom filter instance and the second Bloom filter instance, if yes, perform steps S340 to S370, if not, perform step S330 S380.

Step S340, determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance.

Step S350, judging whether the number of recommended videos to the user is greater than the number of recommended videos requested by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and execute in sequence until the recommended video Until the number of times of the video to be filtered in the candidate results reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if greater, the final video recommendation result is directly generated and sent to the user.

Step S360, monitoring the user's consumption event of the recommended video in the final video recommendation result.

Step S370, when the user's consumption event of any recommended video in the final video recommendation result is detected, add the recommended video to the second Bloom filter instance.

The video consumption events of users will be collected by the video client using the video recommendation system and sent to Kafka. The video recommendation system collects and parses these consumption events, and reads the corresponding user’s second Bloom from the Redis database according to the obtained user ID. The serialized string of the filter instance, as above, if the serialized string of the second Bloom filter instance is successfully read, the second Bloom filter instance will be deserialized from the string, if there is no second Bloom filter instance The serialized string of the Bloom filter instance creates a second Bloom filter instance. Then add the video ID consumed by the user contained in the consumption event to the second Bloom filter instance, and then serialize the second Bloom filter instance as a serialized string and store it in the corresponding position of the Redis database.

In step S380, it is determined that the video to be filtered is not recommended.

In the method for filtering recommended video candidate results disclosed in this embodiment, after determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, the method It also includes: judging whether the number of recommended videos to the user is greater than the number of recommended videos requested by the user, and if not, return to the step of determining the video to be filtered from the recommended video candidate results and perform sequentially until the recommended video Until the number of times of the video to be filtered in the candidate results reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if greater, the final video recommendation result is directly generated and sent to the user. After generating the final video recommendation result and sending it to the user, the method further includes: monitoring the user's consumption event of the recommended video in the final video recommendation result; When there is a consumption event of any recommended video in the recommended video results, the recommended video is added to the second Bloom filter instance. It is possible to further improve the accuracy and efficiency of filtering recommended videos in the system history and recommended videos consumed by users in the recommended video candidate results.

Please refer to accompanying drawing 4. Fig. 4 is a schematic structural diagram of an apparatus for filtering recommended video candidate results provided by an embodiment of the present invention. The apparatus includes:

The receiving module 100 is configured to receive a video recommendation service call request sent by a user, and the video recommendation service call request includes a user identification ID;

The Bloom filter instance determination module 110 is used to determine from the Redis database the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID. Bloom filter instance;

The video to be filtered determining module 120 is used to determine the video to be filtered from the recommended video candidate results;

A processing module 130, configured to inquire whether the first Bloom filter instance and the second Bloom filter instance contain the video to be filtered, and if so, determine that the video to be filtered is a video to be recommended, And adding the video to be filtered to the first Bloom filter instance, if not included, it is determined that the video to be filtered is not recommended.

Optionally, the processing module 130 is also used for:

After determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, determining the system history recommendation included in the first Bloom filter instance Whether the number of videos reaches a first preset threshold, and if so, resets the first Bloom filter instance.

Optionally, the processing module 130 is also used for:

The video recommendation service call request includes the number of recommended videos requested by the user, then after determining that the video to be filtered is the video to be recommended, and adding the video to be filtered to the first Bloom filter instance, it is determined to Whether the number of videos recommended by the user is greater than the number of videos recommended by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and execute in sequence until the video to be filtered is determined from the recommended video candidate results Until the number of times of filtering videos reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if it is greater than, a final video recommendation result is directly generated and sent to the user.

Optionally, the processing module 130 is also used for:

After generating the final video recommendation result and sending it to the user, monitor the user's consumption event of the recommended video in the final video recommendation result; When any recommended video consumption event occurs, the recommended video is added to the second Bloom filter instance.

Optionally, the Bloom filter instance determination module 110 is specifically used for:

generating a storage key value corresponding to the user ID;

Read the Bloom filter serialization string corresponding to the user ID from the Redis database according to the stored key value;

If the reading is successful, obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance;

If the reading fails, create the first Bloom filter instance and the second Bloom filter instance.

It should be noted that the implementation of the specific functions of the above modules has been described in detail in the method embodiments, and will not be repeated in this embodiment.

In summary:

The embodiment of the present invention discloses a method and device for filtering recommended video candidate results, receiving a video recommendation service call request sent by a user, the video recommendation service call request includes a user identification ID; The ID corresponds to the first Bloom filter instance containing recommended videos in system history and the second Bloom filter instance containing recommended videos consumed by users in history; determine the video to be filtered from the recommended video candidate results; query the Whether the video to be filtered is included in the first Bloom filter instance and the second Bloom filter instance, if so, it is determined that the video to be filtered is a video to be recommended, and the video to be filtered is added to If the first Bloom filter instance does not contain it, it is determined that the video to be filtered is not recommended. Based on the above method and device, it is possible to improve the accuracy and efficiency of filtering recommended videos in the system history and recommended videos consumed by users in the recommended video candidate results.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other. As for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and for the related information, please refer to the description of the method part.

Professionals can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the possible For interchangeability, in the above description, the composition and steps of each example have been generally described according to their functions. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present invention.

The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be directly implemented by hardware, software modules executed by a processor, or a combination of both. Software modules can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other Any other known storage medium.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A method for filtering recommended video candidate results, comprising: receiving a video recommendation service invocation request sent by a user, the video recommendation service invocation request including a user identification ID; From the Redis database, determine the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID; Determine the video to be filtered from the recommended video candidate results; Query whether the video to be filtered is contained in the first Bloom filter instance and the second Bloom filter instance, if so, determine that the video to be filtered is a video to be recommended, and set the video to be filtered The video is added to the first Bloom filter instance, if not included, it is determined that the video to be filtered is not recommended. 2. The method according to claim 1, wherein, after determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, the The method also includes: Judging whether the number of system history recommended videos included in the first Bloom filter instance reaches a first preset threshold, and if so, resetting the first Bloom filter instance. 3. The method according to claim 1, wherein the video recommendation service invocation request includes the number of recommended videos requested by the user, then determining that the video to be filtered is a video to be recommended, and adding the video to be recommended After the filtered video is added to the first Bloom filter instance, the method further includes: Judging whether the number of recommended videos to the user is greater than the number of recommended videos requested by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and perform sequentially until the video is selected from the recommended video candidate results When it is determined that the number of videos to be filtered reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if it is greater than, a final video recommendation result is directly generated and sent to the user. 4. The method according to claim 3, characterized in that, after generating the final video recommendation result and sending it to the user, the method further comprises: monitoring the user's consumption event of the recommended video in the final video recommendation result; When the user's consumption event of any recommended video in the final video recommendation result is detected, the recommended video is added to the second Bloom filter instance. 5. The method according to any one of claims 1 to 4, wherein the first Bloom filter instance containing system history recommended videos corresponding to the user ID is determined from the Redis database and a second Bloom filter instance containing recommended videos that the user has consumed historically, including: generating a storage key value corresponding to the user ID; Read the Bloom filter serialization string corresponding to the user ID from the Redis database according to the stored key value; If the reading is successful, obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance; If the reading fails, create the first Bloom filter instance and the second Bloom filter instance. 6. A device for filtering recommended video candidate results, comprising: The receiving module is used to receive the video recommendation service call request sent by the user, and the video recommendation service call request includes the user identification ID; The Bloom filter instance determination module is used to determine from the Redis database that the first Bloom filter instance containing the system history recommended video and the second Bloom filter instance containing the recommended video consumed by the user history corresponding to the user ID Long filter instance; The video to be filtered determination module is used to determine the video to be filtered from the recommended video candidate results; A processing module, configured to query whether the first Bloom filter instance and the second Bloom filter instance contain the video to be filtered, and if so, determine that the video to be filtered is a video to be recommended, and Add the video to be filtered to the first Bloom filter instance, if not included, determine that the video to be filtered is not recommended. 7. The device according to claim 6, wherein the processing module is also used for: After determining that the video to be filtered is a video to be recommended, and adding the video to be filtered to the first Bloom filter instance, determining the system history recommendation included in the first Bloom filter instance Whether the number of videos reaches a first preset threshold, and if so, resets the first Bloom filter instance. 8. The device according to claim 6, wherein the processing module is also used for: The video recommendation service call request includes the number of recommended videos requested by the user, then after determining that the video to be filtered is the video to be recommended, and adding the video to be filtered to the first Bloom filter instance, it is determined to Whether the number of videos recommended by the user is greater than the number of videos recommended by the user, if not, return to the step of determining the video to be filtered from the recommended video candidate results and execute in sequence until the video to be filtered is determined from the recommended video candidate results Until the number of times of filtering videos reaches a second preset threshold, a final video recommendation result is generated and sent to the user; if it is greater than, a final video recommendation result is directly generated and sent to the user. 9. The device according to claim 8, wherein the processing module is also used for: After generating the final video recommendation result and sending it to the user, monitor the user's consumption event of the recommended video in the final video recommendation result; When any recommended video consumption event occurs, the recommended video is added to the second Bloom filter instance. 10. The device according to any one of claims 6 to 9, wherein the Bloom filter instance determination module is specifically used for: generating a storage key value corresponding to the user ID; Read the Bloom filter serialization string corresponding to the user ID from the Redis database according to the stored key value; If the reading is successful, obtain the byte array according to the serialized string and deserialize to obtain the first Bloom filter instance and the second Bloom filter instance; If the reading fails, create the first Bloom filter instance and the second Bloom filter instance.