WO2017181612A1 - Personalized video recommendation method and device - Google Patents

Personalized video recommendation method and device Download PDF

Info

Publication number
WO2017181612A1
WO2017181612A1 PCT/CN2016/101087 CN2016101087W WO2017181612A1 WO 2017181612 A1 WO2017181612 A1 WO 2017181612A1 CN 2016101087 W CN2016101087 W CN 2016101087W WO 2017181612 A1 WO2017181612 A1 WO 2017181612A1
Authority
WO
WIPO (PCT)
Prior art keywords
video
feature
user
user portrait
model corresponding
Prior art date
Application number
PCT/CN2016/101087
Other languages
French (fr)
Chinese (zh)
Inventor
孙浩川
Original Assignee
乐视控股(北京)有限公司
乐视网信息技术(北京)股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to CN201610244378.5A priority Critical patent/CN105930425A/en
Priority to CN201610244378.5 priority
Application filed by 乐视控股(北京)有限公司, 乐视网信息技术(北京)股份有限公司 filed Critical 乐视控股(北京)有限公司
Publication of WO2017181612A1 publication Critical patent/WO2017181612A1/en

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/73Querying
    • G06F16/735Filtering based on additional data, e.g. user or group profiles
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content

Abstract

A personalized video recommendation method and device, comprising: acquiring a user portrait characteristic corresponding to a user identifier (101); acquiring, according to the user portrait characteristic, a video characteristic model corresponding to the user portrait characteristic (102); acquiring, according to the video characteristic model corresponding to the user portrait characteristic, a video matching the video characteristic model corresponding to the user portrait characteristic from a video database and recommending the same (103). The method enables the user to rapidly and accurately acquire a video the user is interested in, thus increasing the efficiency of acquiring videos, satisfying the requirement for personalized videos of the user, and significantly enhancing user experience with respect to searching for and recommending videos for the user.

Description

个性化视频推荐方法及装置Personalized video recommendation method and device
交叉引用cross reference
本申请引用于2016年04月18日递交的名称为“个性化视频推荐方法及装置”的第201610244378.5号中国专利申请,其通过引用被全部并入本申请。The present application is hereby incorporated by reference in its entirety in its entirety in its entirety in the entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire entire all
技术领域Technical field
本发明实施例涉及计算机技术领域,尤其涉及一种个性化视频推荐方法及装置。The embodiments of the present invention relate to the field of computer technologies, and in particular, to a personalized video recommendation method and apparatus.
背景技术Background technique
随着每天有海量的视频特征上载到互联网,如何分析用户的兴趣以及如何推荐用户可能感兴趣的视频是一个很大的挑战,当前已有的视频推荐方法是用户主动选择喜欢的视频类别,然后系统根据用户选择来推荐相同类别的视频,需要用户通过视频网站手动操作进行设置,用户体验度较低。With the uploading of a large number of video features to the Internet every day, how to analyze the user's interests and how to recommend videos that may be of interest to users is a big challenge. The current video recommendation method is that the user actively selects the favorite video category, and then The system recommends the same category of video according to the user's choice, and the user needs to manually set it through the video website, and the user experience is low.
发明内容Summary of the invention
本发明实施例提供一种个性化视频推荐方法及装置,可以实现根据不同用户的兴趣主动向用户推荐感兴趣的视频,用户体验度较高。The embodiment of the invention provides a personalized video recommendation method and device, which can actively recommend a video of interest to the user according to the interest of different users, and the user experience is high.
本发明实施例还提供一种个性化视频推荐电子设备、一种非暂态计算机存储介质以及一种计算机程序产品。The embodiment of the invention further provides a personalized video recommendation electronic device, a non-transitory computer storage medium and a computer program product.
本发明实施例提供一种个性化视频推荐方法,包括:The embodiment of the invention provides a personalized video recommendation method, including:
获取与用户标识对应的用户画像特征;Obtaining a user portrait feature corresponding to the user identifier;
根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;Obtaining a video feature model corresponding to the user portrait feature according to the user portrait feature;
根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与 所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。Obtaining from the video database according to the video feature model corresponding to the user portrait feature The video of the video feature model corresponding to the user portrait feature is matched and recommended.
可选地,获取与用户标识对应的用户画像特征之前还包括:Optionally, before acquiring the user portrait feature corresponding to the user identifier, the method further includes:
根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征点和/或用户历史搜索的视频特征点,所述视频特征点包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率;Obtaining, according to the user identifier, user history behavior data corresponding to the user identifier, where the user history behavior data includes a video feature point played by the user history and/or a video feature point of the user history search, where the video feature point includes The region to which the video belongs, the type of video, the channel of the video, the time of video release, and/or the video clickthrough rate;
根据与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区。Calculating, according to the user historical behavior data corresponding to the user identifier, a user portrait feature corresponding to the user identifier, where the user portrait feature includes a user age, gender, occupation, a channel that has watched the video, and a type of the viewed video. , watched the video's tags, and/or the area where the video was viewed.
可选地,根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型包括:Optionally, acquiring a video feature model corresponding to the user portrait feature according to the user image feature includes:
根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。Determining, according to the user image feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point, and obtaining a video feature model corresponding to the user portrait feature.
可选地,根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐,包括:Optionally, the video matching the video feature model corresponding to the user portrait feature is obtained from the video database according to the video feature model corresponding to the user image feature, and the recommendation is performed, including:
根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集;Obtaining, according to a video feature model corresponding to the user portrait feature, a video candidate set that matches a video feature model corresponding to the user portrait feature from a video database;
根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。And sorting the video of the video candidate set according to the weight of each video feature point in the video feature model.
可选地,所述的方法还包括:Optionally, the method further includes:
根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。And filtering, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule.
本发明实施例还提供一种个性化视频推荐装置,包括:The embodiment of the invention further provides a personalized video recommendation device, including:
第一获取模块,用于获取与用户标识对应的用户画像特征; a first acquiring module, configured to acquire a user portrait feature corresponding to the user identifier;
第二获取模块,用于根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;a second acquiring module, configured to acquire a video feature model corresponding to the user portrait feature according to the user portrait feature;
第三获取模块,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。And a third acquiring module, configured to acquire a video matching the video feature model corresponding to the user portrait feature from the video database according to a video feature model corresponding to the user portrait feature, and perform recommendation.
可选地,所述装置还包括:Optionally, the device further includes:
第四获取模块,用于根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征点和/或用户历史搜索的视频特征点,所述视频特征点包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率;a fourth obtaining module, configured to acquire, according to the user identifier, user historical behavior data corresponding to the user identifier, where the user historical behavior data includes a video feature point played by the user history and/or a video feature point of the user history search The video feature point includes a region to which the video belongs, a video type, a channel of the video, a video publishing time, and/or a video click rate;
计算模块,用于根据所述第四获取模块获取的与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区。a calculation module, configured to calculate, according to the user historical behavior data corresponding to the user identifier acquired by the fourth acquiring module, a user portrait feature corresponding to the user identifier, where the user portrait feature includes a user age, gender, Occupation, channels that have watched videos, types of videos watched, tags that have watched videos, and/or areas where videos have been viewed.
可选地,所述第二获取模块具体用于:Optionally, the second obtaining module is specifically configured to:
根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。Determining, according to the user image feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point, and obtaining a video feature model corresponding to the user portrait feature.
可选地,所述第三获取模块包括:Optionally, the third obtaining module includes:
获取单元,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集;An acquiring unit, configured to acquire, according to a video feature model corresponding to the user portrait feature, a video candidate set that matches a video feature model corresponding to the user portrait feature from a video database;
排序单元,用于根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。a sorting unit, configured to sort the videos of the video candidate set according to weights of video feature points in the video feature model.
可选地,所述的装置还包括:Optionally, the device further includes:
过滤模块,用于根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。The filtering module is configured to filter, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule.
本发明实施例还提供一种非暂态计算机存储介质,存储有计算机可执行 指令,所述计算机可执行指令用于执行本申请上述任一项个性化视频推荐方法。Embodiments of the present invention also provide a non-transitory computer storage medium, which is stored in a computer executable The computer executable instructions are used to perform any of the above-described personalized video recommendation methods of the present application.
本发明实施例提供一种个性化视频推荐电子设备,包括:至少一个处理器;以及,An embodiment of the present invention provides a personalized video recommendation electronic device, including: at least one processor;
与所述至少一个处理器通信连接的存储器;其中,a memory communicatively coupled to the at least one processor; wherein
所述存储器存储有可被所述一个处理器执行的指令,所述指令被所述至少一个处理器执行,以使所述至少一个处理器能够:The memory stores instructions executable by the one processor, the instructions being executed by the at least one processor to enable the at least one processor to:
获取与用户标识对应的用户画像特征;Obtaining a user portrait feature corresponding to the user identifier;
根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;Obtaining a video feature model corresponding to the user portrait feature according to the user portrait feature;
根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。And according to the video feature model corresponding to the user portrait feature, the video matching the video feature model corresponding to the user portrait feature is obtained from the video database and recommended.
本发明实施例还提供了一种计算机程序产品,所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序,所述计算机程序包括程序指令,当所述程序指令被计算机执行时,使所述计算机执行本申请上述任一项个性化视频推荐方法。Embodiments of the present invention also provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer The computer is caused to perform any of the above-described personalized video recommendation methods of the present application.
本发明实施例通过获取与用户标识对应的用户画像特征;根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。使得用户可以快速准确地获取自己感兴趣的视频,提高了视频推荐和获取的效率,符合用户的个性化视频需求,大大提高用户视频搜索推荐的体验度。The embodiment of the present invention acquires a user portrait feature corresponding to the user identifier, and acquires a video feature model corresponding to the user portrait feature according to the user portrait feature; and the video feature model corresponding to the user portrait feature, from the video A video matching the video feature model corresponding to the user portrait feature is obtained in the database and recommended. The user can quickly and accurately obtain the video of interest, improve the efficiency of video recommendation and acquisition, meet the user's personalized video requirements, and greatly improve the user experience of video search recommendation.
附图说明DRAWINGS
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍,显而易见地,下面描述中的附图是本发明的一些实施例,对于本领域普通技术人员来讲,在 不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description of the drawings used in the embodiments or the prior art description will be briefly described below. Obviously, the drawings in the following description Some embodiments of the present invention, for those of ordinary skill in the art, Other drawings may also be obtained from these drawings without the use of creative labor.
图1是本发明实施例提供的一种个性化视频推荐方法的流程示意图;1 is a schematic flowchart of a personalized video recommendation method according to an embodiment of the present invention;
图2为本发明实施例提供的基于用户画像特征的逻辑回归模型的训练方法示意图;2 is a schematic diagram of a training method of a logistic regression model based on user portrait features according to an embodiment of the present invention;
图3为图2所示实施例中视频数据的抽取和视频数据的清洗步骤的具体实现方法示意图;3 is a schematic diagram of a specific implementation method of extracting video data and cleaning steps of video data in the embodiment shown in FIG. 2;
图4为图2所示实施例中数据的特征构造步骤的具体实现方法示意图;4 is a schematic diagram of a specific implementation method of a feature construction step of data in the embodiment shown in FIG. 2;
图5为图2所示实施例中模型训练步骤的具体实现方法示意图;FIG. 5 is a schematic diagram of a specific implementation method of a model training step in the embodiment shown in FIG. 2; FIG.
图6是本发明实施例提供的一种个性化视频推荐装置的结构示意图;以及FIG. 6 is a schematic structural diagram of a personalized video recommendation apparatus according to an embodiment of the present invention;
图7为本发明实施例提供的一种个性化视频推荐电子设备的结构示意图。FIG. 7 is a schematic structural diagram of a personalized video recommendation electronic device according to an embodiment of the present invention.
具体实施方式detailed description
为使本发明实施例的目的、技术方案和优点更加清楚,下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. It is a partial embodiment of the invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.
图1是本发明实施例提供的一种个性化视频推荐方法的流程示意图。该方法可以由个性化搜索装置来执行,所述装置可由软件来实现,可作为实现搜索引擎的一部分被内置在具有搜索功能的终端设备上。其中,终端设备可以是智能手机、平板电脑、笔记本电脑、台式电脑、个人数字助理等。参见图1,本实施例提供的个性化视频推荐方法具体包括如下操作:FIG. 1 is a schematic flowchart of a personalized video recommendation method according to an embodiment of the present invention. The method can be performed by a personalized search device, which can be implemented by software, and can be built into a terminal device having a search function as part of implementing a search engine. The terminal device may be a smart phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, or the like. Referring to FIG. 1 , the personalized video recommendation method provided in this embodiment specifically includes the following operations:
步骤101、获取与用户标识对应的用户画像特征;Step 101: Acquire a user portrait feature corresponding to the user identifier.
具体地,当用户打开视频网站(如乐视视频)时,即可享视频网站的后台服务器发送携带有该用户的用户标识的视频请求,其中,用户标识例如可 以是用户的会员账号(如乐视会员账号),也可以是该用户使用的用户设备(如IPAD)的硬件标识;后台服务器根据该用户标识,获取与该用户标识对应的用户画像特征。Specifically, when the user opens a video website (such as LeTV video), the background server of the video website can send a video request carrying the user identifier of the user, wherein the user identifier can be, for example, The user's member account (such as the music account member account) may also be the hardware identifier of the user equipment (such as the IPAD) used by the user; the background server obtains the user portrait feature corresponding to the user identifier according to the user identifier.
为此,步骤101之前包括:To this end, before step 101, it includes:
根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征点和/或用户历史搜索的视频特征点,所述视频特征点例如包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率;Obtaining, according to the user identifier, user history behavior data corresponding to the user identifier, where the user history behavior data includes a video feature point played by the user history and/or a video feature point of the user history search, where the video feature point is, for example, Including the region to which the video belongs, the type of video, the channel of the video, the time of video release, and/or the video click rate;
根据与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的兴趣标签和/或观看过视频的地区。Calculating, according to the user historical behavior data corresponding to the user identifier, a user portrait feature corresponding to the user identifier, where the user portrait feature includes a user age, gender, occupation, a channel that has watched the video, and a type of the viewed video. , watched the video’s interest tags, and/or the area where the video was viewed.
需要说明的是,用户画像特征是个性化推荐的基础,一个好的推荐系统必须有一个准确的用户画像特征作为基础。用户画像特征的构建包括以下几个方面:It should be noted that the user portrait feature is the basis of personalized recommendation, and a good recommendation system must have an accurate user portrait feature as the basis. The construction of user portrait features includes the following aspects:
用户标识,对于登录用户采用用户ID来进行标识,对于没有登录的用户采用用户设备ID进行标识,之后可以生成用户ID和用户设备ID的对应关系;The user ID is used to identify the user ID by using the user ID, and the user ID is used for the user who is not logged in, and then the corresponding relationship between the user ID and the user ID can be generated.
用户历史行为数据的收集,在个性化搜索场景下的用户画像特征主要依赖用户的历史点击播放行为和搜索关键词等信息,根据用户ID将同一个用户的历史行为数据聚集到一起作为之后用户画像特征算法的输入;The collection of user historical behavior data, the user portrait feature in the personalized search scenario mainly depends on the user's historical click play behavior and search keyword information, and the historical behavior data of the same user is gathered together according to the user ID as the subsequent user portrait. Input of the feature algorithm;
其中,用户画像的特征包括但不限于用户的基本属性(如性别,年龄段,职业等等),还可以包括观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区;Among them, the characteristics of the user portrait include, but are not limited to, the basic attributes of the user (such as gender, age, occupation, etc.), and may also include the channel on which the video has been viewed, the type of video viewed, the label of the viewed video, and/or the viewing. The area of the video;
其中,用户画像特征的生成算法需要考虑以下几个方面:Among them, the algorithm for generating user portrait features needs to consider the following aspects:
用户兴趣计算,通常为视频观看时间比例*视频的权重。而视频的权重计算方法通常为:总用户数/看过此视频的用户数+1; User interest calculation, usually the video viewing time ratio * the weight of the video. The weight calculation method of the video is usually: the total number of users / the number of users who have seen this video +1;
时间衰减,对于用户近期看过的视频给予比较近的权重,随用户行为历史时间递减;Time decay, giving relatively close weights to videos that users have recently seen, decreasing with user history time;
曝光未点击降权,对于用户曝光但是没有点击的历史行为降权。The exposure is not clicked down, and the historical behavior of the user is exposed but not clicked.
步骤102、根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;Step 102: Acquire a video feature model corresponding to the user portrait feature according to the user portrait feature.
具体地,根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。其中,视频特征点例如包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率。Specifically, according to the user portrait feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point are determined, and a video feature model corresponding to the user portrait feature is obtained. The video feature points include, for example, a region to which the video belongs, a video type, a channel of the video, a video publishing time, and/or a video click rate.
需要说明的是,本发明实施例采用倒排索引为个性化推荐提供候选集,尽管这个候选集中并不是每一个视频都很准确,但是这个候选集力求有一个比较的召回,也就是尽量将用户喜欢的视频都包含。主要的方式包括:It should be noted that the embodiment of the present invention uses the inverted index to provide a candidate set for personalized recommendation. Although this candidate set is not accurate for each video, the candidate set strives to have a comparative recall, that is, try to user. Favorite videos are included. The main ways include:
基于协同过滤的ItemCF算法,即为每一个视频计算出点击过这个视频的用户之后最有可能点击的其他视频,具体的计算方式如下:视频A对于视频B的重要性=在点击A的用户中同时点击了B的用户数/点击了A的用户数。The ItemCF algorithm based on collaborative filtering calculates the other videos that are most likely to be clicked after the user who clicked on the video for each video. The specific calculation method is as follows: The importance of video A for video B = among the users who click A The number of users who clicked B at the same time / the number of users who clicked A.
基于标签(tag)的方法,将热门tag的视频聚类,然后对每一个类中选取热度最高的若干个视频。Based on the tag method, the videos of the popular tags are clustered, and then the videos with the highest heat are selected for each class.
需要说明的是,在本发明的另一实施例中,采用基于用户画像特征的逻辑回归模型提供了对上述候选集的排序,确保最终为用户推荐的top n个视频是最符合用户个性化的需求。图2为本发明实施例提供的基于用户画像特征的逻辑回归模型的训练方法示意图,如图2所示,基本步骤如下:It should be noted that, in another embodiment of the present invention, a logistic regression model based on user portrait features is provided to provide ordering of the candidate sets, and to ensure that the top n videos finally recommended by the user are most suitable for user personalization. demand. FIG. 2 is a schematic diagram of a training method of a logistic regression model based on user portrait features according to an embodiment of the present invention. As shown in FIG. 2, the basic steps are as follows:
步骤201.视频数据的抽取和视频数据的清洗;Step 201. Extracting video data and cleaning the video data;
图3为图2所示实施例中视频数据的抽取和视频数据的清洗步骤的具体实现方法示意图,如图3所示包括:FIG. 3 is a schematic diagram of a specific implementation method of extracting video data and cleaning steps of video data in the embodiment shown in FIG. 2, and FIG. 3 includes:
步骤2011.从视频数据库抽取视频特征,视频特征包括但不限于视频的频道、类型、所属地区、发布时间、更新时间和视频在过去某段时间的点击 率。Step 2011. Extract video features from the video database, including but not limited to the channel, type, region, release time, update time, and click of the video in a certain period of time. rate.
步骤2012.从视频数据库抽取用户画像特征,例如用户的年龄,性别,职业,观看过视频的频道,观看过视频的类型,观看过视频的兴趣标签,观看过视频所属的地区。Step 2012. Extract user image features from the video database, such as the user's age, gender, occupation, channel that has watched the video, watched the type of video, watched the interest tag of the video, and watched the region to which the video belongs.
步骤2013.从视频数据库抽取某一段时间的用户观看历史视频的历史行为,包括用户ID,推荐视频的列表,推荐时间,是否点击过推荐的某一个视频。Step 2013. Extract the historical behavior of the user watching the historical video for a certain period of time from the video database, including the user ID, the list of recommended videos, the recommended time, and whether a certain video has been clicked.
步骤2014.根据预设的清洗规则去除符合请求规则的视频数据;如推荐的视频用户点击率很低,这样导致正负样本不均匀,会影响后续模型的训练,需要我们在负样本中随机抽取一定比例的数据,使得正负样本的比例达到1:1,其中,正样本例如为用户点击的推荐视频,负样本为用户没有点击的推荐视频。Step 2014. According to the preset cleaning rule, the video data that meets the request rule is removed; if the recommended video user has a low click rate, the positive and negative samples are uneven, which may affect the training of the subsequent model, and we need to randomly extract from the negative sample. A certain proportion of data makes the ratio of positive and negative samples reach 1:1, wherein the positive sample is, for example, a recommended video clicked by the user, and the negative sample is a recommended video that the user does not click.
步骤202.数据的特征构造;Step 202. Feature construction of the data;
图4为图2所示实施例中数据的特征构造步骤的具体实现方法示意图,如图4所示,包括:FIG. 4 is a schematic diagram of a specific implementation method of the feature construction steps of the data in the embodiment shown in FIG. 2, as shown in FIG. 4, including:
步骤2021.连续特征离散化,本发明实施例需要对连续的特征离散化,比如视频过去某段时间的点击率,一般采用的方法是等频分割,就是对需要进行离散化的特征上所有的样本排序,平均分为若干等份,用其所在的指标(index)取代原来特征的值。Step 2021. The continuous feature discretization, the embodiment of the present invention needs to discretize the continuous feature, such as the click rate of the video in a certain period of time, generally adopting the method of equal frequency segmentation, that is, all the features that need to be discretized The sample is sorted and divided into several equal parts, and the value of the original feature is replaced by the index (index).
步骤2022.特征one-hot编码,比如某一个视频的类型可以分别属于战争、爱情、生活。需要对类别特征进行one-hot编码,即使用一位来表示类别中的某一个值,把原来的一维特征转化为n维特征,这里的n就是原来类别特征中所有的取值。Step 2022. Feature one-hot coding, for example, the type of a certain video may belong to war, love, and life. The one-hot encoding of the category feature is required, that is, one bit is used to represent a certain value in the category, and the original one-dimensional feature is converted into an n-dimensional feature, where n is all the values in the original category feature.
步骤2023.Feature Cross(特征交叉),将用户画像特征和视频特征进行交叉处理。由于特征交叉需要消耗大量的计算时间,这里考虑到性能问题,特征交叉时优选进行三维特征的交叉。下表列举了主要特征交叉规则: Step 2023. Feature Cross, cross-processing user portrait features and video features. Since the feature intersection requires a large amount of computation time, here considering the performance problem, it is preferable to perform the intersection of the three-dimensional features when the features intersect. The following table lists the main feature intersection rules:
序号Serial number 第一维First dimension 第二维Second dimension 第三维Third dimension
11 用户感兴趣频道User interest channel 视频所属频道Video belongs to the channel  
22 用户感兴趣类型User interest type 视频类型Video type  
33 用户感兴趣类型User interest type 视频TAGVideo TAG  
44 用户感兴趣TAGInterested in TAG 视频TAGVideo TAG  
55 用户感兴趣地区User area of interest 视频地区Video area  
66 用户感兴趣类型User interest type 视频类型Video type 视频发布时间Video release time
77 用户感兴趣类型User interest type 视频类型Video type 推荐时间Recommended time
88 用户感兴趣类型User interest type 视频频道Video channel 推荐时间Recommended time
步骤203.模型训练;Step 203. Model training;
图5为图2所示实施例中模型训练步骤的具体实现方法示意图,如图5所示,包括:FIG. 5 is a schematic diagram of a specific implementation method of a model training step in the embodiment shown in FIG. 2, as shown in FIG. 5, including:
步骤2031.特征编码,由于上述生成的特征都是用字符串表示,虽然易于查看,但是计算性能较差,本发明采用每个特征出现的次序对特征进行从0开始编码,将字符串特征转为为int型整数,提高计算性能。Step 2031. Feature encoding, since the generated features are all represented by a character string, although easy to view, but the calculation performance is poor, the present invention uses the order in which each feature appears to encode the feature from 0, and converts the character string into Improve computational performance for int-type integers.
步骤2032.对之前特征交叉后的数据进行拟合,例如使用FTRL算法,最后得到一个可以表达用户对视频喜爱程度的视频特征模型,视频特征模型表现形式例如是KV对,key表示为特征名称,value表示为特征的权重。Step 2032. Fitting the data after the feature intersection, for example, using the FTRL algorithm, and finally obtaining a video feature model that can express the user's preference for the video. The video feature model representation is, for example, a KV pair, and the key is represented as a feature name. Value is expressed as the weight of the feature.
步骤103、根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。Step 103: Obtain a video matching the video feature model corresponding to the user portrait feature from the video database according to the video feature model corresponding to the user portrait feature and perform recommendation.
具体地,根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集;根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。Specifically, the video feature set matching the video feature model corresponding to the user portrait feature is obtained from the video database according to the video feature model corresponding to the user portrait feature; and according to each video feature point in the video feature model Weighting, sorting the video of the video candidate set.
举例来说,某一视频用户通过IPAD打开乐视视频收看的视频包括小马 宝莉、哆啦爱探险、灰姑娘、白雪公主等迪斯尼公主系列的动画视频,通过对该用户的历史行为数据的挖掘,可以知道该视频用户的用户画像特征包括儿童,年龄为18岁以下、女性、观看过视频类型为动画片、观看过视频的频道(电影或动漫)、观看过视频的兴趣标签(公主系列)、观看过视频的地区(美国)等特征信息;根据该儿童用户画像特征,基于该儿童用户画像特征的逻辑回归模型,得到与该儿童用户画像特征对应的视频特征模型,其中,该视频特征模型中包括与该儿童用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,例如,与该儿童用户画像特征对应的视频特征模型包括美国(视频所属地区)、公主片(视频类型)、动漫频道(视频的频道)、视频发布时间和/或视频点击率等视频特征点;从而可以从视频数据库中获取与该儿童用户画像特征对应的视频特征模型匹配的视频候选集,为了使得推荐的视频更加符合用户的个性化需求,本发明实施例中对视频特征模型中的多个视频特征点可以设置对应的权重,例如根据视频发布时间,将最近发布的视频的权重设置的较高(对应地,发布时间较远的视频的权重较低);根据视频点击率,将点击率高的视频的权重设置的较高(对应地,点击率低的视频的权重较低);根据视频类型,将公主片的视频权重设置较高(对应地,非公主片的视频的权重设置较低);根据视频频道,将动漫频道的视频的权重设置较高(对应地,非动漫频道的视频的权重设置较低);本发明实施例不一一举例了,通过举例说明,本发明实施例中,视频特征模型中的每个视频特征点的权重设置是根据不同用户的用户画像特征而可以做相应的调整。因此,用户画像特征不同,视频特征模型也是不同。视频特征模型中每个视频特征点的权重也是不同,从而可以更加准确地推荐和获取符合该用户的个性化需求的视频。因此,最终推荐给用户的视频(或者说展示在用户设备中的视频)为根据与该儿童用户画像特征对应的视频特征模型中每个视频特征的权重对视频候选集中视频进行排序,假设视频美人鱼属于动漫视频、公主片、最近发布、美国地区、视频点击率高,则可以将该视频美人鱼作为视频候选集中第一位视频(特点视频)推荐给该儿童用户;假设视频长发公主属于动漫视频、公主片、美国地区、视频点击率一般、发布时间远,则可以将该视频长发公主作为候选集中非热点视频(如第三位之后的 视频序列)推荐给该儿童用户。For example, a video user who opens a video of LeTV video via IPAD includes a pony. Animated videos of the Disney Princess series, such as Polaroid, Dora Love, Cinderella, Snow White, etc., by mining the historical behavior data of the user, we can know that the user image characteristics of the video user include children, under the age of 18, Female, watched video types such as cartoons, channels that watched videos (movies or anime), interest tags that watched videos (princess series), areas where the video was viewed (USA), and other characteristics based on the child's user image And obtaining a video feature model corresponding to the child user portrait feature based on the logistic regression model of the child user portrait feature, wherein the video feature model includes a plurality of video feature points corresponding to the child user portrait feature and each video The weight corresponding to the feature point, for example, the video feature model corresponding to the child user portrait feature includes the United States (the region to which the video belongs), the princess (the video type), the animation channel (the channel of the video), the video release time, and/or the video click. Video feature points, etc.; thus can be drawn from the video database with the child user In the embodiment of the present invention, a plurality of video feature points in the video feature model may be set with corresponding weights, for example, according to the video, in order to make the recommended video more suitable for the user's individual needs. Release time, the weight of the recently released video is set higher (correspondingly, the weight of the video with a longer release time is lower); according to the video click rate, the weight of the video with a high click rate is set higher (correspondingly The video with low click-through rate has a lower weight; according to the video type, the video weight of the princess is set higher (correspondingly, the weight of the non-Princess video is set lower); according to the video channel, the video of the anime channel is The weight setting is higher (correspondingly, the weight setting of the video of the non-anime channel is lower); the embodiments of the present invention are not exemplified, by way of example, in the embodiment of the present invention, each video feature in the video feature model The weight setting of the points can be adjusted according to the characteristics of the user images of different users. Therefore, the user's portrait features are different and the video feature model is different. The weight of each video feature point in the video feature model is also different, so that the video that meets the personalized needs of the user can be recommended and obtained more accurately. Therefore, the video that is ultimately recommended to the user (or the video displayed in the user device) is to sort the video candidate set video according to the weight of each video feature in the video feature model corresponding to the child user portrait feature, assuming the video mermaid Belonging to anime video, princess film, recently released, US region, high video click rate, you can recommend the video mermaid as the first video (feature video) in the video candidate set to the child user; suppose the video long hair princess belongs to anime video , Princess film, the United States, video click-through rate, release time, you can use the video long hair princess as a candidate for non-hot spots video (such as after the third place) The video sequence is recommended for this child user.
可选地,步骤103之前或之后还包括:Optionally, before or after step 103, the method further includes:
根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。例如,本发明实施例中,预设的视频过滤规则包括色情视频特征点和反动视频特征点等等,这样在用户设备显示推荐的视频链接之前需要将符合视频过滤规则中的视频特征点的视频过滤掉。And filtering, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule. For example, in the embodiment of the present invention, the preset video filtering rule includes a pornographic video feature point, a reaction video feature point, and the like, so that the video that meets the video feature point in the video filtering rule needs to be displayed before the user equipment displays the recommended video link. Filtered.
本发明实施例通过根据检测到的视频请求中携带的用户标识,获取与所述用户标识对应的用户画像特征;根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。使得用户可以快速准确地获取自己感兴趣的视频,提高了视频推荐和获取的效率,符合用户的个性化视频需求,大大提高用户视频搜索推荐的体验度。The embodiment of the present invention acquires a user portrait feature corresponding to the user identifier according to the user identifier carried in the detected video request, and acquires a video feature model corresponding to the user portrait feature according to the user portrait feature; The video feature model corresponding to the user portrait feature acquires a video matching the video feature model corresponding to the user portrait feature from the video database and performs recommendation. The user can quickly and accurately obtain the video of interest, improve the efficiency of video recommendation and acquisition, meet the user's personalized video requirements, and greatly improve the user experience of video search recommendation.
图6是本发明实施例提供的一种个性化视频推荐装置的结构示意图,如图6所示,包括:FIG. 6 is a schematic structural diagram of a personalized video recommendation apparatus according to an embodiment of the present invention. As shown in FIG. 6, the method includes:
第一获取模块21,用于,获取与用户标识对应的用户画像特征;The first obtaining module 21 is configured to acquire a user portrait feature corresponding to the user identifier;
第二获取模块22,用于根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;The second obtaining module 22 is configured to acquire a video feature model corresponding to the user portrait feature according to the user portrait feature;
第三获取模块23,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。The third obtaining module 23 is configured to obtain a video matching the video feature model corresponding to the user portrait feature from the video database according to the video feature model corresponding to the user portrait feature and perform recommendation.
可选地,所述的装置还包括:Optionally, the device further includes:
第四获取模块24,用于根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征和/或用户历史搜索的视频特征,所述视频特征包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率; The fourth obtaining module 24 is configured to obtain, according to the user identifier, user historical behavior data corresponding to the user identifier, where the user historical behavior data includes a video feature played by the user history and/or a video feature of the user history search. The video features include a region to which the video belongs, a video type, a channel of the video, a video publishing time, and/or a video click rate;
计算模块25,用于根据所述第四获取模块获取的与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区。The calculating module 25 is configured to calculate a user portrait feature corresponding to the user identifier according to the user historical behavior data corresponding to the user identifier acquired by the fourth acquiring module, where the user portrait feature includes the user age and gender , career, channels that have watched videos, types of videos watched, tags that have watched videos, and/or areas where videos have been viewed.
可选地,所述第二获取模块22具体用于:Optionally, the second obtaining module 22 is specifically configured to:
根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。Determining, according to the user image feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point, and obtaining a video feature model corresponding to the user portrait feature.
可选地,所述第三获取模块23包括:Optionally, the third obtaining module 23 includes:
获取单元231,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集;The obtaining unit 231 is configured to acquire, according to a video feature model corresponding to the user portrait feature, a video candidate set that matches a video feature model corresponding to the user portrait feature from a video database;
排序单元232,用于根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。The sorting unit 232 is configured to sort the video candidate set videos according to the weights of the video feature points in the video feature model.
可选地,所述的装置还包括:Optionally, the device further includes:
过滤模块26,用于根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。The filtering module 26 is configured to filter, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule.
本发明实施例的个性化视频推荐装置根据检测到的视频请求中携带的用户标识,获取与所述用户标识对应的用户画像特征;根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。使得用户可以快速准确地获取自己感兴趣的视频,提高了视频获取的效率,符合用户的个性化视频需求,大大提高用户视频搜索体验度。The personalized video recommendation device of the embodiment of the present invention acquires a user portrait feature corresponding to the user identifier according to the user identifier carried in the detected video request, and acquires a feature corresponding to the user portrait feature according to the user image feature. The video feature model obtains a video matching the video feature model corresponding to the user portrait feature from the video database according to the video feature model corresponding to the user portrait feature and performs recommendation. The user can quickly and accurately obtain the video that he is interested in, improve the efficiency of video acquisition, meet the user's personalized video requirements, and greatly improve the user's video search experience.
图7为本发明实施例提供的一种个性化视频推荐电子设备的结构示意图,本实施例所述设备可以为个性化视频推荐服务器或个性化视频推荐服务器中的一部分,该设备可以包括: FIG. 7 is a schematic structural diagram of a personalized video recommendation electronic device according to an embodiment of the present invention. The device may be part of a personalized video recommendation server or a personalized video recommendation server, and the device may include:
一个或多个处理器501以及存储器502,图7中以一个处理器501为例。One or more processors 501 and memory 502, one processor 501 is exemplified in FIG.
个性化视频推荐电子设备还可以包括:输入装置503和输出装置504。The personalized video recommendation electronic device may further include: an input device 503 and an output device 504.
处理器501、存储器502、输入装置503和输出装置504可以通过总线或者其他方式连接。The processor 501, the memory 502, the input device 503, and the output device 504 can be connected by a bus or other means.
存储器502作为一种非暂态计算机可读存储介质,可用于存储非暂态软件程序、非暂态计算机可执行程序以及模块,如本发明实施例中的个性化视频推荐方法对应的程序指令/模块。处理器501通过运行存储在存储器502中的非暂态软件程序、指令以及模块,从而执行服务器的各种功能应用以及数据处理,即实现上述方法实施例个性化视频推荐方法。The memory 502 is used as a non-transitory computer readable storage medium, and can be used for storing a non-transitory software program, a non-transitory computer executable program, and a module, such as a program instruction corresponding to the personalized video recommendation method in the embodiment of the present invention. Module. The processor 501 executes various functional applications and data processing of the server by running non-transitory software programs, instructions, and modules stored in the memory 502, that is, implementing the personalized video recommendation method of the above method embodiment.
存储器502可以包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需要的应用程序;存储数据区可存储根据个性化视频推荐装置的使用所创建的数据等。此外,存储器502可以包括高速随机存取存储器,还可以包括非暂态存储器,例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施例中,存储器502可选包括相对于处理器501远程设置的存储器,这些远程存储器可以通过网络连接至个性化视频推荐的处理装置。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。The memory 502 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created according to use of the personalized video recommendation device, and the like. Moreover, memory 502 can include high speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 502 can optionally include a memory remotely located relative to the processor 501 that can be connected to the processing device of the personalized video recommendation over a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
输入装置503可接收输入的数字或字符信息,以及产生与个性化视频推荐装置的用户设置以及功能控制有关的键信号输入。输出装置504可包括显示屏等显示设备。The input device 503 can receive the input digital or character information and generate a key signal input related to user settings and function control of the personalized video recommendation device. Output device 504 can include a display device such as a display screen.
所述一个或者多个模块存储在所述存储器502中,当被所述一个或者多个处理器501执行时,执行上述任意方法实施例中的个性化视频推荐方法。The one or more modules are stored in the memory 502, and when executed by the one or more processors 501, perform a personalized video recommendation method in any of the above method embodiments.
上述产品可执行本发明实施例所提供的方法,具备执行方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节,可参见本发明实施例所提供的方法。The above product can perform the method provided by the embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. For technical details that are not described in detail in this embodiment, reference may be made to the method provided by the embodiments of the present invention.
本发明实施例的电子设备以多种形式存在,包括但不限于:The electronic device of the embodiment of the invention exists in various forms, including but not limited to:
(1)移动通信设备:这类设备的特点是具备移动通信功能,并且以提供话 音、数据通信为主要目标。这类终端包括:智能手机(例如iPhone)、多媒体手机、功能性手机,以及低端手机等。(1) Mobile communication equipment: This type of equipment is characterized by its mobile communication function and Sound and data communication are the main goals. Such terminals include: smart phones (such as iPhone), multimedia phones, functional phones, and low-end phones.
(2)超移动个人计算机设备:这类设备属于个人计算机的范畴,有计算和处理功能,一般也具备移动上网特性。这类终端包括:PDA、MID和UMPC设备等,例如iPad。(2) Ultra-mobile personal computer equipment: This type of equipment belongs to the category of personal computers, has computing and processing functions, and generally has mobile Internet access. Such terminals include: PDAs, MIDs, and UMPC devices, such as the iPad.
(3)便携式娱乐设备:这类设备可以显示和播放多媒体内容。该类设备包括:音频、视频播放器(例如iPod),掌上游戏机,电子书,以及智能玩具和便携式车载导航设备。(3) Portable entertainment devices: These devices can display and play multimedia content. Such devices include: audio, video players (such as iPod), handheld game consoles, e-books, and smart toys and portable car navigation devices.
(4)服务器:提供计算服务的设备,服务器的构成包括处理器、硬盘、内存、系统总线等,服务器和通用的计算机架构类似,但是由于需要提供高可靠的服务,因此在处理能力、稳定性、可靠性、安全性、可扩展性、可管理性等方面要求较高。(4) Server: A device that provides computing services. The server consists of a processor, a hard disk, a memory, a system bus, etc. The server is similar to a general-purpose computer architecture, but because of the need to provide highly reliable services, processing power and stability High reliability in terms of reliability, security, scalability, and manageability.
(5)其他具有数据交互功能的电子装置。(5) Other electronic devices with data interaction functions.
相应地,本发明实施例还提供一种计算机可读存储介质,所述计算机可读存储介质中存储有用于执行上述实施例方法的程序。Correspondingly, an embodiment of the present invention further provides a computer readable storage medium, where the program for executing the method of the foregoing embodiment is stored.
以上所描述的装置实施例仅仅是示意性的,其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下,即可以理解并实施。The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, ie may be located A place, or it can be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment. Those of ordinary skill in the art can understand and implement without deliberate labor.
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件。基于这样的理解,上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品可以存储在计算机可读存储介质中,如ROM/RAM、磁碟、光盘等,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。 Through the description of the above embodiments, those skilled in the art can clearly understand that the various embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on such understanding, the above-described technical solutions may be embodied in the form of software products in essence or in the form of software products, which may be stored in a computer readable storage medium such as ROM/RAM, magnetic Discs, optical discs, etc., include instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the methods described in various embodiments or portions of the embodiments.
最后应说明的是:以上实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。 It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not limited thereto; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that The technical solutions described in the foregoing embodiments are modified, or the equivalents of the technical features are replaced. The modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (13)

  1. 一种个性化视频推荐方法,其特征在于,包括:A personalized video recommendation method, comprising:
    获取与用户标识对应的用户画像特征;Obtaining a user portrait feature corresponding to the user identifier;
    根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;Obtaining a video feature model corresponding to the user portrait feature according to the user portrait feature;
    根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。And according to the video feature model corresponding to the user portrait feature, the video matching the video feature model corresponding to the user portrait feature is obtained from the video database and recommended.
  2. 根据权利要求1所述的方法,其特征在于,获取与用户标识对应的用户画像特征之前还包括:The method according to claim 1, wherein before the acquiring the user portrait feature corresponding to the user identifier, the method further comprises:
    根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征点和/或用户历史搜索的视频特征点,所述视频特征点包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率;Obtaining, according to the user identifier, user history behavior data corresponding to the user identifier, where the user history behavior data includes a video feature point played by the user history and/or a video feature point of the user history search, where the video feature point includes The region to which the video belongs, the type of video, the channel of the video, the time of video release, and/or the video clickthrough rate;
    根据与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区。Calculating, according to the user historical behavior data corresponding to the user identifier, a user portrait feature corresponding to the user identifier, where the user portrait feature includes a user age, gender, occupation, a channel that has watched the video, and a type of the viewed video. , watched the video's tags, and/or the area where the video was viewed.
  3. 根据权利要求1所述的方法,其特征在于,根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型包括:The method according to claim 1, wherein the acquiring a video feature model corresponding to the user portrait feature according to the user portrait feature comprises:
    根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。Determining, according to the user image feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point, and obtaining a video feature model corresponding to the user portrait feature.
  4. 根据权利要求3所述的方法,其特征在于,根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐,包括:The method according to claim 3, wherein the video matching the video feature model corresponding to the user portrait feature is obtained from the video database and recommended according to the video feature model corresponding to the user portrait feature, including :
    根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集; Obtaining, according to a video feature model corresponding to the user portrait feature, a video candidate set that matches a video feature model corresponding to the user portrait feature from a video database;
    根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。And sorting the video of the video candidate set according to the weight of each video feature point in the video feature model.
  5. 根据权利要求4所述的方法,其特征在于,还包括:The method of claim 4, further comprising:
    根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。And filtering, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule.
  6. 一种个性化视频推荐装置,其特征在于,包括:A personalized video recommendation device, comprising:
    第一获取模块,用于获取与用户标识对应的用户画像特征;a first acquiring module, configured to acquire a user portrait feature corresponding to the user identifier;
    第二获取模块,用于根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;a second acquiring module, configured to acquire a video feature model corresponding to the user portrait feature according to the user portrait feature;
    第三获取模块,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。And a third acquiring module, configured to acquire a video matching the video feature model corresponding to the user portrait feature from the video database according to a video feature model corresponding to the user portrait feature, and perform recommendation.
  7. 根据权利要求6所述的装置,其特征在于,还包括:The device according to claim 6, further comprising:
    第四获取模块,用于根据所述用户标识,获取与所述用户标识对应的用户历史行为数据,所述用户历史行为数据包括用户历史播放的视频特征点和/或用户历史搜索的视频特征点,所述视频特征点包括视频所属地区、视频类型、视频的频道、视频发布时间和/或视频点击率;a fourth obtaining module, configured to acquire, according to the user identifier, user historical behavior data corresponding to the user identifier, where the user historical behavior data includes a video feature point played by the user history and/or a video feature point of the user history search The video feature point includes a region to which the video belongs, a video type, a channel of the video, a video publishing time, and/or a video click rate;
    计算模块,用于根据所述第四获取模块获取的与所述用户标识对应的用户历史行为数据,计算得到与所述用户标识对应的用户画像特征,所述用户画像特征包括用户年龄、性别、职业、观看过视频的频道、观看过视频的类型、观看过视频的标签和/或观看过视频的地区。a calculation module, configured to calculate, according to the user historical behavior data corresponding to the user identifier acquired by the fourth acquiring module, a user portrait feature corresponding to the user identifier, where the user portrait feature includes a user age, gender, Occupation, channels that have watched videos, types of videos watched, tags that have watched videos, and/or areas where videos have been viewed.
  8. 根据权利要求6所述的装置,其特征在于,所述第二获取模块具体用于:The device according to claim 6, wherein the second obtaining module is specifically configured to:
    根据所述用户画像特征,确定与所述用户画像特征对应的多个视频特征点以及每个视频特征点对应的权重,得到与所述用户画像特征对应的视频特征模型。Determining, according to the user image feature, a plurality of video feature points corresponding to the user portrait feature and weights corresponding to each video feature point, and obtaining a video feature model corresponding to the user portrait feature.
  9. 根据权利要求8所述的装置,其特征在于,所述第三获取模块包括: The device according to claim 8, wherein the third obtaining module comprises:
    获取单元,用于根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频候选集;An acquiring unit, configured to acquire, according to a video feature model corresponding to the user portrait feature, a video candidate set that matches a video feature model corresponding to the user portrait feature from a video database;
    排序单元,用于根据所述视频特征模型中各视频特征点的权重,对所述视频候选集中视频进行排序。a sorting unit, configured to sort the videos of the video candidate set according to weights of video feature points in the video feature model.
  10. 根据权利要求9所述的装置,其特征在于,还包括:The device according to claim 9, further comprising:
    过滤模块,用于根据预设的视频过滤规则中包括的视频特征点,将所述视频候选集中符合所述视频过滤规则中的视频特征点的视频过滤掉。The filtering module is configured to filter, according to the video feature points included in the preset video filtering rule, the video candidate set that matches the video feature points in the video filtering rule.
  11. 一种个性化视频推荐电子设备,其特征在于,包括:A personalized video recommendation electronic device, comprising:
    至少一个处理器;以及,At least one processor; and,
    与所述至少一个处理器通信连接的存储器;其中,a memory communicatively coupled to the at least one processor; wherein
    所述存储器存储有可被所述一个处理器执行的指令,所述指令被所述至少一个处理器执行,以使所述至少一个处理器能够:The memory stores instructions executable by the one processor, the instructions being executed by the at least one processor to enable the at least one processor to:
    获取与用户标识对应的用户画像特征;Obtaining a user portrait feature corresponding to the user identifier;
    根据所述用户画像特征,获取与所述用户画像特征对应的视频特征模型;Obtaining a video feature model corresponding to the user portrait feature according to the user portrait feature;
    根据与所述用户画像特征对应的视频特征模型,从视频数据库中获取与所述用户画像特征对应的视频特征模型匹配的视频并进行推荐。And according to the video feature model corresponding to the user portrait feature, the video matching the video feature model corresponding to the user portrait feature is obtained from the video database and recommended.
  12. 一种非暂态计算机可读存储介质,其特征在于,所述非暂态计算机可读存储介质存储计算机指令,所述计算机指令用于使所述计算机执行权利要求1-5任一所述方法。A non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores computer instructions for causing the computer to perform the method of any of claims 1-5 .
  13. 一种计算机程序产品,所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序,所述计算机程序包括程序指令,当所述程序指令被计算机执行时,使所述计算机执行权利要求1-5任一所述方法。 A computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to execute The method of any of claims 1-5.
PCT/CN2016/101087 2016-04-18 2016-09-30 Personalized video recommendation method and device WO2017181612A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201610244378.5A CN105930425A (en) 2016-04-18 2016-04-18 Personalized video recommendation method and apparatus
CN201610244378.5 2016-04-18

Publications (1)

Publication Number Publication Date
WO2017181612A1 true WO2017181612A1 (en) 2017-10-26

Family

ID=56838385

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/101087 WO2017181612A1 (en) 2016-04-18 2016-09-30 Personalized video recommendation method and device

Country Status (2)

Country Link
CN (1) CN105930425A (en)
WO (1) WO2017181612A1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108102A (en) * 2018-01-02 2018-06-01 联想(北京)有限公司 Image recommendation method and electronic equipment
CN112468853A (en) * 2020-11-26 2021-03-09 未来电视有限公司 Television resource recommendation method and device, computer equipment and storage medium
CN113326434A (en) * 2021-06-15 2021-08-31 深圳前海微众银行股份有限公司 Information recommendation method, equipment and storage medium
CN113379572A (en) * 2021-06-07 2021-09-10 北京房江湖科技有限公司 House source explanation method and device, computer readable storage medium and electronic equipment

Families Citing this family (42)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105930425A (en) * 2016-04-18 2016-09-07 乐视控股(北京)有限公司 Personalized video recommendation method and apparatus
CN106658096A (en) * 2016-11-17 2017-05-10 百度在线网络技术(北京)有限公司 Method and device for pushing live program
CN106528813B (en) * 2016-11-18 2018-12-11 腾讯科技(深圳)有限公司 A kind of multimedia recommendation method and device
CN108134955A (en) * 2016-12-01 2018-06-08 深圳市优朋普乐传媒发展有限公司 A kind of method and device of program sequence
CN106682216B (en) * 2016-12-30 2020-06-19 广东惠禾科技发展有限公司 Method and equipment for sorting candidate sets based on multi-source information
CN106993226A (en) * 2017-03-17 2017-07-28 深圳市金立通信设备有限公司 A kind of method and terminal of recommendation video
CN106937172A (en) * 2017-03-23 2017-07-07 百度在线网络技术(北京)有限公司 Interactive approach and device during video playback based on artificial intelligence
CN107426604A (en) * 2017-06-16 2017-12-01 环球智达科技(北京)有限公司 A kind of information search method being applied in intelligent television system
CN107609487B (en) * 2017-08-17 2021-01-22 北京三快在线科技有限公司 User head portrait generation method and device
CN107688639A (en) * 2017-08-24 2018-02-13 努比亚技术有限公司 Using recommendation method, server and computer-readable recording medium
CN107341272A (en) * 2017-08-25 2017-11-10 北京奇艺世纪科技有限公司 A kind of method for pushing, device and electronic equipment
CN110019943B (en) * 2017-09-11 2021-09-14 中国移动通信集团浙江有限公司 Video recommendation method and device, electronic equipment and storage medium
CN107748759A (en) * 2017-09-21 2018-03-02 广州华多网络科技有限公司 Video pushing method and its equipment
CN109784961A (en) * 2017-11-13 2019-05-21 阿里巴巴集团控股有限公司 A kind of data processing method and device
CN107908740A (en) * 2017-11-15 2018-04-13 百度在线网络技术(北京)有限公司 Information output method and device
CN107885889A (en) * 2017-12-13 2018-04-06 聚好看科技股份有限公司 Feedback method, methods of exhibiting and the device of search result
CN108038237A (en) * 2017-12-27 2018-05-15 广州市云润大数据服务有限公司 A kind of information recommendation method and system
CN108243357A (en) * 2018-01-25 2018-07-03 北京搜狐新媒体信息技术有限公司 A kind of video recommendation method and device
CN108776676B (en) * 2018-02-02 2020-05-05 腾讯科技(深圳)有限公司 Information recommendation method and device, computer readable medium and electronic device
CN108573032A (en) * 2018-03-27 2018-09-25 麒麟合盛网络技术股份有限公司 Video recommendation method and device
CN110363604B (en) * 2018-04-09 2022-06-07 北京京东尚科信息技术有限公司 Page generation method and device
CN110418203A (en) * 2018-04-27 2019-11-05 郑州市富连网电子科技有限公司深圳分公司 TV programme suggesting method, television set and storage medium
CN108647293B (en) * 2018-05-07 2022-02-01 广州虎牙信息科技有限公司 Video recommendation method and device, storage medium and server
CN108833949A (en) * 2018-06-25 2018-11-16 深圳市华慧品牌管理有限公司 The video recommendation method and device of robot are accompanied for children
CN109191090A (en) * 2018-07-27 2019-01-11 阿里巴巴集团控股有限公司 Means of payment recommended method, device, equipment and computer readable storage medium
CN109658129A (en) * 2018-11-22 2019-04-19 北京奇虎科技有限公司 A kind of generation method and device of user's portrait
CN109376306B (en) * 2018-12-04 2021-04-16 挖财网络技术有限公司 Service recommendation method and system based on tag panorama
CN109583961A (en) * 2018-12-04 2019-04-05 北京唐冠天朗科技开发有限公司 A kind of method and system of identity-based identification information matching information on services
CN109714643B (en) * 2018-12-06 2021-08-13 北京达佳互联信息技术有限公司 Recommendation method, system, server and storage medium for video data
CN109410675B (en) * 2018-12-12 2021-03-12 广东小天才科技有限公司 Exercise recommendation method based on student portrait and family education equipment
CN111385596A (en) * 2018-12-29 2020-07-07 广州市百果园信息技术有限公司 Live content isolation method and device
CN110321474B (en) * 2019-05-21 2022-01-11 北京奇艺世纪科技有限公司 Recommendation method and device based on search terms, terminal equipment and storage medium
CN112312142B (en) * 2019-07-31 2022-07-05 北京沃东天骏信息技术有限公司 Video playing control method and device and computer readable storage medium
CN110557659B (en) * 2019-08-08 2021-06-18 北京达佳互联信息技术有限公司 Video recommendation method and device, server and storage medium
CN110598016B (en) * 2019-09-11 2021-08-17 腾讯科技(深圳)有限公司 Method, device, equipment and medium for recommending multimedia information
CN110602514B (en) * 2019-09-12 2022-04-01 腾讯科技(深圳)有限公司 Live channel recommendation method and device, electronic equipment and storage medium
CN110996142B (en) * 2019-11-08 2021-12-07 北京奇艺世纪科技有限公司 Video recall method and device, electronic equipment and storage medium
CN110990695A (en) * 2019-11-22 2020-04-10 厦门美柚股份有限公司 Recommendation system content recall method and device
CN111163342A (en) * 2020-01-14 2020-05-15 恒信东方文化股份有限公司 Intelligent interaction system and method thereof
CN111552888A (en) * 2020-05-29 2020-08-18 腾讯科技(深圳)有限公司 Content recommendation method, device, equipment and storage medium
CN111741104B (en) * 2020-06-18 2021-10-08 腾讯科技(深圳)有限公司 Method for determining response message, method for configuring response message, device, equipment and storage medium
CN112347302A (en) * 2020-11-06 2021-02-09 四川长虹电器股份有限公司 Video recall method based on inverted index

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102999493A (en) * 2011-09-08 2013-03-27 百度在线网络技术(北京)有限公司 Method and device for achieving video resource recommendation
CN103714130A (en) * 2013-12-12 2014-04-09 深圳先进技术研究院 Video recommendation system and method thereof
CN104063481A (en) * 2014-07-02 2014-09-24 山东大学 Film individuation recommendation method based on user real-time interest vectors
CN104469430A (en) * 2014-12-24 2015-03-25 武汉泰迪智慧科技有限公司 Video recommending method and system based on context and group combination
CN104834686A (en) * 2015-04-17 2015-08-12 中国科学院信息工程研究所 Video recommendation method based on hybrid semantic matrix
CN105930425A (en) * 2016-04-18 2016-09-07 乐视控股(北京)有限公司 Personalized video recommendation method and apparatus

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012084025A1 (en) * 2010-12-21 2012-06-28 Telefonaktiebolaget Lm Ericsson (Publ) Method and apparatus for providing temporal context for recommending content for consumption by a user device
CN104331459B (en) * 2014-10-31 2018-07-06 百度在线网络技术(北京)有限公司 A kind of network resource recommended method and device based on on-line study
CN105354277B (en) * 2015-10-30 2020-11-06 中国船舶重工集团公司第七0九研究所 Recommendation method and system based on recurrent neural network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102999493A (en) * 2011-09-08 2013-03-27 百度在线网络技术(北京)有限公司 Method and device for achieving video resource recommendation
CN103714130A (en) * 2013-12-12 2014-04-09 深圳先进技术研究院 Video recommendation system and method thereof
CN104063481A (en) * 2014-07-02 2014-09-24 山东大学 Film individuation recommendation method based on user real-time interest vectors
CN104469430A (en) * 2014-12-24 2015-03-25 武汉泰迪智慧科技有限公司 Video recommending method and system based on context and group combination
CN104834686A (en) * 2015-04-17 2015-08-12 中国科学院信息工程研究所 Video recommendation method based on hybrid semantic matrix
CN105930425A (en) * 2016-04-18 2016-09-07 乐视控股(北京)有限公司 Personalized video recommendation method and apparatus

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108102A (en) * 2018-01-02 2018-06-01 联想(北京)有限公司 Image recommendation method and electronic equipment
CN112468853A (en) * 2020-11-26 2021-03-09 未来电视有限公司 Television resource recommendation method and device, computer equipment and storage medium
CN113379572A (en) * 2021-06-07 2021-09-10 北京房江湖科技有限公司 House source explanation method and device, computer readable storage medium and electronic equipment
CN113326434A (en) * 2021-06-15 2021-08-31 深圳前海微众银行股份有限公司 Information recommendation method, equipment and storage medium

Also Published As

Publication number Publication date
CN105930425A (en) 2016-09-07

Similar Documents

Publication Publication Date Title
WO2017181612A1 (en) Personalized video recommendation method and device
US10832738B2 (en) Computerized system and method for automatically generating high-quality digital content thumbnails from digital video
CN106021586B (en) Information processing method and server
KR101571443B1 (en) Video Recommendation based on Video Co-occurrence Statistics
CN106326391B (en) Multimedia resource recommendation method and device
WO2017133343A1 (en) Picture processing method and apparatus, and electronic device
TW201447797A (en) Method and system for multi-phase ranking for content personalization
CN112074857A (en) Combining machine learning and social data to generate personalized recommendations
CN109033149B (en) Information recommendation method and device, server and storage medium
US20170024389A1 (en) Method and system for multimodal clue based personalized app function recommendation
US10387915B2 (en) Digital magazine recommendations by topic
US20190278819A1 (en) Systems and methods of providing recommendations of content items
CN102165441A (en) Method, system, and apparatus for ranking media sharing channels
US10482142B2 (en) Information processing device, information processing method, and program
Yamasaki et al. Social popularity score: Predicting numbers of views, comments, and favorites of social photos using only annotations
WO2020238502A1 (en) Article recommendation method and apparatus, electronic device and storage medium
WO2022041979A1 (en) Information recommendation model training method and related device
WO2016165414A1 (en) Method and device for push information
CN107943910B (en) Personalized book recommendation method based on combined algorithm
CN111259192A (en) Audio recommendation method and device
Li et al. A rank aggregation framework for video multimodal geocoding
WO2021155691A1 (en) User portrait generating method and apparatus, storage medium, and device
US20170185690A1 (en) System and method for providing content recommendations based on personalized multimedia content element clusters
US20160132771A1 (en) Application Complexity Computation
CN111242239A (en) Training sample selection method and device and computer storage medium

Legal Events

Date Code Title Description
NENP Non-entry into the national phase in:

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16899197

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 16899197

Country of ref document: EP

Kind code of ref document: A1