ContentWise Impressions

ContentWise Impressions 数据集是来自 Over-The-Top 媒体服务的电影和电视剧的隐式交互和印象的集合，该服务通过 Internet 提供其媒体内容。该数据集与其他已经可用的多媒体推荐数据集的区别在于印象的可用性，即向用户显示的推荐、其大小以及开源。数据集中的项目表示服务提供给用户的多媒体内容，并由匿名数字标识符表示。这些项目是指属于四个相互排斥的类别的电视和电影产品：电影、电影和连续剧剪辑、电视电影或节目以及电视剧集。交互表示用户对服务中的项目执行的操作，并与发生时的时间戳相关联。交互包含印象的标识符，但推荐来自服务提供商添加的行的情况除外。交互分为四种不同的类型：查看、详细信息、评级和购买。印象是指呈现给用户并由其系列标识的推荐项目。展示次数由数字标识符、屏幕上的列表位置、推荐列表的长度和推荐系列标识符的有序列表组成，其中最相关的项目位于首位。

The ContentWise Impressions dataset is a collection of implicit interactions and impressions of films and television series from an Over-The-Top (OTT) media service that delivers its media content over the Internet. What distinguishes this dataset from other existing multimedia recommendation datasets is the availability of impression data, including the recommendations shown to users, the scale of these recommendations, and the dataset's open-source nature. Items in the dataset represent multimedia content provided to users by the service, and are identified by anonymous numeric identifiers. These items correspond to television and film products belonging to four mutually exclusive categories: movies, movie and series clips, television films or specials, and television episodes. Interactions represent actions performed by users on items within the service, and are associated with timestamps marking the time of occurrence. Interactions include the identifier of the corresponding impression, except for cases where the recommendation originates from rows added by the service provider. Interactions are divided into four distinct types: view, details, rating, and purchase. An impression refers to the recommended items presented to a user, and is identified by its series identifier. Each impression consists of a numeric identifier, the on-screen list position, the length of the recommendation list, and an ordered list of recommendation series identifiers, with the most relevant items placed first.

搜集汇总

背景与挑战

背景概述

ContentWise Impressions是一个开源数据集，收集了OTT媒体服务中电影和电视剧的隐式交互与印象数据，包括项目、用户交互（如查看和评级）以及推荐列表。该数据集的特点在于提供了详细的印象信息，如推荐位置和列表长度，区别于其他多媒体推荐数据集。