内容来源:“Recommender Systems Specialization” course of 明尼苏达大学 (University of Minnesota) at Coursera
Learning Objectives
- To understand the different types of recommender systems
- A framework for analyzing recommender systems in general(一个分析推荐系统的通用框架)
- A specific overview of different recommendation algorithms
Analytical Framework
- Dimentions of Analysis
– Domain
– Purpose
– Recommendation Context
– Whose Opinions
– Personalization Level
– Privacy and Trustworthiness
– Interfaces
– Recommendation Algorithms
Domains of Recommendation
明确将要推荐的内容是什么?
- Content to Commerce and Beyond
– News, information, “text”
– Products, vendors, bundles (打包销售)
– Matchmaking (other people)
– Sequences (e.g., music playlists) (推荐序列,考虑items的顺序) - One particularly interesting property (不同的领域,如何对待已经经历过的item)
– New items (e.g., movies, books, …) (对于电影/书籍,一般只推荐新的、未经历过的item)
– Re-recommend old ones (e.g., groceries, music) (可能会重复推荐已听过的歌或购买过的商品)
Purposes of Recommendation
- The recommendations themselves
– Sales
– Information
(引导用户来做出特定的选择,获取利益) - Education of user/customer (例如软件指令推荐,目的是通过教育我们指令信息来提高软件使用效率)
- Build a community of users/customers
around products or content (例如豆瓣,通过给出电影列表以及它们的相关信息,来帮助做出选择,而不是引导我们做出特定的决定,同时试图让用户加入社区提供反馈,例如电影评价,来优化社区内容)
Recommendation Context
- What is the User doing at the time of
recommendation?
– Shopping
– Listening to Music
– Hanging out with other people(recommend for group rather than for individuals) - How does the context constrain the
recommender?
– Groups, automatic consumption (vs. suggestion), level of attention, level of interruption?
(例如Hanging out with other people,推荐系统应该recommend for group rather than for individuals。the context affects the level of attention or the degree to which the user is willing to accept an interruption.)
Whose Opinion?
Recommender systems in the broad sense are based on somebody’s opinion, Then whose opinion?
- Experts(一般用在需专业性知识的网站,例白酒推荐)
- Ordinary “phoaks”(一般用在普适性场景,需要大家的意见。例如知乎)
- People like you(一般用在商业场景,需要推荐用户喜欢的内容。例如淘宝)
Personalization Level
- Generic / Non-Personalized
– Everyone receives same recommendations - Demographic
– Matches a target group(例如给男性和女性推荐不同内容) - Ephemeral
– Matches current activity(只要行为相同,不同的用户可能得到相同的推荐。例如购买了键盘,会同时给你推荐鼠标或键盘膜) - Persistent
– Matches long-term interests
Privacy and Trustworthiness
•- Who knows what about me?
– Personal information revealed(有多少个人信息是我需要提供的)
– Identity(系统必须知道我是谁吗?)
– Deniability of preferences(我有方法能够拒绝部分偏好吗)
– Information disclosure(我能确保我的信息不会被推荐系统所泄露吗)
- Is the recommendation honest?
– Biases built-in by operator(操作员内置的偏差是多少?) - “business rules”(商业规则,例如商家永远只会推荐有货的商品给你)
– Vulnerability to external manipulation(易受外部操纵,例如水军刷票行为)
– Transparency of “recommenders”; Reputation(推荐过程透明化以及明确用户需要提供数据的程度,能够让用户觉得是否相信推荐结果,提高用户满意度)
Interfaces
- Types of Output
– Predictions
– Recommendations
– Filtering
– Organic vs. explicit presentation - Agent/Discussion Interface(Different styles of interface as to how the system is presenting itself,输出不同决定系统所呈现的界面风格不同,例如是像一个代理般提供帮助,还是像一个对话般,能够接受你的反馈。)
- Types of Input
– Explicit
– Implicit
Recommendation Algorithms
- Non-Personalized Summary Statistics
- Content-Based Filtering
– Information Filtering
– Knowledge-Based - Collaborative Filtering
– User-User
– Item-Item
– Dimensionality Reduction - Others
– Critique / Interview Based Recommendations
– Hybrid Techniques