A framework for diversifying recommendation lists by user interest expansion

Abstract

Recommender systems have been widely used to discover users’ preferences and recommend interesting items to users during this age of information load. Researchers in the field of recommender systems have realized that the quality of a top-N recommendation list involves not only relevance but also diversity. Most traditional recommendation algorithms are difficult to generate a diverse item list that can cover most of his/her interests for each user, since they mainly focus on predicting accurate items similar to the dominant interests of users. Additionally, they seldom exploit semantic information such as item tags and users’ interest labels to improve recommendation diversity. In this paper, we propose a novel recommendation framework which mainly adopts an expansion strategy of user interests based on social tagging information. The framework enhances the diversity of users’ preferences by expanding the sizes and categories of the original user-item interaction records, and then adopts traditional recommendation models to generate recommendation lists. Empirical evaluations on three real-world data sets show that our method can effectively improve the accuracy and diversity of item recommendation.

Online social image ranking in diversified preferences

“…The insights may help to design better result ranking strategies and evaluation metrics. Besides, diversity is also enforced in other applications such as image recommendation [32], movie recommendation [33], and general purpose recommendation tasks [34]. Similar as the retrieval task, it has been shown that more diversity can bring user with better experience.…”

Empirical Research on the Impact of Personalized Recommendation Diversity

“…While personalized recommendation diversity reduces accuracy to a certain degree but improves users' subjective evaluation of the recommendation system [11]. Some "dark information" that users may be interested in has been ignored, thus recommendation are unable to meet their real needs [12].…”

Balancing accuracy and diversity in recommendations using matrix completion framework

“…Most existing works evaluate the capability of a RS to diversify recommendations on two criteria -individual diversity (ID) [19] and aggregate diversity (AD) [16]. Individual diversity [20] refers to the breadth or range of (varied) items suggested to a given user.…”

“…In addition to exploiting the available rating values, we also use item metadata (in our case movie genres) to explicitly enforce desired diversifying criteria. To the best of our knowledge, there are very few works in the past [19,36,38] which exploit available secondary data (metadata) to establish diversity-accuracy tradeoff. In [19], authors use user tagging (user metadata) information to develop expansion strategy for user interest and suggest diversified items.…”

“…To the best of our knowledge, there are very few works in the past [19,36,38] which exploit available secondary data (metadata) to establish diversity-accuracy tradeoff. In [19], authors use user tagging (user metadata) information to develop expansion strategy for user interest and suggest diversified items. In [36], taxonomic description of books from amazon is exploited to introduce diversity in recommendation for the book crossing dataset via topic diversification.…”