Personal tools
You are here: Home Papers and abstracts Keynote speech

Keynote speech

Information Presentation of Ranked Datasets, Sihem Amer-Yahia, Yahoo! Research

We examine the question of information presentation of large ranked datasets in two contexts: recommendations in social content sites and online search of large structured datasets. We argue that despite the intuition that structure could be used to improve linear ranking, it is not always available, or sufficient, to help users explore ranked content, online.

On social sites such as Facebook and Yahoo! Travel, effective content recommendation is becoming increasingly important. A well-known side effect of maximizing accuracy is over-specialization. We discuss the limitations of using structure for diversifying recommendations and formalize explanation-based diversity. We show that our algorithms achieve a good compromise between accuracy and heterogeneity on Yahoo! Movies datasets.

In online applications such as and, users define structured profiles which are used to retrieve a ranked list of matches. We argue here that neither linear ranking nor attribute-based grouping is adequate for effective data exploration. We formalize rank-aware clustering and evaluate our algorithm over large datasets from Yahoo! Personals.

Document Actions
« June 2019 »