Abstract

In this presentation, I’ll give a short overview of work of mine in the area of Web Science.

First, we’ll discuss how using web search result clicks on political blogs can be used to annotate queries with a political leaning. Combining this with trend detection, we can algorithmically obtain a summary of political issues in a given week. We further enrich this data set by mapping queries to entities and to fact-checked statements to study the relation of what a query is about and whether it relates to a true statement. We then apply a similar methodology to hashtags on Twitter to analyze which “side” dominates the debate on a particular topic and how the association between hashtags and political leaning changes over time. http://politicalsearchtrends.sandbox.yahoo.com

Next, we’ll look at the affect that demographic variables, such as age, gender or income, have on web search behavior. We’ll show that there are distinct differences which can be exploited to (i) improve web search results and to (ii) help with query suggestions. We also show that different groups exhibit different, topic-dependent adoption patterns with certain groups issuing a query weeks before another group. This has applications to predict content relevance over time as well as to identify information hermits who are least likely to, say, search for vaccine information during a flu epidemic. http://clues.yahoo.com/analysis

Finally, I’ll briefly mention other projects from the same domain, such as using email data to track international migration flows or using social media data to study anorexia as a social disease.