Recommendation strategies based on user-generated data

The challenge of discovering useful information from data has drawn much attention from researchers and scientists. In this work, we further explore the challenge by studying potential strategies and difficulties of harnessing human-generated data (content and activity). Facing the challenge of big data, our goal is to help web users satisfy their needs by providing recommendations and to guide them through the overwhelming amount of information on the internet. Three aspects are targeted: crowd-sourcing, detection of trending topics, and the win-win principle. Correspondingly, three main applications are proposed: recommending similar items, popular items, and profitable items. For crowd-sourcing, we demonstrate the use of social tag data to find similar items. For detection of trending topics, we study bias sampling strategies to track trending news from accredited experts on Twitter. For the win-win principle, we investigate how to design query suggestions that maximize value to all interested parties. When a user is searching for solutions in a unfamiliar domain and is having a difficult time in forming effective queries, our work harnesses user-generated data to offer guidance, leading them to explore the sea of information more efficiently.