CS 224 w Final Report : Community-Based Yelp Personalization

People use Yelp to search for everything and get recommendations. When we are reading a review, we want to know if the review itself and the reviewer are credible. Thus, it helps if we can know the credibility of the reviewer. In turn, a reviewer might be motivated to write more high-quality reviews. The user scores can be incorporated into calculating new (and hopefully more accurate) scores for the businesses. In addition, people in different communities have very different opinions about a single business. For example, a community of Chinese elderly people and another one of American youngsters can have totally different comments for a Chinese restaurant. However, Yelp currently only offers one single score for a business. With the goal of better reflecting different tastes from different communities, we want to generate communities from all users and put users that have similar tastes into one community. Then for all users in one community, we will put some bias on the new score that we generated for each business based on the other users review and their credibility in the same community. In this way, score of a business is personalized to each community. This community-based score is expected to be most informative and can be utilized to make personalized recommendation.