Feed Ranking Refinement with Similitary Distribution in Blog Distillation

摘要

Blog Distillation is the process of finding a blog with a principle and recurring interest. In this paper, two baselines are used to validate the results of our experiments. A set of features of individual feed is firstly constructed by decision tree to represent the similarity distribution of every feed against certain interest. Features are then selected by computing their centroid distances to standard centroids of relevant feeds and irrelevant feeds. Later, SVM classifier is used to predict and re-rank the top 250 results of two baselines. The result shows that our algorithm can effectively present the feeds' similarity distribution and re-rank them into a new order which has much better MAP.