Abstract
To attract first attention at a glance, news titles are often short and contain important abstract information of web news. Topic feature extraction of web news title can greatly help news processing system to improve efficiency and accuracy before process whole news text. After segmentation and tagging, some words are wrongly truncated into discontinuous characters and phrases are into separate words as well. This paper proposes a topic feature extracting model from Chinese web news titles on phrase granularity. Titles are truncated into tagged key words before using frequent patterns to combine words into phrases, which are topic features. We conduct experimental studies on corpus of Chinese news titles between March 2011 and June 2011. The result showed that our topic extraction approach can yield quite reasonable topic feature phrases.