Clusters Merging Method for Short Texts Clustering

Under push of Mobile Internet, new social media such as microblog, we chat, question answering systems are constantly emerging. They produce huge amounts of short texts which bring forward new challenges to text clustering. In response to the features of large amount and dynamic growth of short texts, a two-stage clustering method was putted forward. This method adopted a sliding window sliding on the flow of short texts. Inside the slide window, hierarchical clustering method was used, and between the slide windows, clusters merging method based on information gain was adopted. Experiment indicated that this method is fast and has a higher accuracy.