An effect of term selection and expansion for classifying short documents

Abstract

Many web sites(blogs) over the Internet provide the users the possibility of sharing information like: opinions, news, even their profiles. The peculiarity of this information is that usually the description contains few words. Currently exist a great interest in developing tools that help to process this information in order to organize or categorize it, for helping decision making. Due the importance of this task, in this paper it is explored, through a set of experiments the effect of simple expansion and term selection over two Data Sets. It is applied the Absolute Term Frequency (ATF) term selection technique over this kind of documents, and it is showed that using a percentage of the terms, to represent the information, the classification result could be improved. At the end of the paper it is showed the classification phase where the document expansion could improve the number of classified instances.