PicAChoo: A Tool for Customizable Feature Extraction Utilizing Characteristics of Textual Data

Abstract

Although documents have hundreds of thousands of unique words, only a small number of words are significantly useful for intelligent services. For this reason, feature extraction has become an important issue to be addressed in various fields, such as information retrieval, text mining, pattern recognition, etc. Numerous supporting tools for feature extraction are available, but most of them deal with text as a simple literal. Unfortunately, text is not just a literal, but a semantically significant unit including linguistic characteristics. So, we need customized extraction methods that consider the characteristics of source documents. PicAChoo stands for 'Pick And Choose', and it provides an environment which enables feature extraction methods using the structure of sentences and the part-of-speech information of words. Moreover, we suggest dynamic composition of different extraction methods without hard-coding.