Text Mining Techniques, Applications, and Issues

S.-H. Liao, P.-H. Chu, and  P.-Y. Hsiao, described that gathering, extracting, pre-processing, text transformation, feature extraction, pattern selection, and evaluation steps are part of text mining process. In addition, different widely used text mining techniques, i.e., clustering, categorization, decision tree categorization, and their application in diverse fields are surveyed. N. Zhong, Y.  Li, and  S.-T. Wu highlighted the issues in text mining applications and techniques. They discussed that dealing with unstructured text is difficult as compared to structured or tabular data using traditional mining tools and techniques. They have shown the applications of text mining process in bioinformatics, business intelligence and national security system. Natural language processing and entity recognition techniques has reduced the issues that occur during text mining process. However, there exist issues which need attention.

A. Henriksson, H. Moen, M. Skeppstedt, V. Daudaravicius, and M. Duneld explored MEDLINE biomedical database by integrating a framework for named entity recognition, classification of text, hypothesis generation and testing, relationship and synonym extraction, extract abbreviations. This new framework helps to eliminate unnecessary details and extract valuable information. B. Laxman and D. Sujatha analyzed the text using text mining patterns and showed term based approaches cannot analyze synonyms and polysemy properly. Moreover, a prototype model was designed for specification of patterns in terms of assigning weight according to their distribution. This approach helps to enhance the efficiency of text mining process. C. P. Chen and C.-Y. Zhang presented a crime detection system using text mining tools and relation discovery algorithm was designed to correlate the term with abbreviation.

R. Rajendra and V. Saransh presented a top down and bottom up approach for web based text mining process. To combine the similar text documents, they apply k-mean clustering technique for bottom up partitioning. To find out the similarity within the document TF-IDF (Term Frequency- Inverse Document Frequency) algorithm has been used to find information regarding specific subjects. K. Sumathy and M. Chidambaram gave an overview of applications, tools and issues arises to mine the text. They discussed that documents may be structured, semi structured or unstructured and extracting useful information is a tiresome task. They presented a generic framework for concept based mining which can be visualized as text refinement and knowledge distillation phases. The intermediate form of entity representation mining depends on specific domain

P.J. Joby and J. Korra presented innovative and efficient pattern discovery techniques. They used the pattern evolving and discovering techniques to enhance the effectiveness of discovering relevant and appropriate information. They performed BM25 and vector support machine based filtering on router corpus volume 1 and text retrieval conference data to estimate the effectiveness of the suggested technique. Z. Wen,  T. Yoshida, and X.  Tang performed various experiments of classification using multi-word features on the text. They proposed a hand-crafted method to extract multi-word features from the data set. To classify and extract multi-word text they divide text into linear and nonlinear polynomial form in support of vector machine that improve the effectiveness of the extracted data.