Classification of Interviews - A Case Study on Cancer Patients

摘要

With the rapid expansion of Web 2.0, a variety of documents abound online. Thus, it is important to find methods that can annotate and organize documents in meaningful ways to expedite the search process. A considerable amount of research on document classification has been conducted. However, this paper introduces the classification of interviews of cancer patients into several cancer diseases based on the features collected from the corpus. We have developed a corpus of 727 interviews collected from a web archive of medical articles. The TF-IDF features of unigram, bigram, trigram and emotion words as well as the SentiWordNet and Cosine similarity features have been used in training and testing of the classification systems. We have employed three different classifiers like £-NN, Decision Tree and Naive Bayes for classifying the documents into different classes of cancer. The experimental results obtain maximum accuracy of 99.31% tested on 73 documents of the test data.