Files
Abstract
Topic based classification and searches have always been a hefty challenge along the corridors of data mining. Reading a large amount of articles and indentifying them of to be the same genre or precisely one subject matter is nearly impossible. With the ever popular need for refinement and quick results we have cropped up a technique to apply graph clustering and probabilistic theory along with known data mining concepts to develop a relationship between words that present high instances of existing together across a majority of documents. These words or topics as we call them form a Topic Graph. A Graph is thus a set of words with a high frequent, high probabilistic relationship amongst them. In more technical theory, it is a highly connected graph with words as nodes and relationships between these words as edges. We can apply these concepts of Topic Graphs to refine and categorize search result along with creating new Graphs if the need arises. One of the possible resulting applications should be able to provide precise and specific search answers satisfying users requests.