Book Classification Via Fuzzy Logic Jeremy Keer Project

Book Classification Via Fuzzy Logic Jeremy Keer

Project Goals • Develop a fuzzy logic rule set to classify the content of fantasy and science fiction books based upon genre and hardness • Allow a user to classify a book with cursory knowledge with the purpose of finding whether it is similar to other styles they have enjoyed • If possible, use clustering ANN for a visual output

Why Use Fuzzy Logic? • Line between genres can be blurry at times • Existing deeper classification systems tend to rely on tags that ultimately don’t offer much information on the actual content of the story

Data Set Generation • Data is gathered via questionnaires • Questions must be formulated such that answers can be given reasonably in the form of a scale • Two types of data sets used: – Shallow set answerable based largely on summary and tags – Deep set meant to be answered by someone who has read the book so that it can be accurately be classified

Classification Methods • Data is analyzed and given values on two spectrums: – Science Fiction vs Fantasy: while sometimes seeming to be a binary choice, tends to actually be much more complex – Hardness: hardest is “real life, ” becomes progressively softer as more fantastical elements are included

Expected Results • Questions work well, but would like to add more to improve classification – Finding questions that fit well to scales is the major problem • Reducing output regions of rule set should work for classification – For example, three regions on the SF/Fantasy scale and five regions for each of these for hardness