Author clustering with the Aid of a Simple Distance Measure

A simple distance measure has been applied to the author clustering problem to determine which documents are written by the same author. This simple distance measure works with the probability distribution of character sequences of a document, making it insensitive to language differences. The top most frequent features k, where k is chosen to be 300, determine the distribution where punctuation is present. Also, the uppercase letters are transformed to lowercase symbols, while a threshold of 3.0 remains for the symmetric distance score. In addition, character 2-grams are chosen due to their best outcomes. Using the BCubed F-score provided, it achieves a score of 0.54 on the training set and a score of 0.53 on the test set with a relative low MAP score. Obtaining clusters from links still shows problems.