A Language-Based Approach to Categorical Analysis

With the digitization of media, computers can be employed to help us with the process of classification, both by learning from our behavior to perform the task for us and by exposing new ways for us to think about our information. Given that most of our media comes in the form of electronic text, research in this area focuses on building automatic text classification systems. The standard representation employed by these systems, known as the bag-of-words approach to information retrieval, represents documents as collections of words. As a byproduct of this model, automatic classifiers have difficulty distinguishing between different meanings of a single word. This research presents a new computational model of electronic text, called a synchronic imprint, which uses structural information to contextualize the meaning of words. Every concept in the body of a text is described by its relationships with other concepts in the same text, allowing classification systems to distinguish between alternative meanings of the same word. This representation is applied to both the standard problem of text classification and also to the task of enabling people to better identify large bodies of text. The latter is achieved through the development of a visualization tool named flux that models synchronic imprints as a spring network. Thesis Advisor: Walter Bender Executive Director, MIT Media Laboratory The author gratefully thanks the Motorola Fellows Program for its support and input into the development of this research