Bigram proximity matrix
Angel R. Martinez · Wiley Interdisciplinary Reviews Computational Statistics · 2010
Abstract A transformation of the text stream called a bigram proximity matrix (BPM) has been developed. The BPM is used to encode free‐form text so computational techniques can be applied to this type of information resource. For example, one could classify the encoded documents using k nearest neighbor (k‐NN) discrimination, group the documents according to their topic, search for latent topics, and more. The hope is that encoding text documents using the BPM will preserve the meaning better than the bag‐of‐words method commonly used in information retrieval and is easier to execute than other natural language processing applications. In this article, we provide a brief introduction to natural language processing and definitions for some of the concepts. Next, we define the BPM and give some examples of how it has been used. We then provide a weighted variant of the BPM and conclude with extensions to this type of document encoding. Copyright © 2010 John Wiley & Sons, Inc. This article is categorized under: Statistical Learning and Exploratory Methods of the Data Sciences > Text Mining