Collective Biographies of Women: A Machine Learning Approach to Paragraph Annotation

Abstract

The Collective Biographies of Women Project (CBW) seeks to annotate a large corpus of nineteenth and twentieth century British and American biographical texts about women. These annotations, applied at the paragraph level, draw from a controlled vocabulary known as BESS, Biographical Elements and Structure Schema. The BESS vocabulary terms are grouped into five major categories – Stage Of life, Persona, Event, Topos, Discourse. The corpus is drawn from 1,270 known books, comprising around 13,000 chapters of about 8,000 women. Because manual annotation is painstaking, time-consuming, and error-prone, there is a need to automate the annotation process for the entire corpus. Using the BESS vocabulary as labels and the currently annotated paragraphs as a training set, we developed a supervised machine learning classifier to aid in this process. Employing several methods, including Logistic Regression, Random Forest and Language models, we achieved an accuracy of ∼87%. In addition to aiding in the work of annotation, we have made recommendations about further developing the BESS vocabulary.