Doctor of Philosophy
Juan Luis Suárez
The digitization of several million books by Google in 2011 meant the popularization of a new kind of humanities research powered by the treatment of cultural objects as data. Culturomics, as it is called, was born, and other initiatives resonated with such a methodological approach, as is the case with the recently formed Digital Humanities or Cultural Analytics. Intrinsically, these new quantitative approaches to culture all borrow from techniques and methods developed under the wing of the exact sciences, such as computer science, machine learning or statistics. There are numerous examples of studies that take advantage of the possibilities that treating objects as data has to offer for the understanding of the human. This new data science that is now applied to the current trends in culture can also be replicated to study more traditional humanities. Led by proper intellectual inquiry, an adequate use of technology may bring answers to questions intractable by other means, or add evidence to long held assumptions based on a canon built from few examples. This dissertation argues in favor of such approach. Three different case studies are considered. First, in the more general sense of the big and smart data, we collected and analyzed more than 120,000 pictures of paintings from all periods of art history, to gain a clear insight on how the beauty of depicted faces, in the framework of neuroscience and evolutionary theory, has changed over time. A second study covers the nuances of modes of emotions employed by the Spanish Golden Age playwright Calderón de la Barca to empathize with his audience. By means of sentiment analysis, a technique strongly supported by machine learning, we shed some light into the different fictional characters, and how they interact and convey messages otherwise invisible to the public. The last case is a study of non-traditional authorship attribution techniques applied to the forefather of the modern novel, the Lazarillo de Tormes. In the end, we conclude that the successful application of cultural analytics and computer science techniques to traditional humanistic endeavours has been enriching and validating.
de la Rosa Pérez, Javier, "Making Machines Learn. Applications of Cultural Analytics to the Humanities" (2016). Electronic Thesis and Dissertation Repository. 3486.
List of creators (Ch. 2)
faces.xlsx (5046 kB)
List of faces detected (Ch. 2)
paintings.xlsx (1878 kB)
List of paintings (Ch. 2)
scores.csv (490 kB)
Scores to build classifier (Ch. 3)
characters_categories.csv (13 kB)
Annotation of characters categories (Ch. 3)
normalized_corpus.zip (5116 kB)
Normalized corpus (Ch. 4)