Genre is often applied as a static classification: fiction, non-fiction, mystery, romance, biography, and so on. But the edges of genre are “blurry” (Underwood). The classification of genre can change over time and situation. Ideally, genre and all classifications could be modeled dynamically during content analysis. How can IBM’s Watson Content Analytics (WCA) help analyze genre? Here is a simple demonstration.
In WCA I created a collection of 1368 public domain novels from Open Library. For this demonstration, I obtained author metadata and expressed it as a WCA facet. I did not obtain existing genre metadata. I will demonstrate that I can use author gender to dynamically classify genre for a specific analytical question. In particular, I follow the research of Matthew Jockers and the Nebraska Literary Lab. Can genre be distinguished by the gender of the author? How is action and agency treated differently in male and female genres? This simple demonstration does not answer these questions, but shows how WCA can be used to give insight into literature.
In Figure 1, the WCA Author facet is used to filter the collection to ten male authors: Walter Scott, Robert Louis Stevenson, and others. The idea is to dynamically generate a male genre by the selection of male authors. (Simple, but note that a complex array of facets could be used to quickly define a male genre.)
In Figure 2, the WCA Parts-of-Speech analysis lists frequently used verbs in the collection susbset, the male genre: tempt, condemn, struggle. Some values might be considered action verbs, but further analysis is required.
In Figure 3, the verb “struggle” is seen in the context of its source, the Waverly novels: “the Bohemian struggled to detain Quentin”, “to struggle with the sea”. This view can be used to determine the gender of characters, the actions they are performing, and interpret agency.
In Figure 4, a new search is performed, this time filtering for female authors: Jane Austen, Maria Edgeworth, Susan Ferrier, and others. In this case, the idea is to dynamically generate a female genre by selecting female authors.
In Figure 5, the WCA Parts-of-Speech analysis lists frequently used verbs in the female genre: mix, soothe, furnish. At a glance, there is an obvious difference in quality from the verbs in the male genre.
Finally in Figure 6, the verb “furnish” is seen in the context of its source in Jane Austen’s Letters, “Catherine and Lydia … a walk to Meryton was necessary to amuse their morning hours and furnish conversation.” In this case, furnish does not refer to the literal furnishing of a house, but to the facilitation of dialog. As before, detailed content inspection is needed to analyze and interpret agency.