Social Network Analysis of Alice in Wonderland

TLDR: This paper annotates Lewis Carroll's Alice in Wonderland using a well-defined annotation scheme which has been used in previous computational models that extract social events from news articles and builds novel types of networks in which links between characters are different types of social events.

Abstract: We present a network analysis of a literary text, Alice in Wonderland. We build novel types of networks in which links between characters are different types of social events. We show that analyzing networks based on these social events gives us insight into the roles of characters in the story. Also, static network analysis has limitations which be- come apparent from our analysis. We propose the use of dynamic network analysis to over- come these limitations. tify these limitations, few have done so with a strict and specific rubric for categorizing interactions. In this paper, we annotate Lewis Carroll's Alice in Wonderland using a well-defined annotation scheme which we have previously developed on newswire text Agarwal et al. (2010). It is well suited to deal with the aforementioned limitations. We show that using different types of networks can be useful by al- lowing us to provide a model for determining point- of-view. We also show that social networks allow characters to be categorized into roles based on how they function in the text, but that this approach is limited when using static social networks. We then build and visualize dynamic networks and show that static networks can distort the importance of char- acters. By using dynamic networks, we can build a fuller picture of how each character works in a liter- ary text. Our paper uses an annotation scheme that is well- defined and has been used in previous computational models that extract social events from news articles (Agarwal and Rambow, 2010). This computational model may be adapted to extract these events from literary texts. However, the focus of this paper is not to adapt the previously proposed computational model to a new domain or genre, but to first demon- strate the usefulness of this annotation scheme for the analysis of literary texts, and the social networks derived from it. All results reported in this paper are based on hand annotation of the text. Further- more, we are investigating a single text, so that we do cannot draw conclusions about the usefulness of our methods for validating theories of literature. We summarize the contributions of this paper: • We manually extract a social network from Al-