Analyses of dramatic network simulations by using Markov chains

Official URL: https://dx.doi.org/10.1109/BESC53957.2021.9635385

Progress in computational linguistics has made artificial literature (ALit) more and more possible. ALit is an area of study which aims to enable the computer to generate structural literary works, just like the reference texts. This similarity may be a direct linguistic similarity or a structural similarity. The aim of this research is to extract the structural characteristics of dramatic works, by studying the great theatrical pieces. This is accomplished by analyzing texts using the speech sequence simulation technique. As we further analyze the text and make them quantifiable, we get closer to generate structured literary texts. To do this, the speech sequences of the characters of the play Antigone of Sophocles are simulated and presented. It's called a dramatic network. When a scene starts, the characters join the piece through conversations. If we enumerate the order of character speeches in a scene, it is observed that the pieces form a Markov chain. Each play may be considered a stochastic process. If each character is taken as a state, the transition from one state to another is the transition from one character's speech to the next. In this theoretical framework, the probability of occurrence in each state is calculated and the probabilities of transition between states are demonstrated. Finally, if this play is simulated iterative, we can obtain the histogram of simulation length which is approximately the power law. As a result, the analyses made by considering only the speech sequences have structural importance of the simulations that we claim the alpha characteristics. We observe a similar α characteristic for different theatrical pieces with low-pass filtering in the frequency domain. We claim an alpha approximation among the plays.