Document Type



Doctor of Philosophy (PhD)


Computer Science

First Advisor's Name

Mark A. Finlayson

First Advisor's Committee Title

Major Advisor/Committee Chair

Second Advisor's Name

Phillip Carter

Second Advisor's Committee Title

Committee Member

Third Advisor's Name

Giri Narasimhan

Third Advisor's Committee Title

Committee Member

Fourth Advisor's Name

Monique Ross

Fourth Advisor's Committee Title

Committee Member

Fifth Advisor's Name

Ning Xie

Fifth Advisor's Committee Title

Committee Member


nlp, artificial intelligence, culture, cognition, machine learning

Date of Defense



To understand how people communicate, we must understand how they leverage shared stories and all the knowledge, information, and associations contained within those stories. I examine three classes of narrative elements that convey a wealth of cultural knowledge: Propp's morphology, motifs, and discourse structure. Propp's morphology communicates how roles and actions drive a narrative forward; motifs fill those roles and actions with specific, remarkable events; discourse groups these into a coherent structure to convey a point.

My thesis has three aims: first, to demonstrate that people can reliably detect and identify all three of these narrative elements; second, to develop automatic detectors for discourse and motifs; third, to demonstrate the deep relation between these narrative elements and other theories of narrative structure and knowledge representation that I refer to as the \textit{continuum of communication}.

The first step of my work answers two key questions about Propp's morphology by demonstrating the reliability of annotators applying Propp's scheme across a variety of experiments, in a double-blind annotation study. Additionally, I demonstrate a shortcoming in Propp's scheme, demonstrating areas in which there are elements present in the folktales he analyzed that are not part of his morphology.

The second step of my work, showing that people familiar with motifs can reliably detect when they are being used to share information and associations, approaches this problem by performing a large-scale annotation study of 21,000 examples into four categories performed by three pairs of annotators over a period of 11 weeks. I show that, in a double-blind annotation study, people familiar with the motifs had a moderate to high degree of agreement, demonstrating the reliability of humans at this task.

The third step demonstrates the reliability of applying a theory of news discourse structure to news articles via a double-blind annotation study and, using the results of this annotation, demonstrate a preliminary detector of the news discourse function of paragraphs in news articles.

The fourth step of my work, detecting motific usage automatically, consists of a large-scale pipeline that achieves moderate performance. This pipeline is the first work towards automatically detecting motific usage of motifs and beats out simple baselines while comparing favorably too and generalizing better than a simple neural network baseline system. Additionally, the pipeline uses explainable features that can be used in future work to further develop our understanding of how humans automatically detect motifs.

Finally, I describe an exploration of the broader scope of narrative elements that communicate information between individuals who share a cultural or sub-cultural background. This work is based off of a small-scale, in-lab annotation of posts from the “incel” subculture, a niche internet community with extremist elements and, at times, disturbing content. This small annotation has revealed a complex landscape encompassing fourteen categories, more than three times the number of elements as the large-scale annotation, many of which resemble the moving parts of other theories on narrative structure and cognition, including Vladimir Propp's morphology of folktales and Silvan Tomkins' script theory. I describe these relations and provide a rough continuum of the landscape of narrative communication.



Previously Published In

Yarlott, W. V. H., & Finlayson, M. A. (2016). Learning a better motif index: Toward automated motif extraction. In 7th Workshop on Computational Models of Narrative (CMN 2016). Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik.

Yarlott, W. V. H., & Finlayson, M. A. (2016). ProppML: A complete annotation scheme for proppian morphologies. In 7th Workshop on Computational Models of Narrative (CMN 2016). Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik.

Yarlott, W. V. H., Cornelio, C., Gao, T., & Finlayson, M. (2018, August). Identifying the discourse function of news article paragraphs. In Proceedings of the Workshop Events and Stories in the News 2018 (pp. 25-33).

Yarlott, W. V. H., Ochoa, A., Acharya, A., Bobrow, L., Estrada, D. C., Gomez, D., Zheng, J., McDonald, D., Miller, C., & Finlayson, M. A. (2021). Finding Trolls Under Bridges: Preliminary Work on a Motif Detector. In Proceedings of the Ninth Annual Conference on Advances in Cognitive Systems.



Rights Statement

Rights Statement

In Copyright. URI:
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).