Deutsch Intern
    Data Science Chair

    Final Paper submitted to EACL 2021: "Detecting Scenes in Fiction: A new Segmentation Task"

    01/26/2021

    We have just submitted the final version of our full paper "Detecting Scenes in Fiction: A new Segmentation Task", which has been accepted for publication at EACL 2021.

    We have just submitted the final version of our full paper "Detecting Scenes in Fiction: A new Segmentation Task", which has been accepted for publication at EACL 2021.

    In the paper, we introduce the task of detecting scenes in literary fiction, present annotation guidelines, an annotated corpus as well as baseline experiments towards automatically solving the task.

    Scene Segmentation is a challenging new task in NLP with a wide variety of possible applications.
    Analysis of long texts still tends to be very challenging. Thus, breaking them down into shorter, coherent parts can be a valuable step towards dealing with these texts, for example greatly simplifying coreference resolution.
    In itself, scenes can be used as a means of analysing the plot of a novel, as part of an author's "fingerprint" (some authors may have a tendency towards more, shorter scenes, while others write few longer scenes) or as a building block for book recommendations.

    We hope that our paper sparks research towards this task in the future and are currently working on a follow-up paper, developing better automatic solutions.

    You can read a preprint version of the paper here.

    Back