Network Analysis by Automated Request Annotations

This repository provides a python package which automates the detection of speech requests in German dramas and uses these annotations for network analysis.

Ressources

gerdracor_request_annotations/

Speech request annotations for each <sp> element in the TEI encoded plays of GerDraCor. Each <sp> element is represented in this structure:

[
    {
        "speaker":"danton",
        "speeches":"Bürger Biron, vormals Herzog von Lauzun, Du bist von der Rheinarmee versetzt zur Armee der Vendée und erhältst ein Armeekorps.",
        "scene":1,
        "request_classification":"False"
    },
    {
        "speaker":"biron",
        "speeches":"In die Vendée?! Pfui Teufel! Gegen Landsleute!",
        "scene":1,
        "request_classification":"True"
    }
]

These annotations are created by the requestnetpy.request_clf (see below).

network_metrics/

Network stats based on the request annotations.

requestnetpy/

.gold_annotation_preprocessing

Preprocessing functions to create training data from CATMA annotations.

.network_graphs

Network class which creates three types of drama networks:

• configuration networks based on the character's copresence
• communcation networks based on every character speech
• request networks based on the request annotations

Additionaly, every network type can be plotted as Plotly graph (see: network_ploty.ipynb) and standard network stats for each networks type are provided.

.request_clf

A Support Vector Machine classifier pipeline to detect dramatic speeches as requests.

.xml_parsing

Perprocessing for the TEI encoded plays from GerDraCor.

character_gender_gerdracor.json

The gender of all characters in the GerDraCor corpus.

gerdracor-metadata.csv

Corpus metadata.

manual_annotations.json

Manual request annotations of four German plays:

• kleist-der-zerbrochene-krug
• kleist-die-familie-schroffenstein
• schlegel-canut
• wedekind-fruehlings-erwachen

{
    play: "kleist-der-zerbrochene-krug",
    text: "Ei, was zum Henker, sagt, Gevatter Adam! Was ist mit Euch geschehn? Wie seht Ihr aus? ",
    start_point: 1147,
    end_point: 1242,
    speaker: "licht",
    request: "True",
    manual_annotations: [
        {
            annotation: "Ei, was zum Henker, sagt, Gevatter Adam!",
            tag: "request",
            start_point: 1154,
            end_point: 1194,
            speaker: "licht",
            addressee: "adam"
        },
        {
            annotation: "Was ist mit Euch geschehn?",
            tag: "question",
            start_point: 1195,
            end_point: 1221,
            speaker: "licht",
            addressee: "adam"
        },
        {
            annotation: "Wie seht Ihr aus?",
            tag: "question",
            start_point: 1222,
            end_point: 1239,
            speaker: "licht",
            addressee: "adam"
        }
    ]
}

network_plots.ipynb

Jupyte Notebook to illustrate the network types with network graph as interactive plotly plots.

Prerequisites

networkx==2.6.2
nltk==3.6.7
plotly==5.5.0
pandas==1.4.0
scikit-learn==1.0.2
scipy==1.7.3
spacy==3.2.1