Annotating events in Mythological Sequences

• Hylemes_extract.csv: data excerpt for the annotation study on event types in mythological sequences
• Statements are derived according to the hylistic theory (Zgoll, 2019)
• For access to the full data, contact: [email protected]
• Full data: 93 durative-initial, 1338 durative-constant, 438 durative-resultative, 4443 single-point (= punktuell)
• submitted as a paper@LAW-XVII 2023

Performance

Binary Classifer

	precision	recall	f1-score	support
durativ	0.83	0.75	0.79	430
punktuell	0.91	0.94	0.93	1148
accuracy			0.89	1578
macro avg	0.87	0.85	0.86	1578
weighted avg	0.89	0.89	0.89	1578

Durative Classifier

	precision	recall	f1-score	support
durativ-initial	0.62	0.37	0.47	27
durativ-konstant	0.80	0.87	0.84	326
durativ-resultativ	0.58	0.49	0.53	116
accuracy			0.75	469
macro avg	0.67	0.58	0.61	469
weighted avg	0.74	0.75	0.74	469

Fine-grained Classifer

	precision	recall	f1-score	support
durativ-initial	0.50	0.17	0.25	30
durativ-konstant	0.73	0.67	0.70	297
durativ-resultativ	0.55	0.46	0.50	103
punktuell	0.90	0.95	0.93	1148
accuracy			0.85	1578
macro avg	0.67	0.56	0.59	1578
weighted avg	0.84	0.85	0.84	1578

Running the classifiers

You can run the pre-trained classifiers like this:

from joblib import load

# replace with your own path
path = 'my_path' 
filenames = ['binary_classifier.joblib', 'classifier_all_classes.joblib', 'classifier_durative.joblib']

for filename in filenames:
    loaded_model = load(path+filename)
    pred = loaded_model.predict(['Die Götter wohnen auf dem Olymp.']) 
    print(pred)

Literature

Zgoll, Christian. Tractatus mythologicus: Theorie und Methodik zur Erforschung von Mythen als Grundlegung einer allgemeinen, transmedialen und komparatistischen Stoffwissenschaft, Berlin, Boston: De Gruyter, 2019. https://doi.org/10.1515/9783110541588