Automatic Detection and Characterization of Propaganda Techniques from Diplomats (DIPROMATS)

The DIPROMATS task (webpage)presented at IberLEF 2023 focused on indentifying tweets that have a propaganda technique.

The following code can generate an instance of the system used in the competition.

>>> from EvoMSA.competitions import Comp2023
>>> D = # Training set
>>> comp2023 = Comp2023(lang='es')
>>> ins = comp2023.stack_3_bows(D)

Performance in Cross-validation (Spanish)

Configuration	Performance	p-value
Comp2023.stack_3_bows	0.6551	1.0000
Comp2023.stack_3_bow_tailored_all_keywords	0.6544	0.4180
Comp2023.stack_3_bows_tailored_keywords	0.6515	0.2200
Comp2023.stack_2_bow_tailored_all_keywords	0.6514	0.2500
Comp2023.stack_bows	0.6488	0.1120
Comp2023.stack_2_bow_keywords	0.6486	0.1360
Comp2023.stack_2_bow_tailored_keywords	0.6486	0.1360
Comp2023.stack_bow_keywords_emojis_voc_selection	0.6486	0.1520
Comp2023.stack_bow_keywords_emojis	0.6485	0.1480
Comp2023.stack_2_bow_all_keywords	0.6484	0.1180
Comp2023.bow_training_set	0.6290	0.0080
Comp2023.bow	0.6136	0.0000
Comp2023.bow_voc_selection	0.6123	0.0000

The following code can generate an instance of the system used in the competition.

>>> from EvoMSA.competitions import Comp2023
>>> D = # Training set
>>> tailored = 'IberLEF2023_DIPROMATS_task1'
>>> comp2023 = Comp2023(lang='en', tailored=tailored)
>>> ins = comp2023.stack_3_bow_tailored_all_keywords(D)

Performance in Cross-validation (English)

Configuration	Performance	p-value
Comp2023.stack_3_bow_tailored_all_keywords	0.6498	1.0000
Comp2023.stack_3_bows_tailored_keywords	0.6489	0.2260
Comp2023.stack_2_bow_tailored_keywords	0.6471	0.1280
Comp2023.stack_2_bow_all_keywords	0.6448	0.0440
Comp2023.stack_2_bow_tailored_all_keywords	0.6446	0.0140
Comp2023.stack_2_bow_keywords	0.6443	0.0240
Comp2023.stack_3_bows	0.6386	0.0080
Comp2023.stack_bow_keywords_emojis_voc_selection	0.6381	0.0000
Comp2023.stack_bow_keywords_emojis	0.6377	0.0040
Comp2023.stack_bows	0.6327	0.0000
Comp2023.bow_training_set	0.6043	0.0000
Comp2023.bow	0.5961	0.0000
Comp2023.bow_voc_selection	0.5922	0.0000