Explainable Detection of Online Sexism (EDOS)

The EDOS task presented at SemEval 2023 aims at the detection of sexism.

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='en')
>>> ins = comp2023.stack_2_bow_keywords(D)

Performance in Cross-validation (A)

Configuration	Performance	p-value
Comp2023.stack_2_bow_keywords	0.7622	1.0000
Comp2023.stack_3_bows_tailored_keywords	0.7580	0.2220
Comp2023.stack_2_bow_tailored_keywords	0.7567	0.0960
Comp2023.stack_2_bow_tailored_all_keywords	0.7532	0.1100
Comp2023.stack_3_bows	0.7517	0.0720
Comp2023.stack_2_bow_all_keywords	0.7503	0.0600
Comp2023.stack_bow_keywords_emojis	0.7502	0.0280
Comp2023.stack_3_bow_tailored_all_keywords	0.7487	0.0300
Comp2023.stack_bows	0.7486	0.0540
Comp2023.stack_bow_keywords_emojis_voc_selection	0.7478	0.0100
Comp2023.bow	0.7398	0.0060
Comp2023.bow_training_set	0.7354	0.0020
Comp2023.bow_voc_selection	0.7350	0.0000

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='en')
>>> ins = comp2023.stack_bow_keywords_emojis(D)

Performance in Cross-validation (B)

Configuration	Performance	p-value
Comp2023.stack_bow_keywords_emojis	0.5247	1.0000
Comp2023.stack_2_bow_keywords	0.5123	0.1580
Comp2023.stack_bow_keywords_emojis_voc_selection	0.5088	0.1540
Comp2023.stack_2_bow_tailored_keywords	0.5064	0.1040
Comp2023.stack_2_bow_all_keywords	0.5002	0.1440
Comp2023.stack_2_bow_tailored_all_keywords	0.4969	0.1000
Comp2023.stack_3_bow_tailored_all_keywords	0.4950	0.0960
Comp2023.stack_3_bows	0.4929	0.0760
Comp2023.stack_3_bows_tailored_keywords	0.4924	0.0080
Comp2023.stack_bows	0.4909	0.1000
Comp2023.bow	0.4597	0.0340
Comp2023.bow_training_set	0.4450	0.0140
Comp2023.bow_voc_selection	0.4427	0.0140

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='en')
>>> ins = comp2023.stack_2_bow_all_keywords(D)

Performance in Cross-validation (C)

Configuration	Performance	p-value
Comp2023.stack_2_bow_all_keywords	0.3236	1.0000
Comp2023.stack_2_bow_tailored_all_keywords	0.3145	0.0980
Comp2023.stack_bow_keywords_emojis	0.3123	0.2760
Comp2023.stack_2_bow_tailored_keywords	0.3069	0.1460
Comp2023.stack_3_bow_tailored_all_keywords	0.3035	0.0020
Comp2023.stack_bow_keywords_emojis_voc_selection	0.2943	0.0580
Comp2023.stack_3_bows_tailored_keywords	0.2924	0.0240
Comp2023.stack_2_bow_keywords	0.2870	0.0120
Comp2023.bow_voc_selection	0.2700	0.0140
Comp2023.bow	0.2685	0.0140
Comp2023.stack_3_bows	0.2556	0.0000
Comp2023.bow_training_set	0.2530	0.0080
Comp2023.stack_bows	0.2486	0.0000