Misha24-10/MultiCoNER-2-recognition-model

An example of using an ensemble of models is shown in the main.py file

In low lavel classification on MultiCoNER II in test set:

Класс	Precision	Recall	F1
Facility	0,7464	0,7321	0,7392
OtherLOC	0,7932	0,7068	0,7475
HumanSettlement	0,899	0,8948	0,8969
Station	0,8318	0,8125	0,8221
VisualWork	0,8528	0,8319	0,8422
MusicalWork	0,8025	0,7813	0,7917
WrittenWork	0,7766	0,728	0,7515
ArtWork	0,6374	0,5528	0,5921
Software	0,8476	0,8201	0,8336
MusicalGRP	0,8185	0,8207	0,8196
PublicCorp	0,7853	0,7572	0,771
PrivateCorp	0,7362	0,6896	0,7121
AerospaceManufacturer	0,6774	0,7541	0,7137
SportsGRP	0,8715	0,8938	0,8825
CarManufacturer	0,7617	0,7902	0,7757
ORG	0,7617	0,7371	0,7492
Scientist	0,5338	0,4886	0,5102
Artist	0,7971	0,8369	0,8165
Athlete	0,8094	0,802	0,8057
Politician	0,7115	0,6194	0,6622
Cleric	0,7349	0,6239	0,6748
SportsManager	0,678	0,6097	0,6421
OtherPER	0,5354	0,5915	0,562
Clothing	0,6326	0,6876	0,659
Vehicle	0,6699	0,6608	0,6653
Food	0,6814	0,6634	0,6723
Drink	0,6859	0,7203	0,7027
OtherPROD	0,7033	0,6638	0,683
Medication/Vaccine	0,7943	0,816	0,805
MedicalProcedure	0,7481	0,7375	0,7428
AnatomicalStructure	0,7765	0,7567	0,7664
Symptom	0,6086	0,7178	0,6587
Disease	0,7977	0,7719	0,7846
Macro Average Performance	0,7423	0,7294	0,7349

In high lavel classification on MultiCoNER II in test set:

Класс	Precision	Recall	F1
LOC	0,8866	0,8732	0,8798
Medicine	0,794	0,7927	0,7934
GRP	0,8489	0,8419	0,8454
PROD	0,7449	0,7247	0,7347
PER	0,9346	0,939	0,9368
CW	0,8507	0,8162	0,8331
Macro Average Performance	0,8433	0,8313	0,8372

MultiCoNER II features complex NER in these languages:

classification entities in low level between languages overall Macro F1-score: