baseline

Baseline results

#	Language	precision	recall	f-measure	lev. distance
1	English	20.99	28.79	24.28	2.69
2	Hungarian	20.88	27.81	23.85	3.54
3	Czech	22.10	19.72	20.84	2.94
4	Spanish	15.76	17.91	16.76	5.20
5	Russian	13.23	14.13	13.67	7.62
6	French	11.08	14.00	12.37	4.32
7	Italian	8.12	10.54	9.18	5.35
8	Latin	6.76	13.17	8.94	4.14

Russian - Multilingual BERT Tokenizer (cased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	6.77	16.10	9.53	1.38
001	no	no	yes	12.00	23.08	15.79	2.92
010	no	yes	no	14.39	21.71	17.31	4.62
011	no	yes	yes	14.74	18.25	16.31	7.12
100	yes	no	no	6.85	10.19	8.19	5.85
101	yes	no	yes	10.68	15.99	12.80	6.06
110	yes	yes	no	15.89	14.55	15.19	8.94
111	yes	yes	yes	7.70	7.63	7.66	9.61
all	-	-	-	13.23	14.13	13.67	7.62

Italian - Multilingual BERT Tokenizer (cased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	2.70	6.97	3.89	1.59
001	no	no	yes	14.67	23.40	18.03	3.11
010	no	yes	no	14.11	19.16	16.25	3.77
011	no	yes	yes	10.53	13.95	12.00	5.79
100	yes	no	no	3.32	5.49	4.14	4.87
101	yes	no	yes	11.76	13.19	12.44	4.47
110	yes	yes	no	11.37	11.97	11.66	6.56
111	yes	yes	yes	11.11	10.64	10.87	7.09
all	-	-	-	8.12	10.54	9.18	5.35

Hungarian - Multilingual BERT Tokenizer (cased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	2.61	7.36	3.85	1.83
001	no	no	yes	9.89	20.37	13.31	2.78
010	no	yes	no	22.96	31.70	26.63	2.84
011	no	yes	yes	19.01	25.56	21.80	3.60
100	yes	no	no	16.75	27.17	20.72	2.97
101	yes	no	yes	17.76	27.27	21.51	3.92
110	yes	yes	no	25.53	28.36	26.87	3.88
111	yes	yes	yes	23.93	27.89	25.76	4.66
all	-	-	-	20.88	27.81	23.85	3.54

French - Multilingual BERT Tokenizer (cased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	6.55	16.23	9.33	1.48
001	no	no	yes	30.63	57.44	39.96	2.85
010	no	yes	no	12.60	18.44	14.97	3.81
011	no	yes	yes	16.31	22.22	18.81	5.45
100	yes	no	no	6.08	9.33	7.36	3.85
101	yes	no	yes	15.06	16.45	15.72	5.06
110	yes	yes	no	14.15	14.32	14.23	5.30
111	yes	yes	yes	6.67	5.68	6.13	8.37
all	-	-	-	11.08	14.00	12.37	4.32

Spanish - Multilingual BERT Tokenizer (cased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	5.19	12.82	7.39	1.47
001	no	no	yes	9.89	15.52	12.08	2.76
010	no	yes	no	16.02	21.29	18.28	3.37
011	no	yes	yes	5.26	5.26	5.26	4.83
100	yes	no	no	11.52	15.79	13.32	4.62
101	yes	no	yes	10.57	11.82	11.16	5.59
110	yes	yes	no	20.21	19.29	19.74	6.11
111	yes	yes	yes	28.92	28.57	28.74	7
all	-	-	-	15.76	17.91	16.76	5.2

English - BERT Tokenizer (uncased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	2.02	5.55	2.96	2.11
001	no	no	yes	46.60	64.11	53.97	1.42
010	no	yes	no	25.29	33.90	28.97	2.75
011	no	yes	yes	36.85	38.93	37.86	2.96
100	yes	no	no	12.16	19.14	14.87	2.73
101	yes	no	yes	51.57	50.12	50.83	1.51
110	yes	yes	no	24.92	26.17	25.53	3.31
111	yes	yes	yes	34.81	28.30	31.22	3.22
all	-	-	-	20.99	28.79	24.28	2.69

Mongolian - BERT Tokenizer (uncased)

category	inflection	derivation	compound	precision	recall	f_measure	distance
000	no	no	no	1.26	3.75	1.89	2.14
001	no	no	yes	-	-	-	-
010	no	yes	no	4.86	9.62	6.46	4.45
011	no	yes	yes	-	-	-	-
100	yes	no	no	6.22	12.52	8.31	3.50
101	yes	no	yes	0.00	0.00	0.00	5.33
110	yes	yes	no	6.45	10.18	7.90	5.94
111	yes	yes	yes	-	-	-	-
all	-	-	-	5.89	10.59	7.57	4.51

Language	Precison	Recall	F-measure	Lev. distance
Czech	36.76	30.35	33.25	21.01
English	63.68	65.77	64.71	5.50
Mongolian	20.00	29.95	23.99	28.86