Evaluation of cutoff policies for term extraction

Journal of the Brazilian Computer Society

Table 5 Precision (P), recall (R), F-measure (F), and list sizes using threshold-based cutoff points

Threshold	Bigrams					Trigrams
cutoff	P (%)	R (%)	F (%)	\(\mid \mathcal {EL} \mid \)	\(\mid \mathcal {EL} \cap \mathcal {RL} \mid \)	P (%)	R (%)	F (%)	\(\mid \mathcal {EL} \mid \)	\(\mid \mathcal {EL} \cap \mathcal {RL} \mid \)
points
0	10	100	18	15,487	1534	15	100	26	18,174	2660
1	11	98	20	13,589	1502	15	97	26	17,227	2577
2	39	96	56	3698	1470	77	93	84	3245	2483
3	72	92	81	1959	1416	87	46	60	1395	1213
4	83	69	75	1277	1061	90	28	42	820	737
5	88	52	65	909	798	91	18	31	536	490
6	91	42	57	704	639	92	13	22	365	335
7	94	33	49	541	506	94	10	17	273	256
8	94	27	42	439	414	94	7	14	209	196
9	95	23	37	364	347	94	6	11	159	150
10	96	20	33	313	301	93	5	9	131	122
11	97	17	28	263	254	93	4	8	113	105
12	97	14	24	220	213	93	3	6	86	80
13	97	12	21	188	183	95	3	5	73	69
14	98	11	19	167	163	95	2	4	62	59
15	98	9	17	143	142	96	2	4	52	50