Skip to main content

Table 5 Comparison of ClusTop algorithm against various baselines, in terms of Topic Coherence (TC) and Pointwise Mutual Information (PMI) for the top 15 and 20 keywords

From: A clustering-based topic model using word networks and word embeddings

Algorithm

Top 15 Keywords/Unigrams

Top 20 Keywords/Unigrams

Dataset A

Dataset B

Dataset C

Average

Dataset A

Dataset B

Dataset C

Average

TC

PMI

TC

PMI

TC

PMI

Rank@15

TC

PMI

TC

PMI

TC

PMI

Rank@20

ClusTop-Word-NA

− 409.7 (21)

− 157.4 (17)

− 382.6 (19)

− 115.5 (16)

− 403.5 (18)

− 147.6 (21)

(18.7)

− 768.2 (21)

− 318.1 (19)

− 695.1 (19)

− 234.8 (17)

− 732.7 (18)

− 247.8 (21)

(19.2)

ClusTop-BiG-NA

− 357.2 (20)

− 108.3 (15)

− 360.3 (17)

− 77.2 (12)

− 447.9 (20)

− 116.5 (18)

(17.0)

− 645.5 (20)

− 212.3 (16)

− 633.7 (17)

− 147.8 (13)

− 794.7 (20)

− 165.4 (18)

(17.3)

ClusTop-TriG-NA

− 289.2 (16)

− 83.1 (13)

− 382.0 (18)

− 86.2 (15)

− 451.6 (21)

− 145.4 (20)

(17.2)

− 530.2 (16)

− 170.0 (14)

− 674.8 (18)

− 183.1 (14)

− 804.0 (21)

− 222.8 (19)

(17.0)

ClusTop-BiHa-NA

− 179.9 (13)

− 24.2 (10)

− 324.4 (15)

− 82.8 (14)

− 389.5 (17)

− 98.6 (17)

(14.3)

− 319.3 (13)

− 64.1 (10)

− 569.2 (15)

− 189.6 (15)

− 692.0 (17)

− 141.3 (17)

(14.5)

ClusTop-Hash-NA

− 38.4 (2)

− 4.0 (5)

− 116.4 (5)

− 11.3 (7)

− 98.6 (6)

13.1 (4)

(4.8)

− 63.7 (2)

− 11.0 (6)

− 195.2 (7)

− 9.7 (6)

− 148.8 (9)

28.2 (4)

(5.7)

ClusTop-Noun-NA

− 139.1 (8)

− 23.3 (9)

− 208.5 (11)

− 31.4 (9)

− 219.9 (15)

− 29.6 (11)

(10.5)

− 242.1 (8)

− 63.7 (9)

− 363.5 (11)

− 86.9 (9)

− 384.7 (15)

− 36.4 (10)

(10.3)

ClusTop-H2VG-NA

− 160.2 (10)

− 77.7 (12)

− 183.4 (9)

− 70.6 (11)

− 75.6 (3)

− 24.8 (10)

(9.2)

− 292.1 (11)

− 127.5 (11)

− 295.8 (10)

− 107.8 (11)

− 75.6 (3)

− 24.8 (8)

(9.0)

ClusTop-H2VW-NA

− 49.9 (3)

72.7 (2)

− 61.4 (2)

96.8 (1)

− 19.3 (1)

52.5 (2)

(1.8)

− 73.4 (3)

111.3 (2)

− 95.7 (2)

161.4 (1)

− 19.3 (1)

52.5 (3)

(2.0)

ClusTop-H2VF-NA

− 59.5 (4)

87.5 (1)

− 60.0 (1)

94.2 (2)

− 20.7 (2)

54.8 (1)

(1.8)

− 92.4 (4)

145.3 (1)

− 95.0 (1)

157.7 (2)

− 20.7 (2)

54.8 (2)

(2.0)

ClusTop-Word-AH

− 311.9 (18)

− 172.7 (19)

− 480.2 (21)

− 306.9 (21)

− 177.0 (14)

− 14.8 (8)

(16.8)

− 567.7 (17)

− 336.0 (21)

− 903.7 (21)

− 562.7 (21)

− 298.8 (14)

− 50.2 (13)

(17.8)

ClusTop-Hash-AH

− 32.3 (1)

− 4.2 (6)

− 99.8 (3)

− 16.3 (8)

− 90.7 (5)

8.7 (5)

(4.7)

− 54.8 (1)

− 9.5 (5)

− 169.0 (3)

− 22.2 (7)

− 143.6 (6)

21.9 (5)

(4.5)

ClusTop-Noun-AH

− 309.9 (17)

− 173.6 (20)

− 444.0 (20)

− 242.1 (19)

− 131.8 (9)

− 17.5 (9)

(15.7)

− 575.1 (18)

− 316.4 (18)

− 811.4 (20)

− 425.6 (19)

− 221.7 (12)

− 35.0 (9)

(16.0)

ClusTop-H2VG-AH

− 166.4 (11)

− 96.1 (14)

− 304.6 (14)

− 127.6 (17)

− 143.0 (12)

− 45.3 (13)

(13.5)

− 303.3 (12)

− 166.6 (13)

− 530.9 (14)

− 195.1 (16)

− 147.6 (8)

− 46.9 (12)

(12.5)

ClusTop-H2VW-AH

− 167.4 (12)

− 166.8 (18)

− 281.9 (13)

− 237.7 (18)

− 116.9 (8)

− 79.1 (15)

(14.0)

− 274.1 (10)

− 268.2 (17)

− 476.9 (12)

− 390.7 (18)

− 122.6 (5)

− 82.5 (15)

(12.8)

ClusTop-H2VF-AH

− 201.1 (14)

− 197.3 (21)

− 350.3 (16)

− 299.5 (20)

− 134.5 (10)

− 95.8 (16)

(16.2)

− 335.0 (14)

− 329.3 (20)

− 591.2 (16)

− 494.6 (20)

− 160.1 (10)

− 115.3 (16)

(16.0)

ClusTop-Word-AM

− 326.6 (19)

− 57.3 (11)

− 278.7 (12)

− 42.5 (10)

− 422.5 (19)

− 143.4 (19)

(15.0)

− 599.9 (19)

− 155.4 (12)

− 492.7 (13)

− 122.3 (12)

− 766.5 (19)

− 239.6 (20)

(15.8)

ClusTop-Hash-AM

− 152.7 (9)

− 17.5 (8)

− 102.1 (4)

− 1.8 (5)

− 350.0 (16)

− 50.1 (14)

(9.3)

− 262.2 (9)

− 56.7 (8)

− 169.8 (4)

− 26.5 (8)

− 615.2 (16)

− 67.0 (14)

(9.8)

ClusTop-Noun-AM

− 60.7 (5)

− 6.0 (7)

− 150.6 (8)

− 10.5 (6)

− 138.9 (11)

29.7 (3)

(6.7)

− 99.3 (5)

− 11.4 (7)

− 257.4 (8)

− 2.4 (5)

− 227.4 (13)

58.0 (1)

(6.5)

ClusTop-H2VG-AM

− 273.6 (15)

− 123.9 (16)

− 193.0 (10)

− 78.1 (13)

− 152.4 (13)

− 43.1 (12)

(13.2)

− 480.9 (15)

− 202.2 (15)

− 289.8 (9)

− 103.9 (10)

− 167.3 (11)

− 46.4 (11)

(11.8)

ClusTop-H2VW-AM

− 95.4 (6)

4.1 (4)

− 128.1 (7)

25.2 (3)

− 89.3 (4)

5.5 (6)

(5.0)

− 146.2 (6)

9.5 (4)

− 192.2 (6)

53.3 (3)

− 102.6 (4)

12.2 (7)

(5.0)

ClusTop-H2VF-AM

− 96.2 (7)

4.3 (3)

− 122.6 (6)

14.9 (4)

− 108.3 (7)

4.1 (7)

(5.7)

− 149.5 (7)

11.3 (3)

− 185.6 (5)

30.2 (4)

− 146.6 (7)

13.0 (6)

(5.3)

LDA-Orig

− 722.5 (24)

− 659.1 (24)

− 695.0 (24)

− 619.0 (24)

− 593.5 (24)

− 442.7 (24)

(24.0)

− 1279.6 (24)

− 1133.6 (24)

− 1262.1 (24)

− 1102.0 (24)

− 1083.2 (24)

− 790.0 (24)

(24.0)

LDA-Hash

− 607.9 (22)

− 435.4 (22)

− 611.4 (22)

− 449.2 (22)

− 500.7 (22)

− 271.5 (22)

(22.0)

− 1134.2 (22)

− 794.2 (22)

− 1120.9 (22)

− 805.6 (22)

− 930.7 (22)

− 463.1 (22)

(22.0)

LDA-Ment

− 611.6 (23)

− 465.7 (23)

− 613.9 (23)

− 477.7 (23)

− 540.9 (23)

− 316.9 (23)

(23.0)

− 1141.5 (23)

− 850.2 (23)

− 1130.6 (23)

− 855.9 (23)

− 981.9 (23)

− 566.6 (23)

(23.0)

  1. The rank of an algorithm’s performance for each metric are provided in brackets