Skip to main content

Table 4 Comparison of ClusTop algorithm against various baselines, in terms of Topic Coherence (TC) and Pointwise Mutual Information (PMI) for the top 5 and 10 keywords

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

Algorithm

Top 5 Keywords/Unigrams

Top 10 Keywords/Unigrams

Dataset A

Dataset B

Dataset C

Average

Dataset A

Dataset B

Dataset C

Average

TC

PMI

TC

PMI

TC

PMI

Rank@5

TC

PMI

TC

PMI

TC

PMI

Rank@10

ClusTop-Word-NA

− 37.6 (21)

− 5.5 (15)

− 34.1 (15)

− 7.7 (14)

− 37.9 (18)

− 14.4 (17)

(16.7)

− 171.0 (21)

− 49.2 (16)

− 160.8 (17)

− 39.5 (15)

− 173.4 (18)

− 67.5 (19)

(17.7)

ClusTop-BiG-NA

− 36.6 (20)

7.3 (8)

− 35.9 (17)

1.2 (9)

− 42.5 (22)

− 16.4 (18)

(15.7)

− 153.4 (20)

− 29.6 (13)

− 158.2 (16)

− 25.8 (13)

− 194.8 (21)

− 63.4 (18)

(16.8)

ClusTop-TriG-NA

− 30.9 (17)

10.7 (5)

− 35.8 (16)

− 2.6 (13)

− 42.0 (21)

− 18.2 (19)

(15.2)

− 122.6 (16)

− 16.1 (12)

− 166.5 (19)

− 25.1 (12)

− 194.2 (20)

− 73.5 (21)

(16.7)

ClusTop-BiHa-NA

− 23.3 (12)

19.6 (1)

− 32.3 (14)

4.7 (7)

− 37.9 (18)

− 11.2 (14)

(11.0)

− 81.4 (12)

7.1 (4)

− 140.8 (15)

− 14.9 (11)

− 169.9 (17)

− 50.7 (15)

(12.3)

ClusTop-Hash-NA

− 7.1 (2)

5.8 (9)

− 14.8 (4)

0.3 (11)

− 14.1 (4)

2.6 (6)

(6.0)

− 19.4 (2)

2.3 (8)

− 54.9 (5)

− 6.9 (8)

− 47.8 (4)

4.4 (5)

(5.3)

ClusTop-Noun-NA

− 17.1 (6)

10.6 (6)

− 21.4 (8)

6.9 (5)

− 22.8 (10)

− 0.3 (10)

(7.5)

− 64.7 (8)

2.9 (6)

− 90.5 (10)

− 3.6 (7)

− 97.8 (14)

− 14.1 (10)

(9.2)

ClusTop-H2VG-NA

− 17.9 (8)

− 7.8 (16)

− 22.8 (10)

− 9.3 (15)

− 22.5 (8)

− 7.9 (12)

(11.5)

− 69.3 (9)

− 35.7 (14)

− 87.2 (9)

− 35.5 (14)

− 62.1 (7)

− 20.9 (11)

(10.7)

ClusTop-H2VW-NA

− 9.4 (3)

16.6 (3)

− 11.0 (2)

18.2 (1)

− 8.9 (2)

21.2 (1)

(2.0)

− 28.0 (3)

40.3 (2)

− 32.5 (2)

48.4 (1)

− 18.4 (1)

48.6 (2)

(1.8)

ClusTop-H2VF-NA

− 10.5 (4)

18.8 (2)

− 10.4 (1)

18.0 (2)

− 8.6 (1)

20.3 (2)

(2.0)

− 31.8 (5)

45.7 (1)

− 30.7 (1)

45.5 (2)

− 19.0 (2)

48.7 (1)

(2.0)

ClusTop-Word-AH

− 30.9 (17)

− 1.6 (14)

− 40.2 (19)

− 27.6 (19)

− 24.2 (11)

10.3 (3)

(13.8)

− 137.6 (18)

− 57.7 (17)

− 198.3 (21)

− 131.1 (20)

− 88.5 (12)

9.1 (4)

(15.3)

ClusTop-Hash-AH

− 6.3 (1)

5.4 (10)

− 12.8 (3)

0.9 (10)

− 13.1 (3)

2.6 (6)

(5.5)

− 16.2 (1)

1.5 (9)

− 47.3 (3)

− 7.3 (9)

− 43.7 (3)

2.2 (6)

(5.2)

ClusTop-Noun-AH

− 28.7 (14)

− 11.4 (18)

− 41.8 (20)

− 19.8 (18)

− 17.6 (5)

4.6 (4)

(13.2)

− 132.4 (17)

− 72.3 (19)

− 185.6 (20)

− 102.5 (18)

− 63.8 (8)

− 2.5 (9)

(15.2)

ClusTop-H2VG-AH

− 17.6 (7)

− 9.1 (17)

− 32.0 (13)

− 15.6 (17)

− 29.5 (14)

− 11.3 (15)

(13.8)

− 71.2 (10)

− 41.0 (15)

− 136.3 (14)

− 64.7 (17)

− 97.9 (15)

− 33.7 (14)

(14.2)

ClusTop-H2VW-AH

− 27.2 (13)

− 23.8 (20)

− 38.7 (18)

− 32.1 (20)

− 26.6 (13)

− 19.2 (21)

(17.5)

− 84.0 (13)

− 83.3 (20)

− 133.6 (13)

− 113.9 (19)

− 87.7 (11)

− 60.7 (16)

(15.3)

ClusTop-H2VF-AH

− 29.0 (15)

− 23.8 (21)

− 45.1 (21)

− 38.4 (21)

− 25.7 (12)

− 18.3 (20)

(18.3)

− 97.3 (14)

− 93.3 (21)

− 166.1 (18)

− 144.0 (21)

− 85.1 (10)

− 62.9 (17)

(16.8)

ClusTop-Word-AM

− 34.4 (19)

8.3 (7)

− 30.9 (12)

11.0 (4)

− 37.7 (17)

− 14.3 (16)

(12.5)

− 146.1 (19)

− 5.4 (11)

− 126.8 (12)

8.1 (4)

− 179.9 (19)

− 69.0 (20)

(14.2)

ClusTop-Hash-AM

− 19.7 (11)

11.4 (4)

− 18.5 (5)

16.3 (3)

− 33.7 (16)

− 7.3 (11)

(8.3)

− 73.9 (11)

8.5 (3)

− 52.9 (4)

14.3 (3)

− 153.6 (16)

− 30.4 (12)

(8.2)

ClusTop-Noun-AM

− 11.2 (5)

4.8 (12)

− 19.2 (6)

0.3 (11)

− 22.6 (9)

4.0 (5)

(8.0)

− 29.9 (4)

-0.3 (10)

− 70.1 (8)

− 9.0 (10)

− 70.3 (9)

11.9 (3)

(7.3)

ClusTop-H2VG-AM

− 30.9 (17)

− 13.9 (19)

− 26.7 (11)

− 12.0 (16)

− 29.5 (14)

− 10.9 (13)

(15.0)

− 119.1 (15)

− 59.3 (18)

− 99.4 (11)

− 43.6 (16)

− 95.7 (13)

− 31.9 (13)

(14.3)

ClusTop-H2VW-AM

− 18.7 (9)

3.7 (13)

− 21.6 (9)

3.1 (8)

− 19.9 (7)

0.7 (9)

(9.2)

− 53.3 (6)

2.7 (7)

− 69.6 (7)

6.9 (5)

− 58.8 (6)

0.3 (8)

(6.5)

ClusTop-H2VF-AM

− 19.6 (10)

5.2 (11)

− 20.2 (7)

5.0 (6)

− 19.6 (6)

1.9 (8)

(8.0)

− 54.0 (7)

3.4 (5)

− 65.2 (6)

5.4 (6)

− 58.2 (5)

0.4 (7)

(6.0)

LDA-Orig

− 74.7 (24)

− 74.4 (24)

− 66.9 (24)

− 62.2 (24)

− 54.2 (24)

− 43.1 (24)

(24.0)

− 323.5 (24)

− 307.5 (24)

− 297.1 (24)

− 269.2 (24)

− 251.3 (24)

− 191.9 (24)

(24.0)

LDA-Hash

− 51.2 (22)

− 43.8 (22)

− 55.1 (23)

− 42.9 (22)

− 41.5 (20)

− 23.4 (22)

(21.8)

− 247.1 (22)

− 185.4 (22)

− 256.8 (22)

− 199.2 (22)

− 206.6 (22)

− 112.5 (22)

(22.0)

LDA-Ment

− 52.8 (23)

− 45.9 (23)

− 54.1 (22)

− 45.3 (23)

− 47.3 (23)

− 27.7 (23)

(22.8)

− 250.2 (23)

− 198.8 (23)

− 258.6 (23)

− 206.5 (23)

− 225.5 (23)

− 136.4 (23)

(23.0)

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