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

Mu, Wenchuan; Lim, Kwan Hui; Liu, Junhua; Karunasekera, Shanika; Falzon, Lucia; Harwood, Aaron

doi:10.1186/s40537-022-00585-4

Journal of Big Data

Table 6 Comparison of ClusTop algorithm against various baselines, in terms of Precision (Pre), Recall (Rec) and F-score (FS) for the top 5 keywords/unigrams of each topic

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

Algorithm	Top 5 Keywords/Unigrams
	Dataset A			Dataset B			Dataset C			Average
	Precision	Recall	F-score	Precision	Recall	F-score	Precision	Recall	F-score	Rank@5
ClusTop-Word-NA	.754 ± .002 (22)	.031 ± .000 (6)	.059 ± .000 (6)	.866 ± .003 (21)	.043 ± .000 (5)	.082 ± .001 (5)	.840 ± .009 (18)	.027 ± .001 (7)	.052 ± .001 (7)	(10.8)
ClusTop-BiG-NA	.786 ± .002 (18)	.034 ± .000 (2)	.064 ± .000 (2)	.857 ± .003 (22)	.046 ± .000 (2)	.086 ± .001 (2)	.833 ± .010 (20)	.029 ± .001 (4)	.056 ± .001 (3)	(8.3)
ClusTop-TriG-NA	.791 ± .002 (17)	.034 ± .000 (2)	.064 ± .000 (2)	.871 ± .003 (20)	.046 ± .000 (2)	.087 ± .001 (1)	.822 ± .009 (21)	.031 ± .001 (2)	.058 ± .001 (2)	(7.7)
ClusTop-BiHa-NA	.784 ± .002 (19)	.032 ± .000 (4)	.060 ± .000 (4)	.886 ± .003 (18)	.045 ± .000 (4)	.084 ± .001 (4)	.820 ± .009 (22)	.029 ± .001 (4)	.055 ± .001 (5)	(9.3)
ClusTop-Hash-NA	.898 ± .004 (10)	.023 ± .000 (21)	.044 ± .000 (18)	.916 ± .007 (13)	.032 ± .001 (17)	.062 ± .001 (17)	.936 ± .011 (8)	.022 ± .000 (17)	.042 ± .001 (18)	(15.4)
ClusTop-Noun-NA	.761 ± .002 (21)	.028 ± .000 (8)	.054 ± .000 (8)	.836 ± .003 (24)	.043 ± .000 (5)	.081 ± .001 (6)	.888 ± .008 (14)	.032 ± .001 (1)	.062 ± .001 (1)	(9.8)
ClusTop-H2VG-NA	.924 ± .003 (3)	.023 ± .000 (21)	.045 ± .000 (17)	.978 ± .004 (2)	.032 ± .001 (17)	.062 ± .001 (17)	.976 ± .007 (4)	.023 ± .001 (14)	.044 ± .001 (14)	(12.1)
ClusTop-H2VW-NA	.906 ± .001 (9)	.023 ± .001 (21)	.023 ± .001 (22)	.963 ± .001 (6)	.028 ± .001 (20)	.028 ± .001 (20)	.986 ± .001 (1)	.019 ± .001 (23)	.019 ± .001 (23)	(16.1)
ClusTop-H2VF-NA	.910 ± .001 (7)	.025 ± .001 (15)	.025 ± .001 (19)	.960 ± .001 (7)	.029 ± .001 (19)	.029 ± .001 (19)	.955 ± .001 (6)	.019 ± .001 (23)	.019 ± .001 (23)	(15.3)
ClusTop-Word-AH	.741 ± .003 (24)	.025 ± .000 (15)	.049 ± .000 (13)	.845 ± .005 (23)	.040 ± .000 (9)	.075 ± .001 (9)	.844 ± .010 (17)	.027 ± .001 (7)	.052 ± .001 (7)	(13.8)
ClusTop-Hash-AH	.847 ± .002 (12)	.026 ± .000 (12)	.051 ± .000 (12)	.912 ± .004 (15)	.046 ± .001 (2)	.086 ± .001 (2)	.896 ± .008 (12)	.024 ± .000 (12)	.046 ± .001 (13)	(10.2)
ClusTop-Noun-AH	.802 ± .002 (16)	.024 ± .000 (18)	.046 ± .000 (16)	.873 ± .004 (19)	.037 ± .000 (13)	.071 ± .001 (13)	.872 ± .008 (15)	.029 ± .001 (4)	.056 ± .001 (3)	(13.0)
ClusTop-H2VG-AH	.919 ± .002 (5)	.025 ± .000 (15)	.049 ± .000 (13)	.902 ± .003 (16)	.042 ± .000 (7)	.080 ± .001 (7)	.891 ± .008 (13)	.025 ± .001 (10)	.048 ± .001 (11)	(10.8)
ClusTop-H2VW-AH	.927 ± .001 (1)	.023 ± .001 (21)	.023 ± .001 (22)	.972 ± .001 (4)	.028 ± .001 (20)	.028 ± .001 (20)	.979 ± .001 (3)	.020 ± .001 (21)	.020 ± .001 (21)	(14.8)
ClusTop-H2VF-AH	.918 ± .001 (6)	.025 ± .001 (15)	.025 ± .001 (19)	.965 ± .001 (5)	.027 ± .001 (23)	.027 ± .001 (23)	.948 ± .001 (7)	.021 ± .001 (19)	.021 ± .001 (19)	(15.1)
ClusTop-Word-AM	.748 ± .001 (23)	.034 ± .000 (2)	.065 ± .000 (1)	.929 ± .003 (11)	.036 ± .000 (15)	.069 ± .001 (15)	.758 ± .016 (24)	.022 ± .000 (17)	.043 ± .001 (16)	(13.8)
ClusTop-Hash-AM	.763 ± .002 (20)	.027 ± .000 (10)	.052 ± .000 (10)	.917 ± .003 (12)	.037 ± .000 (13)	.072 ± .001 (12)	.869 ± .011 (16)	.024 ± .000 (12)	.047 ± .001 (12)	(13.0)
ClusTop-Noun-AM	.842 ± .002 (13)	.025 ± .000 (15)	.048 ± .000 (15)	.950 ± .003 (9)	.039 ± .000 (10)	.074 ± .001 (10)	.923 ± .009 (9)	.022 ± .000 (17)	.043 ± .001 (16)	(12.7)
ClusTop-H2VG-AM	.864 ± .002 (11)	.028 ± .000 (8)	.054 ± .000 (8)	.930 ± .003 (10)	.041 ± .000 (8)	.078 ± .001 (8)	.900 ± .008 (10)	.025 ± .000 (10)	.049 ± .001 (9)	(9.1)
ClusTop-H2VW-AM	.924 ± .001 (3)	.023 ± .001 (21)	.023 ± .001 (22)	.976 ± .001 (3)	.027 ± .001 (23)	.027 ± .001 (23)	.981 ± .001 (2)	.020 ± .001 (21)	.020 ± .001 (21)	(15.4)
ClusTop-H2VF-AM	.910 ± .001 (7)	.022 ± .001 (24)	.022 ± .001 (24)	.985 ± .001 (1)	.027 ± .001 (23)	.027 ± .001 (23)	.971 ± .001 (5)	.020 ± .001 (21)	.020 ± .001 (21)	(16.6)
LDA-Orig	.925 ± .001 (2)	.027 ± .000 (10)	.052 ± .000 (10)	.956 ± .002 (8)	.037 ± .000 (13)	.070 ± .001 (14)	.898 ± .010 (11)	.025 ± .000 (10)	.049 ± .001 (9)	(9.7)
LDA-Hash	.821 ± .002 (15)	.031 ± .000 (6)	.059 ± .000 (6)	.916 ± .003 (13)	.036 ± .000 (15)	.069 ± .001 (15)	.837 ± .010 (19)	.028 ± .001 (6)	.054 ± .001 (6)	(11.2)
LDA-Ment	.830 ± .002 (14)	.031 ± .000 (6)	.060 ± .000 (4)	.900 ± .003 (17)	.039 ± .000 (10)	.074 ± .001 (10)	.814 ± .018 (23)	.023 ± .000 (14)	.044 ± .001 (14)	(12.4)

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

Back to article page