Improving topic modeling through homophily for legal documents

Applied Network Science

Table 2 Coherence scores of topics from \({G}'_P\) and \({G}'_S\) against \(w_n\), \(w_j\), cutting and weighting models

	\({G}'_P, w_n\)		\({G}'_P, w_j\)		\({G}'_S, w_n\)		\({G}'_S, w_j\)
LDA		0.131		0.131		0.131		0.131
RTM	\(w_n \rightarrow \infty\)	0.159	\(w_j \rightarrow \infty\)	0.159	\(w_n \rightarrow \infty\)	0.159	\(w_j \rightarrow \infty\)	0.159
Cutting	\(w_n \, \ge\) 100	0.166	\(w_j \, \ge\) 0.75	0.160	\(w_n \, \ge\) 10	0.167	\(w_j \, \ge\) 0.75	0.164
	\(w_n \, \ge\) 50	0.165	\(w_j \, \ge\) 0.50	0.166	\(w_n \, \ge\) 5	0.159	\(w_j \, \ge\) 0.50	0.164
	\(w_n \, \ge\) 5	0.167	\(w_j \, \ge\) 0.25	0.161	\(w_n \, \ge\) 0	0.164	\(w_j \, \ge\) 0.25	0.165
	\(w_n \, \ge\) 0	0.166	\(w_j \, \ge\) 0	0.166			\(w_j \, \ge\) 0	0.164
Weighting			\(w_j^{P}\)	0.180			\(w_j^{S}\)	0.176
			\(w_s^{P}\)	0.170			\(w_s^{S}\)	0.171