Table 2 Comparative evaluation of Theme-generated terms with LDA using the UMass coherence metric on the SNP dataset
From: Discovering themes in biomedical literature using a projection-based algorithm
# Cl | Method | Topic terms | ||
|---|---|---|---|---|
Top 5 | Top 10 | Top 20 | ||
100 | LDA | -21.15 | -108.99 | -541.92 |
100 | Theme uni | -15.64 | -81.73 | -378.95 |
100 | Theme | -13.53 | -80.12 | -397.23 |
1000 | LDA | -33.66 | -181.19 | -942.08 |
1623 | Theme uni | -19.43 | -98.31 | -461.09 |
1066 | Theme | -17.25 | -94.82 | -462.35 |