Table 1 Comparison of knowledge-based topic models
From: Lifelong aspect extraction from big data: knowledge engineering
Features | DF-LDA | SAS | ME-SAS |
|---|---|---|---|
Model approach | Semi-superivsed | Semi-supervised | Hybrid |
Knowledge type | 1–1 mapping | Rule sets | Rule sets |
User support | Must-link/cannot-link | Seed aspects | Seed aspects |
Handling big data | No | No | No |
Learning criteria | Nil | Nil | Nil |
Losing wrong knowledge | Nil | Nil | Nil |
Transitivity issue addressed | No | No | No |
Aspect/sentiment separation | No | Yes | Yes |
Features | MC-LDA | AKL | LTM |
|---|---|---|---|
Model approach | Semi-supervised | Automatic | Automatic |
Knowledge type | Rule sets | Rule sets | Rule sets |
User support | Must-set/cannot-set | Must-set/cannot-set | 1–1 mapping |
Handling big data | No | Yes | Yes |
Learning criteria | Nil | Knowledge clusters | PMI |
Losing wrong knowledge | Nil | Yes | Yes |
Transitivity issue addressed | Yes | Yes | Nil |
Aspect/sentiment separation | Yes | Sentiments pre-processed | Sentiments pre-processed |