LLM post-correction raises LDA topic coherence by 5.86%, but LLM-guided initialization does not improve convergence and yields the worst final topics.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Semantic-Augmented Latent Topic Modeling with LLM-in-the-Loop
LLM post-correction raises LDA topic coherence by 5.86%, but LLM-guided initialization does not improve convergence and yields the worst final topics.