Benchmarks

Benchmark results on the SearchSnippets dataset. Sixteen models were trained with varying iterations (10-2000) using default parameters and 8 topics. Metrics calculated: semantic coherence (u_mass) and perplexity.