0

I am using gensim's LDA and trying to see the perplexity for a certain number of topics.

Perplexity for 1 : -7.903370624873305

Coherence Score for 1 : 0.8044880331838007

Perplexity for 2 : -8.269065851934347

Coherence Score for 2 : 0.7999767615350039

Perplexity for 3 : -8.55527218635052

Coherence Score for 3 : 0.7957008853871509

Perplexity for 4 : -8.769445203605587

Coherence Score for 4 : 0.7950156446303915

Perplexity for 5 : -8.953285444837803

Coherence Score for 5 : 0.7895563460199423

Perplexity for 6 : -9.118557691084591

Coherence Score for 6 : 0.7886679109232323

Perplexity for 7 : -9.347579422466403

Coherence Score for 7 : 0.7862291131217295

Perplexity for 8 : -10.515445720592052

Coherence Score for 8 : 0.7870887700340385

Perplexity for 9 : -15.37602605337135

Coherence Score for 9 : 0.793312181374788

How do I judge here? If it is supposed to be closer to 0, then it would mean that with 1 topic its the best but that doesn't really make sense since I had over 100000 words for training the LDA. Should I choose somewhere in between?

Stayne
  • 21
  • 5

0 Answers0