search

vi3k6i5 / GuidedLDA

semi supervised guided topic model with custom guidedLDA
Mozilla Public License 2.0
497 stars 108 forks source link

which metric I should use? #40

Open vvssttkk opened 4 years ago

vvssttkk commented 4 years ago

hi

in my pool data, i want clusterization for this approach, i will train many models with different num topics after i should choose реу best model

which metrics at your libs i should use for this? the best metrics its coherence but i can't find her in your libs

thakur-nandan commented 4 years ago

@vtrokhymenko Unfortunately there is no metric as such to check the best model, you could try to choose the best model by checking the topic coverage of all the individual topic ids and checking the probabilities of words across topics, best words of the topic and manually evaluating the model against your test set.