Topic Coherence

Calculate topic coherence for topic models.

model_coherence(models, ...)

# S3 method for gensim.models.basemodel.BaseTopicModel
model_coherence(models,
  ...)

# S3 method for list
model_coherence(models, ...)

Arguments

models	A model, i.e.: LDA or LSI, or a list of the latter.
...	Any other options, from the official documentation.

Details

A greater coherence is preferred: a higher value on the get_coherence method, see example.

Examples

# preprocess the corpus
texts <- prepare_documents(corpus)
#> → Preprocessing 9 documents
#> ← 9 documents after perprocessing
dictionary <- corpora_dictionary(texts)
corpus <- doc2bow(dictionary, texts)

# create 2 models to compare
good_lda_model <- model_lda(
  corpus = corpus,
  id2word = dictionary,
  iterations = 50L,
  num_topics = 2L
)
bad_lda_model <- model_lda(
  corpus = corpus,
  id2word = dictionary,
  iterations = 1L,
  num_topics = 5L
)

# create coherence models
good_cm <- model_coherence(
  model = good_lda_model,
  corpus = corpus,
  dictionary = dictionary,
  coherence = 'u_mass'
)
bad_cm <- model_coherence(
  model = bad_lda_model,
  corpus = corpus,
  dictionary = dictionary,
  coherence = 'u_mass'
)

# compare coherence
good_cm$get_coherence()
#> -14.643125063496548
bad_cm$get_coherence()
#> -14.701487465421206