Experimenting with LDA modeling

The BoW data and TF-IDF data are two variants of data formation. The model built on BoW data will result in a different outcome from the model built on TF-IDF. We will build models on both variants.

A model built on BoW data

The basic syntax of LDA is easy. The required input parameter is corpus. We assign the BoW data to build the first model, as illustrated in the following code snippet:

from gensim.models import LdaModellda_bow = LdaModel(bow_corpus,
num_topics=10,
id2word = gensim_dictionary)

I’d like to review several important model inputs, as follows:

• num_topics: This is the number of topics. In this experiment, we will just assign 10. We will learn how to determine the optimal number of topics in the Determining the optimal number of topics section.
• random_state=None: This is helpful for reproducibility.
• id2word: We assign our gensim_dictionary dictionary from our corpus. If we do not assign a dictionary, Gensim...