dou-topic-modeling

Analyzing the topic structure of DOU comments using Latent Dirichlet Allocation (LDA).

Trained model

Instead of training your own LDA, you can use a saved version from the trained folder. You can find an 18-topic snapshot on the Releases page.

Run the following line in the IPython Notebook:

ldamodel = models.ldamodel.LdaModel.load(os.path.join("trained", "checkpoint-18topics.lda"))

Example output

data/clean/clean-comments-10405.txt

                     work   51%
        living-conditions   17%
            relationships   16%
                 cashflow   6%

data/clean/clean-comments-10445.txt

                     work   44%
                 market-2   10%
        living-conditions   9%
                  tractor   8%
             moral-rights   7%
                 cashflow   6%