Question about LDA weights

EJang
EJang Altair Community Member
in Community Q&A

Hi experts,

I ran a LDA analysis and found the weights in the data table and wordcloud do not match.

Here is the data table. Specifically, the weight of the word "flavor" is 284.

Screenshot 2024-12-30 144623.png

However, when I visualized it using Wordcloud, the weight of the word flavor was 1,158. The weight in the table and wordcloud don't match. Anyone knows the reason?

Untitled.png

Answers

  • RolandJones
    RolandJones
    Altair Employee

    Hi @EJang,

    What you will find is that in the background, the wordcloud is creating the sum of the weights across all topicID's for each word before plotting. You can verify this using an Aggregate operator. In your case, it means that "flavor" appears in multiple topics and the weights of each occurrence add up to 1,158.

    Let me know if you require further assistance.

    Best,

    Roland

  • EJang
    EJang Altair Community Member

    Thank you so much for the comment!

    Then, is there any other way to create the wordcloud matching the weight (284) number?

  • RolandJones
    RolandJones
    Altair Employee

    In this case I would look to filter on the column topicID so that you only display results for topic 0. You could then add some automation with loops or similar if you needed to do this for multiple topics. Would this meet your requirements?

Categories