I created a topic extraction with Elbow Model and ran the analysis
successfully. However, when I viewed the topics I created, there are some
multiple topics (e.g. I have two topic terms created under the name “help”,
one of them is created as topic 1 and the other is created as topic 2). How
can I solve this problem since the ideal approach is having one topic for
each topic term.
Can you post an example workflow and data that recreates the problem? That would definitely help in this case.
Thank you very much,
Here is the workflow I have.
topic extraction beg 1.knwf (327.1 KB)
The topics aren’t generated with any description other than topic_0, topic_1, etc… which is working as intended. It’s entirely possible to have the same term appear in multiple topics (although they will likely have different weights). Maybe that’s where your confusion is arising from? It’s unlikely that you’re going to have 1-to-1 topic to term correspondence.
If you want to dig into this further, could you upload a sample of your data as well? I noticed you are adapting the Topic Extraction workflow from the Hub (https://kni.me/w/H8EUf75lnsyAv6-U) to use Twitter data - but I can’t see what’s going on with your topics without your data.
Thanks for your response! So, then what I need to consider is if the weight is larger in topic_0 in comparison to topic_1 for the same topic, then it is more likely to include that term under topic_ 0, right?
You can’t strictly assume that because a term in topic_0 has a higher weight than in topic_1, that you should then associate the term only with the topic where it has a higher weight (although sometimes you might be able to get away with it). The weights are more relevant intra-cluster.
I understand. Thank you very much for your help, highly appreciated!