07-10-2015 10:19 AM
I've been carrying out text mining in E Miner and have added text topics to my process flow. This works fine for the most part, but suddenly after a number of iterations, some of the important topics that had been generated in previous iterations no longer appear, even though they are some of the most frequently occurring topics, and as far as I can see I did not do anything that should exclude them. Has anyone else had this issue or know anything about how to solve it or what causes it?
I have been thinking that maybe once the term reaches a threshold number of mentions it is no longer pulled out as a topic but I'm not sure this makes sense as a consistent reason for this issue.
08-26-2015 04:11 PM
I guess my question back would be whether you have changed any of them to "user topics". If you make any change to a topic, such as its name, any of the topic weights, the term or document cutoff, then the topic is changed into a user topic. Then when you generate new multi-term topics, it intentionally eliminates the one that is closest to the topic you have modified: that way you don't have to look at essentially the same topic twice. Does this correspond to what you are seeing?