LDAvis
LDAvis copied to clipboard
cpsievert
Intertopic Distance Map Label
Is it possible to label the circles with the top term rather than a number? Thanks
This would be easy if we restricted the label to be the most probable term for each topic. Then it could be done on the R side in the createJSON() function.
It would be more complicated if we wanted to label each topic circle with the most relevant term, since that depends on the value of lambda (in the slider), and we'd have to modify the javascript to do this.
Do you have a preference? Would labeling using the most probable term be useful enough?
k
Thank you for your prompt response. Based on a re-read of your paper http://nlp.stanford.edu/events/illvi2014/papers/sievert-illvi2014.pdf http://nlp.stanford.edu/events/illvi2014/papers/sievert-illvi2014.pdf it think the second option " to label each topic circle with the most relevant term” would enable wider utilization of your excellent visaulation/analysis tool. I would anticipate that the ability to vary the lambda values and thus circle labels would facilitate peer and group panel discussions on defining appropriate topics.
Sincerely David Casciotti
On Jul 21, 2016, at 11:13 AM, Kenny Shirley [email protected] wrote:
This would be easy if we restricted the label to be the most probable term for each topic. Then it could be done on the R side in the createJSON() function.
It would be more complicated if we wanted to label each topic circle with the most relevant term, since that depends on the value of lambda (in the slider), and we'd have to modify the javascript to do this.
Do you have a preference? Would labeling using the most probable term be useful enough?
k
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/cpsievert/LDAvis/issues/59#issuecomment-234285360, or mute the thread https://github.com/notifications/unsubscribe-auth/AAUgXkPULPQSVknuN95UxSWkNdB_vWpFks5qX4yngaJpZM4JQuHp.
Thought you might be interested in this: http://webmappingevents.com/data-visulation/sociometric-analysis/ http://webmappingevents.com/data-visulation/sociometric-analysis/
On Jul 22, 2016, at 11:07 AM, DaveCasciotti [email protected] wrote:
Thank you for your prompt response. Based on a re-read of your paper http://nlp.stanford.edu/events/illvi2014/papers/sievert-illvi2014.pdf http://nlp.stanford.edu/events/illvi2014/papers/sievert-illvi2014.pdf it think the second option " to label each topic circle with the most relevant term” would enable wider utilization of your excellent visaulation/analysis tool. I would anticipate that the ability to vary the lambda values and thus circle labels would facilitate peer and group panel discussions on defining appropriate topics.
Sincerely David Casciotti
On Jul 21, 2016, at 11:13 AM, Kenny Shirley <[email protected] mailto:[email protected]> wrote:
This would be easy if we restricted the label to be the most probable term for each topic. Then it could be done on the R side in the createJSON() function.
It would be more complicated if we wanted to label each topic circle with the most relevant term, since that depends on the value of lambda (in the slider), and we'd have to modify the javascript to do this.
Do you have a preference? Would labeling using the most probable term be useful enough?
k
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/cpsievert/LDAvis/issues/59#issuecomment-234285360, or mute the thread https://github.com/notifications/unsubscribe-auth/AAUgXkPULPQSVknuN95UxSWkNdB_vWpFks5qX4yngaJpZM4JQuHp.
Hi! I would like to follow up on this question, as I don't see it resolved and I have a quite similar question. But my question is: is possible to label the circles (topics) with a code I created (3 to 4 digits)?
