[Topic-models] Semantic sense of LDA

Dave Newman newman at uci.edu
Fri Jan 15 16:27:51 EST 2010


On Sat, Jan 16, 2010 at 7:17 AM, Aaron Zinman <aaron at azinman.com> wrote:
> Well I was originally thinking about an analogy of the correlated topic
> model, where by tokens have some explicit correlation to each other.

For this, see:
Incorporating Domain Knowedge into Topic Modeling via Dirichlet Forest
Priors (ICML 2009)
Andrzejewski, Zhu, and Craven
http://pages.cs.wisc.edu/~andrzeje/research/df_lda.html

Dave

> On Jan 15, 2010, at 2:42 PM, Anthony Di Franco wrote:
>
> What exactly do you mean?
>
> On Jan 15, 2010 11:36 AM, "Aaron Zinman" <aaron at azinman.com> wrote:
>
> Computational complexity aside, couldn't you build that into the model
> through statistical correlation against other tokens?
> Aaron
>
> On Jan 15, 2010, at 2:17 PM, Anthony Di Franco wrote: > When it happens for
> less obvious reasons ...
>
>
> _______________________________________________
> Topic-models mailing list
> Topic-models at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/topic-models
>
>



-- 
David Newman
Research Scientist
NICTA Australia, and
Dept. Computer Science, UC Irvine
www.ics.uci.edu/~newman


More information about the Topic-models mailing list