[Topic-models] Training Classifier with multi-labeled data

David Blei david.blei at gmail.com
Wed Mar 3 16:55:54 EST 2010


hi bin

another option is to use supervised LDA in the multinomial regression
setting.  chong wang, fei--fei li, and i wrote about this in the
context of image classification,

   http://www.cs.princeton.edu/~blei/papers/WangBleiFeiFei2009.pdf

chong has placed code on-line,

  http://www.cs.princeton.edu/~chongw/slda/index.html

for a more general perspective on supervised LDA, jon mcauliffe and i
recently submitted this paper to statistical science:

  http://arxiv.org/abs/1003.0783

(the link may not be active yet, but it will be soon.)

best
dave

On Wed, Mar 3, 2010 at 12:46 PM, Daniel Ramage <dramage at cs.stanford.edu> wrote:
> Hi Bin,
>
> One option is to use Labeled LDA,
> http://www.aclweb.org/anthology/D/D09/D09-1026.pdf which constrains each
> document's topic distribution to align with the document's label space.
>  Because the per-document topics in this model are actually observed, it's
> less of a latent and more of a blatant dirichlet allocation.  It's
> competitive with an SVM baseline in our experiments, but state of the art
> discriminative models still beat it.
>
> dan
>
> Liu Bin wrote:
>>
>> Dear Friends,
>>
>> Currently, my job is designing a classifier to classify some short
>> questions.
>> The classifier can be trained by a collection of avialable questions with
>> labels.
>> And a question can be labeled to several related categories.
>> Thus, the topic models might be useful.
>>
>> However, as I have known, the LDA model is an unsupervised model.
>> Is there any alternative way to use topic models to do multi-labeled
>> classification?
>>
>>
>> Thanks!
>>
>> Yours,
>> Bin
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Topic-models mailing list
>> Topic-models at lists.cs.princeton.edu
>> https://lists.cs.princeton.edu/mailman/listinfo/topic-models
>
> _______________________________________________
> Topic-models mailing list
> Topic-models at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/topic-models
>


More information about the Topic-models mailing list