Re: [Senseclusters-users] Comparing classification of sense clusters vs. stock naive bayes

[Ted P. <dul...@gm...>]

Hi Pete,

I think a direct comparison with Naive Bayes will be difficult, since Naive
Bayes is most often used as a supervised model, whereas SenseClusters is
unsupervised. SenseClusters doesn't allow for training of the same sort as
done by Naive Bayes, so again a direct comparison would be difficult.
SenseClusters is really intended for use cases where there is not
pre-existing gold standard available, and so it is perhaps more of a sense
discrimination or sense induction tool, whereas Naive Bayes is more a word
sense disambiguation method.

I hope this makes some sense. I think there are still interesting
comparisons that could be done, although I am not sure if they are in line
with what you hope to do. I am happy to discuss further though, if that
seems like it would be helpful.

Cordially,
Ted

On Mon, Mar 16, 2015 at 4:30 PM, Rail, Pete <pet...@hp...> wrote:

>  Hello All,
>
>
>
> I have installed the SenseClusters family of tools on CentOS and reviewed
> the materials  (including the long video seminar in Italy lead by the
> author).
>
>
>
> I am trying to compare the performance of SenseClusters to the simple Perl
> Algorithm::NaiveBayes here:
> http://search.cpan.org/~kwilliams/Algorithm-NaiveBayes-0.04/lib/Algorithm/NaiveBayes.pm
>
>
>
> Although I’ve read through the SenseClusters material and looked at a few
> samples, I fear I’m out of my depth.
>
>
>
> What I would like to do seems simple, but I don’t see it laid out in the
> instructions (which are actually quite thorough..thanks).
>
>
>
> I would like to learn how I can train SenseClusters to a recognize a given
> list of subjects from the same training set I’m using with my Naïve Bayes
> implementation. That way I can compare the performance against my test set
> to determine how much better SenseClusters performs compared to simple
> methods.
>
>
>
> If there are instructions you can point me to, or advice you can provide,
> that would be much appreciated.
>
>
>
> Thanks.
>
>
>
> *Pete Rail*, pet...@hp...
>
>
>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming The Go Parallel Website,
> sponsored
> by Intel and developed in partnership with Slashdot Media, is your hub for
> all
> things parallel software development, from weekly thought leadership blogs
> to
> news, videos, case studies, tutorials and more. Take a look and join the
> conversation now. http://goparallel.sourceforge.net/
> _______________________________________________
> senseclusters-users mailing list
> sen...@li...
> https://lists.sourceforge.net/lists/listinfo/senseclusters-users
>
>