word space transformations

[wlf...@gmail.com]

Hi,

I want to built a word co-occurence space and use a pmi-weighting for the counts. I built a word co-occurence space with the help of GenericWordSpace, but I just can't figure out how to plug things together to be able transform my space with the help of the PointwiseMutualInformationTransform. What is the way to go?

Thanks and sorry if I'm missing something obvious,

Stefanie

[David Jurgens]

Hi Stefanie,

  Applying a transform to a generic word space is something we should already be supporting but it looks like there's no direct way to do it.  I think this functionality should be fairly quick to implement, so I'll see if I can get it done today.

  Thanks,

  David

--

---
You received this message because you are subscribed to the Google Groups "S-Space Package Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to s-space-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Stefanie Wolf]

Thanks, it would be very useful!

[Stefanie Wolf]

Hi David,

I just wanted to ask if you have made any updates for this issue.

Best,
Stefanie

[David Jurgens]

Hi Stefanie,

  Sorry for the delay!  The code for performing the transform is now tested and checked into the trunk.  Please let me know if you run into any issues with it and thanks for the suggested improvement!

  Best,

  David

[Stefanie Wolf]

Hi David,

thanks a lot. When I tested it on a small example I could see that all the values turned into zero. I solved the problem by using SparseHashdoubleVector in the processSpace-method.

vectorsInOrder.add(new SparseHashDoubleVector(Vectors.asDouble(
                Vectors.subview(e.getValue(), 0, getVectorLength()))));

Let me know if there is anything wrong with doing this.


Best,
Stefanie