khigashi1987 / labeled-lda Goto Github PK

Hi khigashi,
I am trying to use your code to create a topic model with multiply labeled text.
However I am not sure how the input format should look like.

What I am trying is:

from sklearn import cross_validation
from LLDA.llda import LLDAClassifier
import numpy

X_train = numpy.array([[1,2], [3,4]]) # 2 documents - with 4 unique words in total
y_train = numpy.array([[0,1], [1,0]]) # two classes
llda = LLDAClassifier(alpha = 0.5/y_train.shape[1], threshold=0.10)
llda.fit(X_train, y_train)

The console ouput says "Number of words 2" which I expected to be 4. (words 1,2,3,4)
In addition I looked at the phi-values because I wanted to know the topic-word distribution:

phi = numpy.loadtxt(os.path.join(llda.tmp, "fit.phi"))
phi

I exptected to see a 2x4 matrix (topicsxwords) but actually its 2x2...

Is there anything wrong with my input?

Thank you and best regards!