axis mismatch bug fix

jointtsmodel/JST.py

@@ -210,9 +210,11 @@ def conditionalDistribution(self, d, v):
 
              secondFactor[:,s] = ((self.n_dst[d, s, :] + self.gammaVec) / \
             (self.n_ds[d, s] + np.sum(self.gammaVec)))
+
         thirdFactor = (self.n_vts[v,:, :] + self.beta) / \
             (self.n_ts + self.n_vts.shape[0] * self.beta)
-        probabilities_ts *= firstFactor[:, np.newaxis]
+
+        probabilities_ts *= firstFactor[np.newaxis,:]
         probabilities_ts *= secondFactor * thirdFactor
         probabilities_ts /= np.sum(probabilities_ts)
 

jointtsmodel/TSWE.py

@@ -254,9 +254,9 @@ def conditionalDistribution(self, d, v):
         #    forthFactor[k,:] = np.exp(np.dot(self.topic_embeddings[k,:],self.word_embeddings[v,:]))/np.sum(np.exp(np.dot(self.topic_embeddings[k,:],self.word_embeddings.T)))
 
         forthFactor = np.exp(np.dot(self.topic_embeddings,self.word_embeddings[v,:]))/np.sum(np.exp(np.dot(self.topic_embeddings,self.word_embeddings.T)),-1)
-        probabilities_ts *= firstFactor[:, np.newaxis]
+        probabilities_ts *= firstFactor[np.newaxis,:]
         #probabilities_ts *= secondFactor * thirdFactor
-        probabilities_ts *= secondFactor * ((1-self.lambda_)*thirdFactor + self.lambda_*forthFactor)
+        probabilities_ts *= secondFactor * ((1-self.lambda_)*thirdFactor + self.lambda_*forthFactor[:,np.newaxis])
         probabilities_ts /= np.sum(probabilities_ts)
 
         return probabilities_ts

jointtsmodel/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "1.5"
+__version__ = "1.6"
 
 from .JST import JST
 from .RJST import RJST

setup.py

@@ -9,7 +9,7 @@
 if __name__ == "__main__":
     setup(
         name="jointtsmodel",
-        version="1.5",
+        version="1.6",
         description="jointtsmodel - library of joint topic-sentiment models",
         long_description=long_description,
         long_description_content_type='text/markdown',