Update recommender.py

This fixes the issue of program name ambiguity, as well as fixes the output format.

Author: nakuldahiwade

recommender.py

@@ -12,7 +12,6 @@
 tfidf = TfidfVectorizer(stop_words='english')
 
 metadata['shortDescription'] = metadata['shortDescription'].fillna('')
-
 tfidf_matrix = tfidf.fit_transform(metadata['shortDescription'])
 
 # Using cosine similarity between descriptions.
@@ -27,10 +26,13 @@ def get_recommendations(cliName, cosine_sim=cosine_sim, k=None):
  idx = indices[cliName]
 
  # If the Index object is iterable, it means there is ambiguity because of same program name.
- # Curently, we are choosing one and moving ahead, but this will be changed in future.
+ # As the same program can belong to different platforms with the same or different functionality.
+ # The ambiguity is resolved by providing a platformid 
  if hasattr(idx, '__iter__'):
  z = [x for x in idx]
- idx = z[1]
+ for indx in z:
+ if metadata['platformId'].iloc[indx] == platform_id:
+ idx = indx
 
  # Get the pairwsie similarity scores of all programs with that program
  sim_scores = list(enumerate(cosine_sim[idx]))
@@ -44,14 +46,13 @@ def get_recommendations(cliName, cosine_sim=cosine_sim, k=None):
  # Get the program indices
  program_indices = [i[0] for i in sim_scores]
 
- # import pdb; pdb.set_trace()
-
  # Return the top k or 5 most similar programs
  names = metadata['cliName'].iloc[program_indices].tolist()
  description = metadata['shortDescription'].iloc[program_indices].tolist()
- recs = {}
-
+ 
+ recs = dict()
+ recs["recommendations"] = []
  for x, y in zip(names, description):
- recs[x] = y
+ recs["recommendations"].append({"cliName": x, "shortDescription": y})
 
  return recs