Data frame:
pair = collections.defaultdict(collections.Counter)
e.g.
pair = {\'doc1\': {\'word1\':4, \'word2\':3},
new_pair = {} # simple dict at the top level for doc, tab in testing.form.items(): for word, freq in tab.items(): # top-level values is word counters new_pair[doc].setdefault(word, Counter()) += freq