Fix flaky test that calculated wrongly the ratio between keywords len…

…ght. (#72)

This was caused when calculating keywords using stopwords. As there are
more words that are filtering, the list of concepts gets smaller and the error
gets bigger.

Because of this, one of every five runs failed because the ratio between runs
was slightly different than two.

This was fixed expecting that the ratio of the runs is double with 5% tolerance
instead of up to one decimal place.

test/test_keywords.py

@@ -2,6 +2,7 @@
 
 from summa.keywords import keywords
 from summa.preprocessing.textcleaner import deaccent
+from numpy import isclose
 from .utils import get_text_from_test_data
 
 
@@ -94,7 +95,10 @@ def test_keywords_ratio_wstopwords(self):
         selected_docs_20 = keywords(text, ratio=0.2, split=True, additional_stopwords=additional_stoplist)
         selected_docs_40 = keywords(text, ratio=0.4, split=True, additional_stopwords=additional_stoplist)
 
-        self.assertAlmostEqual(float(len(selected_docs_40)) / len(selected_docs_20), 0.4 / 0.2, places=1)
+        actual_ratio = float(len(selected_docs_40)) / len(selected_docs_20)
+        expected_ratio = 0.4 / 0.2
+        # Expect the same ratio with a relative tolerance of 5%.
+        self.assertTrue(isclose(actual_ratio, expected_ratio, rtol=0.5), "Ratio between number of keywords should be 2.")
 
     def test_keywords_consecutive_keywords(self):
         text = "Rabbit populations known to be plentiful, large, and diverse \