def test_disjoint_windows_04(self):
tokens = "a b c d e f g h i j k".split()
wins = ["a b c".split(), "e f g".split(), "i j k".split()]
output = windows.disjoint_windows(tokens,
window_size=3,
strategy="spread")
self.assertEqual([o.tokens for o in output], wins)
def test_disjoint_windows_01(self):
tokens = "a b c d e f g h i j k".split()
wins = ["a b c d".split(), "e f g h".split()]
output = windows.disjoint_windows(tokens,
window_size=4,
strategy="left")
self.assertEqual([o.tokens for o in output], wins)
def test_disjoint_windows_03(self):
tokens = "a b c d e f g h i j k l m n".split()
wins = ["c d e f g".split(), "h i j k l".split()]
output = windows.disjoint_windows(tokens,
window_size=5,
strategy="center")
self.assertEqual([o.tokens for o in output], wins)
def bootstrap(measure, tokens, window_size, strategy="spread", **kwargs):
"""Calculate bootstrap for surface-based measures as explained in
Evert et al. (2017).
kwargs are passed to measure
Evert, Stefan, Sebastian Wankerl, Elmar Nöth (2017). Reliable
measures of syntactic and lexical complexity: The case of Iris
Murdoch. In: Proceedings of the Corpus Linguistics 2017
Conference, Birmingham, UK.
http://purl.org/stefan.evert/PUB/EvertWankerlNoeth2017.pdf
"""
results = []
for window in windows.disjoint_windows(tokens, window_size, strategy):
results.append(measure(window, **kwargs))
if len(results) == 1:
return results[0], 0, results
return statistics.mean(results), misc.confidence_interval(results), results