Tests look strange

[mailgun]

Tests look strange

[justafucker]

What is also unclear to me is that in https://github.com/afedosenko/talon/blob/master/tests/signature/learning/featurespace_test.py

s = '''John Doe
VP Research and Development, Xxxx Xxxx Xxxxx
555-226-2345
[email protected]'''
    sender = 'John <[email protected]>'
    features = fs.features(sender)
    result = fs.apply_features(s, features)
    # note that we don't consider the first line because signatures don't
    # usually take all the text, empty lines are not considered
    eq_(result, [[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
                 [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                 [0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
                 [0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0]])

the last line contains 'john' which mean the last '0' should be '1'

[obukhov-sergey]

Hi @justafucker. Sorry for confusion and thanks for your interest / questions. Will try to explain them.

The 1st test checks that ['Sergey', 'Obukhov'] will be among extracted names - not that they are the only ones extracted. E.g. if you modify the test and add serobnic to the list the test will pass as well.

There is a test that specifically checks that given [email protected] we'll extract sergey: https://github.com/mailgun/talon/blob/master/tests/signature/learning/helpers_test.py#L103

But we definitely encourage you to submit a PR if you find tests / code confusing and wish to contribute / improve them.

Regarding your 2nd question. The algo looks for lines like "John Doe" or "John" or "Doe" i.e. a line should end with extracted name or extracted name should be a detached word. This requirement might seem strange in respect to "[email protected]" but in general it helps to avoid false positives when extracted name happens to be some general sequence of chars that might occurs in a line.