def compare_same_duration_ordinal(dataset, narrative, rel_diff=0.1):
questions = [
'Was the <O1> [sound event,sound] [roughly,approximately] as <D> as the <O2> [sound event,sound]?', # noqa: E501
'Was the <O1> and <O2> [sound events,sounds] [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> [sound event,sound] and the <O2> [sound event,sound], were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> [sound event,sound] and the <O2> [sound event,sound], did they [roughly,approximately] have the same duration?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> and <O2> [sound events,sounds], were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> and <O2> [sound events,sounds], did they [roughly,approximately] have the same duration?', # noqa: E501
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
number_1, ordinal_1 = sample_number(len(lst_events))
duration = sample_duration()
number_2, ordinal_2 = sample_second_number(len(lst_events), number_1)
assert number_1 != number_2, 'Question (compare_same_duration_ordinal) illposed.'
question = question.replace('<O1>', ordinal_1) # insert ordinal
question = question.replace('<D>', duration) # insert duration
question = question.replace('<O2>', ordinal_2) # insert ordinal
question = sanitize_question(question) # correct grammar
lst_duration = get_lst_durations(narrative)
e_1_duration = lst_duration[number_1 - 1]
e_2_duration = lst_duration[number_2 - 1]
rel_duration_diff = compute_rel_diff(np.array(e_1_duration),
np.array(e_2_duration))
# Assert a good margin in relative duration
assert np.sum(np.logical_and(rel_duration_diff > rel_diff,
rel_duration_diff < (2 * rel_diff))) <= 0, \
'Question (compare_same_duration_ordinal) illposed.'
answer = 'yes' if rel_duration_diff <= rel_diff else 'no'
return question, answer
def compare_same_duration_event_ordinal(dataset, narrative, rel_diff=0.1):
questions = [
'Was the <S> <A> [roughly,approximately] as <D> as the <O> [sound event,sound]?', # noqa: E501
'[Comparing,Listening to,Hearing] the <S> <A> and the <O> [sound event,sound], were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the sound of the <S> <A> and the <O> [sound event,sound], were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <S> <A> and the <O> [sound event,sound], did they [roughly,approximately] have the same duration?', # noqa: E501
'[Comparing,Listening to,Hearing] the sound of the <S> <A> and the <O> [sound event,sound], did they [roughly,approximately] have the same duration?', # noqa: E501
'Was the <O> [sound event,sound] [roughly,approximately] as <D> as the <S> <A>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O> [sound event,sound] and the <S> <A>, were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O> [sound event,sound] and the sound of the <S> <A>, were they [roughly,approximately] as <D>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O> [sound event,sound] and the <S> <A>, did they [roughly,approximately] have the same duration?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O> [sound event,sound] and the sound of the <S> <A>, did they [roughly,approximately] have the same duration?', # noqa: E501
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
unique_lst_events = [e for e in lst_events if lst_events.count(e) == 1]
assert len(unique_lst_events) > 0, \
'Question (compare_same_duration_event_ordinal) illposed.'
event = str(np.random.choice(unique_lst_events)) # sample event
source = str(np.random.choice(dataset['sources'][event]))
action = str(np.random.choice(dataset['actions'][event]))
duration = sample_duration()
number, ordinal = sample_second_number(len(lst_events),
lst_events.index(event) + 1)
assert lst_events.count(event) == 1, \
'Question (compare_same_duration_event_ordinal) illposed.'
assert lst_events.index(event) != (number - 1), \
'Question (compare_same_duration_event_ordinal) illposed.'
question = question.replace('<S>', source) # insert source
question = question.replace('<A>', action) # insert action
question = question.replace('<D>', duration) # insert duration
question = question.replace('<O>', ordinal) # insert ordinal
question = sanitize_question(question) # correct grammar
lst_duration = get_lst_durations(narrative)
e_1_duration = lst_duration[lst_events.index(event)]
e_2_duration = lst_duration[number - 1]
rel_duration_diff = compute_rel_diff(np.array(e_1_duration),
np.array(e_2_duration))
# Assert a good margin in relative duration
assert np.sum(np.logical_and(rel_duration_diff > rel_diff,
rel_duration_diff < (2 * rel_diff))) <= 0, \
'Question (compare_same_duration_event_ordinal) illposed.'
answer = 'yes' if rel_duration_diff <= rel_diff else 'no'
return question, answer
def compare_duration_ordinal_event(dataset, narrative, rel_diff=0.1):
questions = [
'Was the <O> [sound event,sound] <RD> than the <S> <A>?',
'Was the <O> [sound event,sound] <RD> than the sound of the <S> <A>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O> [sound event,sound] and the <S> <A>, was the former <RD>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <S> <A> and the <O> [sound event,sound], was the latter <RD>?', # noqa: E501
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
unique_lst_events = [e for e in lst_events if lst_events.count(e) == 1]
assert len(unique_lst_events) > 0, \
'Question (compare_duration_ordinal_event) illposed.'
event = str(np.random.choice(unique_lst_events)) # sample event
source = str(np.random.choice(dataset['sources'][event]))
action = str(np.random.choice(dataset['actions'][event]))
rel_duration = sample_rel_duration()
number, ordinal = sample_second_number(len(lst_events),
lst_events.index(event) + 1)
assert lst_events.count(event) == 1, \
'Question (compare_duration_ordinal_event) illposed.'
assert lst_events.index(event) != (number - 1), \
'Question (compare_duration_ordinal_event) illposed.'
question = question.replace('<S>', source) # insert source
question = question.replace('<A>', action) # insert action
question = question.replace('<RD>', rel_duration) # insert duration
question = question.replace('<O>', ordinal) # insert ordinal
question = sanitize_question(question) # correct grammar
lst_duration = get_lst_durations(narrative)
e_1_duration = lst_duration[number - 1]
e_2_duration = lst_duration[lst_events.index(event)]
# Assert a good margin in relative duration
rel_duration_diff = compute_rel_diff(np.array(e_1_duration),
np.array(e_2_duration))
assert np.sum(rel_duration_diff < rel_diff) <= 0, \
'Question (compare_duration_ordinal_event) illposed.'
if 'short' in question:
answer = 'yes' if e_1_duration < e_2_duration else 'no'
elif 'long' in question:
answer = 'yes' if e_1_duration > e_2_duration else 'no'
else:
assert False, \
'Duration illdefined in Question (compare_duration_ordinal_event).'
return question, answer
def compare_ordinal(dataset, narrative, _):
questions = [
'Was the <O1> [sound event,sound] [the same as,similar to] the <O2> [sound event,sound]?', # noqa: E501
'Was the <O1> [sound event,sound] and <O2> [sound event,sound] [the same,similar]?', # noqa: E501
'Were the <O1> and <O2> [sound events,sounds] [the same,similar]?',
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
number_1, ordinal_1 = sample_number(len(lst_events))
number_2, ordinal_2 = sample_second_number(len(lst_events), number_1)
assert number_1 != number_2, 'Question (compare_ordinal) illposed.'
question = question.replace('<O1>', ordinal_1) # insert ordinal
question = question.replace('<O2>', ordinal_2) # insert ordinal
question = sanitize_question(question) # correct grammar
answer = 'yes' if lst_events[number_1 - 1] == lst_events[number_2 - 1] \
else 'no'
return question, answer
def how_many_event_two_ordinal(dataset, narrative, _):
questions = ['How many times did you [hear,listen to] a sound that [sounded,seemed] like the <O1> [sound event,sound] [or,and] the <O2> [sound event,sound]?', # noqa: E501
'What is the number of times did you [hear,listen to] a sound that [sounded,seemed] like the <O1> [sound event,sound] [or,and] the <O2> [sound event,sound]?', # noqa: E501
'[Hearing,Listening to] the <O1> [sound event,sound] and the <O2> [sound event,sound], how many sounds were [the same,similar]?', # noqa: E501
'[Hearing,Listening to] the <O1> [sound event,sound] and the <O2> [sound event,sound], what is the number of sounds that were [the same,similar]?', # noqa: E501
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
number_1, ordinal_1 = sample_number(len(lst_events))
number_2, ordinal_2 = sample_second_number(len(lst_events), number_1)
question = question.replace('<O1>', ordinal_1) # insert ordinal
question = question.replace('<O2>', ordinal_2) # insert ordinal
question = sanitize_question(question) # correct grammar
event_1 = lst_events[number_1 - 1]
event_2 = lst_events[number_2 - 1]
answer = numbers_to_words((lst_events.count(event_1) - 1) # -1 for base event
+ (lst_events.count(event_2) - 1))
return question, answer
def compare_duration_ordinal(dataset, narrative, rel_diff=0.1):
questions = [
'Was the <O1> [sound event,sound] <RD> than the <O2> [sound event,sound]?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> [sound event,sound] and the <O2> [sound event,sound], was the former <RD>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O1> and <O2> [sound events,sounds], was the former <RD>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O2> [sound event,sound] and the <O1> [sound event,sound], was the latter <RD>?', # noqa: E501
'[Comparing,Listening to,Hearing] the <O2> and <O1> [sound events,sounds], was the latter <RD>?', # noqa: E501
]
question = str(np.random.choice(questions)) # sample question
lst_events = get_lst_events(narrative)
number_1, ordinal_1 = sample_number(len(lst_events))
rel_duration = sample_rel_duration()
number_2, ordinal_2 = sample_second_number(len(lst_events), number_1)
assert number_1 != number_2, 'Question (compare_duration_ordinal) illposed.'
question = question.replace('<O1>', ordinal_1) # insert ordinal
question = question.replace('<RD>', rel_duration) # insert duration
question = question.replace('<O2>', ordinal_2) # insert ordinal
question = sanitize_question(question) # correct grammar
lst_duration = get_lst_durations(narrative)
e_1_duration = lst_duration[number_1 - 1]
e_2_duration = lst_duration[number_2 - 1]
# Assert a good margin in relative duration
rel_duration_diff = compute_rel_diff(np.array(e_1_duration),
np.array(e_2_duration))
assert np.sum(rel_duration_diff < rel_diff) <= 0, \
'Question (compare_duration_ordinal) illposed.'
if 'short' in question:
answer = 'yes' if e_1_duration < e_2_duration else 'no'
elif 'long' in question:
answer = 'yes' if e_1_duration > e_2_duration else 'no'
else:
assert False, 'Duration illdefined in Question (compare_duration_ordinal).'
return question, answer
