Skip to content

Issue 374 add representation tests #381

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
May 9, 2023
Merged

Issue 374 add representation tests #381

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d8a08d9
addd gender representation test
ArshaanNazir May 8, 2023
08f7373
add supported tasks
ArshaanNazir May 8, 2023
0dc9d4a
update helper util funcs for representation tests
ArshaanNazir May 8, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
56 changes: 38 additions & 18 deletions nlptest/transform/representation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ class BaseRepresentation(ABC):
based on the implemented representation measure.
"""
alias_name = None
supported_tasks = ["ner", "text-classification"]
supported_tasks = ["ner", "text-classification","question-answering"]

@staticmethod
@abstractmethod
Expand Down Expand Up @@ -80,6 +80,8 @@ class GenderRepresentation(BaseRepresentation):
"min_gender_representation_count",
"min_gender_representation_proportion"
]

supported_tasks = ["ner", "text-classification","question-answering"]

def transform(test, data, params):
"""
Expand Down Expand Up @@ -155,30 +157,42 @@ async def run(sample_list: List[Sample], model: ModelFactory, **kwargs) -> List[
"""

progress = kwargs.get("progress_bar", False)

classifier = GenderClassifier()
genders = [classifier.predict(sample.original)
for sample in kwargs['raw_data']]
for sample in kwargs['raw_data']:

if "task" in sample.__annotations__:
if "perturbed_context" in sample.__annotations__:
genders = [classifier.predict(sample.original_context)
for sample in kwargs['raw_data']]
else:
genders = [classifier.predict(sample.original_question)
for sample in kwargs['raw_data']]

else:
genders = [classifier.predict(sample.original)
for sample in kwargs['raw_data']]

gender_counts = {
"male": len([x for x in genders if x == "male"]),
"female": len([x for x in genders if x == "female"]),
"unknown": len([x for x in genders if x == "unknown"])
}
"male": len([x for x in genders if x == "male"]),
"female": len([x for x in genders if x == "female"]),
"unknown": len([x for x in genders if x == "unknown"])
}

total_samples = len(kwargs['raw_data'])

for sample in sample_list:
if progress:
if progress:
progress.update(1)
if sample.test_type == "min_gender_representation_proportion":
sample.actual_results = MinScoreOutput(
min_score=round(gender_counts[sample.test_case]/total_samples, 2))
sample.state = "done"
elif sample.test_type == "min_gender_representation_count":
sample.actual_results = MinScoreOutput(
min_score=round(gender_counts[sample.test_case], 2))
sample.state = "done"

if sample.test_type == "min_gender_representation_proportion":
sample.actual_results = MinScoreOutput(
min_score=round(gender_counts[sample.test_case]/total_samples, 2))
sample.state = "done"

elif sample.test_type == "min_gender_representation_count":
sample.actual_results = MinScoreOutput(
min_score=round(gender_counts[sample.test_case], 2))
sample.state = "done"
return sample_list


Expand All @@ -196,6 +210,8 @@ class EthnicityRepresentation(BaseRepresentation):
"min_ethnicity_name_representation_count",
"min_ethnicity_name_representation_proportion"
]

supported_tasks = ["ner", "text-classification","question-answering"]

def transform(test, data, params):
"""
Expand Down Expand Up @@ -324,6 +340,8 @@ class LabelRepresentation(BaseRepresentation):
"min_label_representation_count",
"min_label_representation_proportion"
]

supported_tasks = ["ner", "text-classification"]

def transform(test, data, params):
"""
Expand Down Expand Up @@ -454,7 +472,7 @@ class ReligionRepresentation(BaseRepresentation):
"min_religion_name_representation_count",
"min_religion_name_representation_proportion"
]
supported_tasks = ["ner", "text-classification"]
supported_tasks = ["ner", "text-classification","question-answering"]

def transform(test, data, params):
"""
Expand Down Expand Up @@ -594,6 +612,8 @@ class CountryEconomicRepresentation(BaseRepresentation):
"min_country_economic_representation_count",
"min_country_economic_representation_proportion"
]

supported_tasks = ["ner", "text-classification","question-answering"]

def transform(test, data, params):
"""
Expand Down
20 changes: 19 additions & 1 deletion nlptest/transform/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7071,6 +7071,11 @@ def get_country_economic_representation_dict(data: List[Sample]) -> Dict[str, in
words = [x.span.word for x in sample.expected_results.predictions]
elif isinstance(sample.expected_results, SequenceClassificationOutput):
words = sample.original.split()
else:
if "perturbed_context" in sample.__annotations__:
words = sample.original_context.split()
else:
words = sample.original_question.split()
for i in words:
if check_name(i, [country_economic_dict['High-income']]):
country_economic_representation["high_income"] += 1
Expand Down Expand Up @@ -7101,6 +7106,11 @@ def get_religion_name_representation_dict(data: List[Sample]) -> Dict[str, int]:
words = [x.span.word for x in sample.expected_results.predictions]
elif isinstance(sample.expected_results, SequenceClassificationOutput):
words = sample.original.split()
else:
if "perturbed_context" in sample.__annotations__:
words = sample.original_context.split()
else:
words = sample.original_question.split()
for i in words:
if check_name(i, [religion_wise_names['Muslim']]):
religion_representation["muslim"] += 1
Expand Down Expand Up @@ -7136,6 +7146,11 @@ def get_ethnicity_representation_dict(data: List[Sample]) -> Dict[str, int]:
words = [x.span.word for x in sample.expected_results.predictions]
elif isinstance(sample.expected_results, SequenceClassificationOutput):
words = sample.original.split()
else:
if "perturbed_context" in sample.__annotations__:
words = sample.original_context.split()
else:
words = sample.original_question.split()
for i in words:
if check_name(i, [white_names['first_names'], white_names['last_names']]):
ethnicity_representation["white"] += 1
Expand Down Expand Up @@ -7165,7 +7180,10 @@ def get_entity_representation_proportions(entity_representation):
total_entities = sum(entity_representation.values())
entity_representation_proportion = {}
for k, v in entity_representation.items():
entity_representation_proportion[k] = v / total_entities
if total_entities == 0:
entity_representation_proportion[k] = 0
else:
entity_representation_proportion[k] = v / total_entities

return entity_representation_proportion

Expand Down