def test_occupational_scoped_skill_extractor_candidate_skills():
extractor = SocScopedExactMatchSkillExtractor(sample_ontology())
candidate_skills = sorted(extractor.candidate_skills(sample_job_posting()),
key=lambda cs: cs.skill_name)
assert candidate_skills[0].skill_name == 'organization'
assert candidate_skills[
0].context == 'Organization, Cleanliness, Trainability, team player, good communication skillz, Motivation, a Sense of Responsibility and Pride in your Performance'
assert candidate_skills[0].confidence == 100
def test_occupational_scoped_skill_extractor_candidate_skills():
with utils.makeNamedTemporaryCSV(sample_skills(), '\t') as skills_filename:
extractor = SocScopedExactMatchSkillExtractor(
skill_lookup_path=skills_filename)
candidate_skills = sorted(extractor.candidate_skills(
sample_job_posting()),
key=lambda cs: cs.skill_name)
assert candidate_skills[0].skill_name == 'organization'
assert candidate_skills[
0].context == 'Organization, Cleanliness, Trainability, team player, good communication skillz, Motivation, a Sense of Responsibility and Pride in your Performance'
assert candidate_skills[0].confidence == 100
def _compute_func_on_one(self):
corpus_creator = SimpleCorpusCreator()
skill_extractor = SocScopedExactMatchSkillExtractor(
skill_lookup_path=self.skill_lookup_path)
def func(job_posting):
count_dict = skill_extractor.document_skill_counts(
soc_code=job_posting.get('onet_soc_code', '99-9999.00'),
document=corpus_creator._transform(job_posting))
count_lists = [[k] * v for k, v in count_dict.items()]
flattened = [
count for countlist in count_lists for count in countlist
]
return {self.property_name: flattened}
return func
def test_occupation_scoped_freetext_skill_extractor():
content = [[
'', 'O*NET-SOC Code', 'Element ID', 'ONET KSA', 'Description',
'skill_uuid', 'nlp_a'
],
[
'1', '11-1011.00', '2.a.1.a', 'reading comprehension',
'...', '2c77c703bd66e104c78b1392c3203362',
'reading comprehension'
],
[
'2', '11-1011.00', '2.a.1.b', 'active listening', '...',
'a636cb69257dcec699bce4f023a05126', 'active listening'
]]
with utils.makeNamedTemporaryCSV(content, '\t') as skills_filename:
extractor = SocScopedExactMatchSkillExtractor(
skill_lookup_path=skills_filename)
documents = [
{
'soc_code': '11-1011.00',
'document': 'this is a job that needs active listening',
'expected_value': Counter({'active listening': 1})
},
{
'soc_code': '11-1011.00',
'document': 'this is a reading comprehension job',
'expected_value': Counter({'reading comprehension': 1})
},
{
'soc_code': '11-1011.00',
'document': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'soc_code':
'11-1011.00',
'document':
'this is a reading comprehension and active listening job',
'expected_value':
Counter({
'active listening': 1,
'reading comprehension': 1
})
},
{
'soc_code': '11-1021.00',
'document': 'this is a job that needs active listening',
'expected_value': Counter()
},
{
'soc_code': '11-1021.00',
'document': 'this is a reading comprehension job',
'expected_value': Counter()
},
{
'soc_code': '11-1021.00',
'document': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'soc_code': '11-1021.00',
'document':
'this is a reading comprehension and active listening job',
'expected_value': Counter()
},
{
'soc_code': None,
'document': 'this is a job that needs active listening',
'expected_value': Counter()
},
{
'soc_code': None,
'document': 'this is a reading comprehension job',
'expected_value': Counter()
},
{
'soc_code': None,
'document': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'soc_code': None,
'document':
'this is a reading comprehension and active listening job',
'expected_value': Counter()
},
]
for document in documents:
assert extractor.document_skill_counts(
soc_code=document['soc_code'],
document=document['document']) == document['expected_value']
def test_occupation_scoped_freetext_skill_extractor():
ontology = CompetencyOntology(
competency_name='Sample Framework',
competency_description='A few basic competencies',
edges=[
CompetencyOccupationEdge(
competency=Competency(identifier='2.a.1.a',
name='Reading Comprehension'),
occupation=Occupation(identifier='11-1011.00')),
CompetencyOccupationEdge(
competency=Competency(identifier='2.a.1.b',
name='Active Listening'),
occupation=Occupation(identifier='11-1011.00')),
])
extractor = SocScopedExactMatchSkillExtractor(ontology)
documents = [
{
'id': '1234',
'@type': 'JobPosting',
'onet_soc_code': '11-1011.00',
'description': 'this is a job that needs active listening',
'expected_value': Counter({'active listening': 1})
},
{
'id': '2234',
'@type': 'JobPosting',
'onet_soc_code': '11-1011.00',
'description': 'this is a reading comprehension job',
'expected_value': Counter({'reading comprehension': 1})
},
{
'id': '3234',
'@type': 'JobPosting',
'onet_soc_code': '11-1011.00',
'description': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'id':
'4234',
'@type':
'JobPosting',
'onet_soc_code':
'11-1011.00',
'description':
'this is a reading comprehension and active listening job',
'expected_value':
Counter({
'active listening': 1,
'reading comprehension': 1
})
},
{
'id': '5234',
'@type': 'JobPosting',
'onet_soc_code': '11-1021.00',
'description': 'this is a job that needs active listening',
'expected_value': Counter()
},
{
'id': '6234',
'@type': 'JobPosting',
'onet_soc_code': '11-1021.00',
'description': 'this is a reading comprehension job',
'expected_value': Counter()
},
{
'id': '7234',
'@type': 'JobPosting',
'onet_soc_code': '11-1021.00',
'description': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'id': '8234',
'@type': 'JobPosting',
'onet_soc_code': '11-1021.00',
'description':
'this is a reading comprehension and active listening job',
'expected_value': Counter()
},
{
'id': '9234',
'@type': 'JobPosting',
'onet_soc_code': None,
'description': 'this is a job that needs active listening',
'expected_value': Counter()
},
{
'id': '1334',
'@type': 'JobPosting',
'onet_soc_code': None,
'description': 'this is a reading comprehension job',
'expected_value': Counter()
},
{
'id': '1434',
'@type': 'JobPosting',
'onet_soc_code': None,
'description': 'this is an active and reading listening job',
'expected_value': Counter(),
},
{
'id': '1534',
'@type': 'JobPosting',
'onet_soc_code': None,
'description':
'this is a reading comprehension and active listening job',
'expected_value': Counter()
},
]
for document in documents:
assert extractor.document_skill_counts(
document) == document['expected_value']
from skills_ml.evaluation.skill_extractors import candidate_skills_from_sample, metrics_for_candidate_skills
from skills_ml.evaluation.skill_extraction_metrics import TotalOccurrences, TotalVocabularySize, OntologyCompetencyRecall
from skills_ml.job_postings.common_schema import JobPostingCollectionSample
from tests.utils import sample_factory
sample = sample_factory(JobPostingCollectionSample())
print('Building ONET, may take a while to download')
full_onet = Onet()
skill_extractors = [
SectionExtractSkillExtractor(),
SkillEndingPatternExtractor(only_bulleted_lines=False),
AbilityEndingPatternExtractor(only_bulleted_lines=False),
FuzzyMatchSkillExtractor(full_onet.competency_framework),
ExactMatchSkillExtractor(full_onet.competency_framework),
SocScopedExactMatchSkillExtractor(full_onet)
]
print('Done building ONET! Now subsetting ONET into K,S,A')
metric_ontologies = [
full_onet,
full_onet.filter_by(lambda edge: 'Knowledge' in edge.competency.categories,
competency_name='onet_knowledge',
competency_description='ONET Knowledge'),
full_onet.filter_by(lambda edge: 'Abilities' in edge.competency.categories,
competency_name='onet_ability',
competency_description='ONET Ability'),
full_onet.filter_by(lambda edge: 'Skills' in edge.competency.categories,
competency_name='onet_skill',
competency_description='ONET Skill')
]
metrics = [