def test_model_extracts_nickname_symbol_filter(
model: DefinitionDetectionModel):
prediction_type = "DocDef2"
features = model.featurize("The agent acts with SYMBOL SYMBOL")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0] == 0
assert slots[prediction_type][0] == ["O", "O", "O", "O", "O", "O"]
def test_model_extracts_simple_definitions(model: DefinitionDetectionModel):
prediction_type = "W00"
features = model.featurize("Neural networks are machine learning models.")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"TERM", "TERM", "O", "DEF", "DEF", "DEF", "O"
]
def test_model_extracts_nickname_symbol_separated_by_colon(
model: DefinitionDetectionModel, ):
prediction_type = "DocDef2"
features = model.featurize("The agent acts with a policy : SYMBOL")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O", "O", "O", "O", "O", "DEF", "O", "TERM"
]
def test_model_extracts_nickname_symbol_parentheses(
model: DefinitionDetectionModel):
prediction_type = "DocDef2"
features = model.featurize("The agent acts with policy (SYMBOL)")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O", "O", "O", "O", "DEF", "O", "TERM", "O"
]
def test_model_extract_abbreviation_shortened_word(
model: DefinitionDetectionModel):
prediction_type = "AI2020"
features = model.featurize("We propose Conductive Networks (CondNets)")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O", "O", "DEF", "DEF", "O", "TERM", "O"
]
def test_model_extracts_nickname_for_th_index_pattern(
model: DefinitionDetectionModel):
prediction_type = "DocDef2"
features = model.featurize(
"This process repeats for every SYMBOLth timestep")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O", "O", "O", "O", "O", "TERM", "DEF"
]
def test_model_extract_abbreviation_acronym(model: DefinitionDetectionModel):
prediction_type = "AI2020"
features = model.featurize(
"We use a Convolutional Neural Network (CNN) based architecture")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O",
"O",
"O",
"DEF",
"DEF",
"DEF",
"O",
"TERM",
"O",
"O",
"O",
]
def test_model_extracts_nickname_before_symbol(
model: DefinitionDetectionModel):
prediction_type = "DocDef2"
features = model.featurize(
"The agent acts with a policy SYMBOL in each timestep SYMBOL")
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O",
"O",
"O",
"O",
"O",
"DEF",
"TERM",
"O",
"O",
"DEF",
"TERM",
]
def test_model_extracts_nickname_after_symbol(model: DefinitionDetectionModel):
prediction_type = "DocDef2"
features = model.featurize(
"The architecture consists of SYMBOL dense layers trained with SYMBOL learning rate"
)
intents, slots, _ = model.predict_batch([features], prediction_type)
assert intents[prediction_type][0]
assert slots[prediction_type][0] == [
"O",
"O",
"O",
"O",
"TERM",
"DEF",
"DEF",
"O",
"O",
"TERM",
"DEF",
"DEF",
]
def model():
model = DefinitionDetectionModel(["AI2020", "DocDef2", "W00"])
return model
def model():
model = DefinitionDetectionModel()
return model
def model():
model = DefinitionDetectionModel("DocDef2+AI2020+W00")
return model
