def test_label(nlp: BertPipeline):
# We add the pattern recognizer to the pipeline.
pattern_recognizer = AttentionPatternRecognizer()
nlp.pattern_recognizer = pattern_recognizer
text_1 = ("We are great fans of Slack, but we wish the subscriptions "
"were more accessible to small startups.")
text_2 = "The Slack often has bugs."
text_3 = "best of all is the warm vibe"
aspect = "slack"
examples = [
Example(text_1, aspect),
Example(text_2, aspect),
Example(text_3, aspect)
]
tokenized_examples = nlp.tokenize(examples)
input_batch = nlp.encode(tokenized_examples)
output_batch = nlp.predict(input_batch)
labeled_examples = nlp.label(tokenized_examples, output_batch)
labeled_examples = list(labeled_examples)
labeled_1, labeled_2, labeled_3 = labeled_examples
assert labeled_1.sentiment == Sentiment.positive
assert labeled_2.sentiment == Sentiment.negative
assert isinstance(labeled_1.scores, list)
assert np.argmax(labeled_1.aspect_representation.look_at) == 5
assert np.argmax(labeled_2.aspect_representation.look_at) == 1
# We need to calibrate the model. The prediction should be neutral.
# In fact, the model does not recognize the aspect correctly.
assert labeled_3.sentiment == Sentiment.positive
assert np.allclose(labeled_3.aspect_representation.look_at,
[1.0, 0.16, 0.50, 0.54, 0.34, 0.39, 0.12],
atol=0.01)
def test_predict(nlp: BertPipeline):
text_1 = ("We are great fans of Slack, but we wish the subscriptions "
"were more accessible to small startups.")
text_2 = "We are great fans of Slack"
aspect = "Slack"
examples = [Example(text_1, aspect), Example(text_2, aspect)]
tokenized_examples = nlp.tokenize(examples)
input_batch = nlp.encode(tokenized_examples)
output_batch = nlp.predict(input_batch)
assert output_batch.scores.shape == [2, 3]
assert output_batch.hidden_states.shape == [2, 13, 23, 768]
assert output_batch.attentions.shape == [2, 12, 12, 23, 23]
assert output_batch.attention_grads.shape == [2, 12, 12, 23, 23]
scores = output_batch.scores.numpy()
assert np.argmax(scores, axis=-1).tolist() == [2, 2]
def test_encode(nlp: BertPipeline):
text_1 = ("We are great fans of Slack, but we wish the subscriptions "
"were more accessible to small startups.")
text_2 = "We are great fans of Slack"
aspect = "Slack"
examples = [Example(text_1, aspect), Example(text_2, aspect)]
tokenized_examples = nlp.tokenize(examples)
input_batch = nlp.encode(tokenized_examples)
assert isinstance(input_batch.token_ids, tf.Tensor)
# 101 the CLS token, 102 the SEP tokens.
token_ids = input_batch.token_ids.numpy()
values = [101, 2057, 2024, 2307, 4599, 1997, 19840, 102, 19840, 102]
assert token_ids[1, :10].tolist() == values
assert token_ids[0, :7].tolist() == values[:7]
# The second sequence should be padded (shorter),
# and attention mask should be set.
assert np.allclose(token_ids[1, 10:], 0)
attention_mask = input_batch.attention_mask.numpy()
assert np.allclose(attention_mask[1, 10:], 0)
# Check how the tokenizer marked the segments.
token_type_ids = input_batch.token_type_ids.numpy()
assert token_type_ids[0, -2:].tolist() == [1, 1]
assert np.allclose(token_type_ids[0, :-2], 0)