textattack.constraints package

Constraints

Constraints determine whether a given transformation is valid. Since transformations do not perfectly preserve semantics semantics or grammaticality, constraints can increase the likelihood that the resulting transformation preserves these qualities. All constraints are subclasses of the Constraint abstract class, and must implement at least one of __call__ or call_many.

We split constraints into three main categories.

Semantics: Based on the meaning of the input and perturbation.

Grammaticality: Based on syntactic properties like part-of-speech and grammar.

Overlap: Based on character-based properties, like edit distance.

A fourth type of constraint restricts the search method from exploring certain parts of the search space:

pre_transformation: Based on the input and index of word replacement.

• textattack.constraints.grammaticality package
  • Grammaticality:
  • textattack.constraints.grammaticality.language_models package
    • non-pre Language Models:
    • textattack.constraints.grammaticality.language_models.google_language_model package
      • Google Language Models:
      • Google Language Models from Alzantot
      • GoogLMHelper
        • GoogLMHelper.clear_cache()
        • GoogLMHelper.get_words_probs()
        • GoogLMHelper.get_words_probs_uncached()
        • GoogLMHelper.CACHE_PATH
      • Google 1-Billion Words Language Model
      • GoogleLanguageModel
        • GoogleLanguageModel.check_compatibility()
        • GoogleLanguageModel.extra_repr_keys()
      • A library for loading 1B word benchmark dataset.
      • CharsVocabulary
        • CharsVocabulary.encode_chars()
        • CharsVocabulary.word_to_char_ids()
        • CharsVocabulary.max_word_length
        • CharsVocabulary.word_char_ids
      • LM1BDataset
        • LM1BDataset.get_batch()
        • LM1BDataset.vocab
      • Vocabulary
        • Vocabulary.decode()
        • Vocabulary.encode()
        • Vocabulary.id_to_word()
        • Vocabulary.word_to_id()
        • Vocabulary.bos
        • Vocabulary.eos
        • Vocabulary.size
        • Vocabulary.unk
      • get_batch()
      • Utils for loading 1B word benchmark dataset.
      • LoadModel()
    • textattack.constraints.grammaticality.language_models.learning_to_write package
      • “Learning To Write”
      • AdaptiveSoftmax
      • AdaptiveLoss
        • AdaptiveLoss.forward()
        • AdaptiveLoss.remap_target()
        • AdaptiveLoss.reset()
        • AdaptiveLoss.training
      • AdaptiveSoftmax
        • AdaptiveSoftmax.forward()
        • AdaptiveSoftmax.log_prob()
        • AdaptiveSoftmax.reset()
        • AdaptiveSoftmax.set_target()
        • AdaptiveSoftmax.training
      • Language model helpers
      • QueryHandler
        • QueryHandler.load_model()
        • QueryHandler.query()
        • QueryHandler.try_query()
      • util_reverse()
      • “Learning To Write” Language Model
      • LearningToWriteLanguageModel
        • LearningToWriteLanguageModel.get_log_probs_at_index()
        • LearningToWriteLanguageModel.CACHE_PATH
      • RNN Language Model
      • RNNModel
        • RNNModel.forward()
        • RNNModel.init_hidden()
        • RNNModel.init_weights()
        • RNNModel.training
    • GPT2 Language Models:
    • GPT2
      • GPT2.get_log_probs_at_index()
    • Language Models Constraint
    • LanguageModelConstraint
      • LanguageModelConstraint.extra_repr_keys()
      • LanguageModelConstraint.get_log_probs_at_index()
  • CoLA for Grammaticality
  • COLA
    • COLA.clear_cache()
    • COLA.extra_repr_keys()
  • LanguageTool Grammar Checker
  • LanguageTool
    • LanguageTool.extra_repr_keys()
    • LanguageTool.get_errors()
  • Part of Speech Constraint
  • PartOfSpeech
    • PartOfSpeech.check_compatibility()
    • PartOfSpeech.clear_cache()
    • PartOfSpeech.extra_repr_keys()
• textattack.constraints.overlap package
  • Overlap Constraints
  • BLEU Constraints
  • BLEU
    • BLEU.extra_repr_keys()
  • chrF Constraints
  • chrF
    • chrF.extra_repr_keys()
  • Edit Distance Constraints
  • LevenshteinEditDistance
    • LevenshteinEditDistance.extra_repr_keys()
  • Max Perturb Words Constraints
  • MaxWordsPerturbed
    • MaxWordsPerturbed.extra_repr_keys()
  • METEOR Constraints
  • METEOR
    • METEOR.extra_repr_keys()
• textattack.constraints.pre_transformation package
  • Pre-Transformation:
  • Input Column Modification
  • InputColumnModification
    • InputColumnModification.extra_repr_keys()
  • Max Modification Rate
  • MaxModificationRate
    • MaxModificationRate.extra_repr_keys()
  • Max Word Index Modification
  • MaxWordIndexModification
    • MaxWordIndexModification.extra_repr_keys()
  • Min Word Lenth
  • MinWordLength
  • Repeat Modification
  • RepeatModification
  • Stopword Modification
  • StopwordModification
    • StopwordModification.check_compatibility()
• textattack.constraints.semantics package
  • Semantic Constraints
  • textattack.constraints.semantics.sentence_encoders package
    • Sentence Encoder Constraint
    • textattack.constraints.semantics.sentence_encoders.bert package
      • BERT
      • sBERT for Sentence Similarity
      • BERT
        • BERT.encode()
    • textattack.constraints.semantics.sentence_encoders.infer_sent package
      • infer sent
      • infer sent for sentence similarity
      • InferSent
        • InferSent.encode()
        • InferSent.get_infersent_model()
        • InferSent.MODEL_PATH
        • InferSent.WORD_EMBEDDING_PATH
      • Infer sent model
      • InferSentModel
        • InferSentModel.build_vocab()
        • InferSentModel.build_vocab_k_words()
        • InferSentModel.encode()
        • InferSentModel.forward()
        • InferSentModel.get_batch()
        • InferSentModel.get_w2v()
        • InferSentModel.get_w2v_k()
        • InferSentModel.get_word_dict()
        • InferSentModel.is_cuda()
        • InferSentModel.prepare_samples()
        • InferSentModel.set_w2v_path()
        • InferSentModel.tokenize()
        • InferSentModel.update_vocab()
        • InferSentModel.training
    • textattack.constraints.semantics.sentence_encoders.universal_sentence_encoder package
      • Universal sentence encoder
      • multilingual universal sentence encoder
      • MultilingualUniversalSentenceEncoder
        • MultilingualUniversalSentenceEncoder.encode()
      • universal sentence encoder class
      • UniversalSentenceEncoder
        • UniversalSentenceEncoder.encode()
    • Sentence Encoder Class
    • SentenceEncoder
      • SentenceEncoder.encode()
      • SentenceEncoder.extra_repr_keys()
    • get_angular_sim()
    • get_neg_euclidean_dist()
    • Thought Vector Class
    • ThoughtVector
      • ThoughtVector.clear_cache()
      • ThoughtVector.encode()
      • ThoughtVector.extra_repr_keys()
  • BERT Score
  • BERTScore
    • BERTScore.extra_repr_keys()
    • BERTScore.SCORE_TYPE2IDX
  • Word Embedding Distance
  • WordEmbeddingDistance
    • WordEmbeddingDistance.check_compatibility()
    • WordEmbeddingDistance.extra_repr_keys()
    • WordEmbeddingDistance.get_cos_sim()
    • WordEmbeddingDistance.get_mse_dist()

TextAttack Constraint Class

class textattack.constraints.constraint.Constraint(compare_against_original)

Bases: ReprMixin, ABC

An abstract class that represents constraints on adversial text examples. Constraints evaluate whether transformations from a AttackedText to another AttackedText meet certain conditions.

Parameters:

compare_against_original (bool) – If True, the reference text should be the original text under attack. If False, the reference text is the most recent text from which the transformed text was generated. All constraints must have this attribute.

call_many(transformed_texts, reference_text)

Filters transformed_texts based on which transformations fulfill the constraint. First checks compatibility with latest Transformation, then calls _check_constraint_many

Parameters:

• transformed_texts (list[AttackedText]) – The candidate transformed AttackedText’s.

• reference_text (AttackedText) – The AttackedText to compare against.

check_compatibility(transformation)

Checks if this constraint is compatible with the given transformation. For example, the WordEmbeddingDistance constraint compares the embedding of the word inserted with that of the word deleted. Therefore it can only be applied in the case of word swaps, and not for transformations which involve only one of insertion or deletion.

Parameters:

transformation – The Transformation to check compatibility with.

extra_repr_keys()

Set the extra representation of the constraint using these keys.

To print customized extra information, you should reimplement this method in your own constraint. Both single-line and multi- line strings are acceptable.

Pre-Transformation Constraint Class

class textattack.constraints.pre_transformation_constraint.PreTransformationConstraint

Bases: ReprMixin, ABC

An abstract class that represents constraints which are applied before the transformation.

These restrict which words are allowed to be modified during the transformation. For example, we might not allow stopwords to be modified.

check_compatibility(transformation)

Checks if this constraint is compatible with the given transformation. For example, the WordEmbeddingDistance constraint compares the embedding of the word inserted with that of the word deleted. Therefore it can only be applied in the case of word swaps, and not for transformations which involve only one of insertion or deletion.

Parameters:

transformation – The Transformation to check compatibility with.

extra_repr_keys()

Set the extra representation of the constraint using these keys.

To print customized extra information, you should reimplement this method in your own constraint. Both single-line and multi- line strings are acceptable.