textflint.generation_layer.transformation.SM.swap_word

Replacing its words with antonyms provided by WordNet

class textflint.generation_layer.transformation.SM.swap_word.SwapWord(language='eng')

Bases: textflint.generation_layer.transformation.transformation.Transformation

Transforms an input by replacing its words with antonyms provided by WordNet. Download nltk_data before running.

Implement follow by Stress Test Evaluation for Natural Language Inference For the correctness of trasformation we swap the word has best_sense(Wordnet) to its antonym

https://www.aclweb.org/anthology/C18-1198/

Example:

{
    sentence1: I hate this book.
    sentence2: This book is my favorite.
    y: 0
}

__init__(language='eng')

Parameters

language (string) – language of transformation

transform(sample, n=1, **kwargs)

Transform data sample to a list of Sample.

Parameters

• sample (~SMSample) – Data sample for augmentation

• n (int) – Default is 1. MAX number of unique augmented output

• **kwargs –

Returns

Augmented data

class textflint.generation_layer.transformation.SM.swap_word.Transformation(**kwargs)

Bases: abc.ABC

An abstract class for transforming a sequence of text to produce a list of potential adversarial example.

processor = <textflint.common.preprocess.en_processor.EnProcessor object>

transform(sample, n=1, field='x', **kwargs)

Transform data sample to a list of Sample.

Parameters

• sample (Sample) – Data sample for augmentation.

• n (int) – Max number of unique augmented output, default is 5.

• field (str|list) – Indicate which fields to apply transformations.

• **kwargs (dict) –

  other auxiliary params.

Returns

list of Sample

classmethod sample_num(x, num)

Get ‘num’ samples from x.

Parameters

• x (list) – list to sample

• num (int) – sample number

Returns

max ‘num’ unique samples.

textflint.generation_layer.transformation.SM.swap_word.lesk(context_sentence, ambiguous_word, pos=None, synsets=None)

Return a synset for an ambiguous word in a context.

Parameters

• context_sentence (iter) – The context sentence where the ambiguous word occurs, passed as an iterable of words.

• ambiguous_word (str) – The ambiguous word that requires WSD.

• pos (str) – A specified Part-of-Speech (POS).

• synsets (iter) – Possible synsets of the ambiguous word.

Returns

lesk_sense The Synset() object with the highest signature overlaps.

This function is an implementation of the original Lesk algorithm (1986) [1].

Usage example:

>>> lesk(['I', 'went', 'to', 'the', 'bank', 'to', 'deposit', 'money', '.'], 'bank', 'n')
Synset('savings_bank.n.02')

[1] Lesk, Michael. “Automatic sense disambiguation using machine readable dictionaries: how to tell a pine cone from an ice cream cone.” Proceedings of the 5th Annual International Conference on Systems Documentation. ACM, 1986. http://dl.acm.org/citation.cfm?id=318728