Perturbations

With their success on text classification, we examine the robustness of summarization models against adversarial perturbations, which can be in different levels – character, word, sentence, and document. The space of possible modifications at every level is huge. We show how an attacker, leveraging the biases in summarization models, can implement sentence exclusion attack which can also result in quality degradation.

Character Level

The main motivation behind applying these perturbations is that they can be used to simulate common typo errors and input noise that can occur in real-world scenarios.

Charater swap

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation word_delete

Charater Delete

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation character_delete

Charater Replace

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation character_replace

Charater Insert

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation character_insert

Charater Repeat

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation character_repeat

Word Level

Just like the character level, it simulates common typo errors and input noise that can occur in real-world scenarios.

Word Delete

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation word_delete

Word Synonym

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation word_synonym

Replacement with Homoglyphs

Homoglyphs are visually similar characters/ words that are less noticeable to human readers and can be used for deceptive purposes. To assess the models’ performance, one character or word at a time was replaced with its homoglyph counterpart.

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation word_homograph

Sentence/Document Reordering

In natural language, the order of sentences and paragraphs is important as they account for understanding the context. This can be disrupted due to formatting issues or intentional manipulation. We evaluate the models’ robustness against such changes in structure by moving one of the sentences in a document from the top to the bottom, and in the case of documents, by placing the top document at the bottom

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation document_reorder

Sentence Paraphrasing

Paraphrasing is a common phenomenon in natural language, and models should be able to handle paraphrased expressions while capturing the core meaning. With this perturbation, we test the models’ ability to summarize effectively without any change, while replacing the original sentence with its paraphrased version.

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation sentence_paraphrase

Replacement of original sentences

Once these perturbations are executed at the character, word, and sentence level, we replace the original sentences with the sentences containing them. In case of document perturbations, we just rearrange the order of documents and observe the model’s capability to identify the document again.

summarization_robustness --model t5-small --dataset yaolu/multi_x_science_sum --split validation --size 50 --perturbation sentence_reorder