# Finding Friends and Flipping Frenemies: Automatic Paraphrase Dataset Augmentation Using Graph Theory

arXiv: https://arxiv.org/abs/2011.01856 · HTML: https://mlsanity.com/paper/2011.01856/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 3 Nov 2020.

## Built on by (most cited first, 8 of 8)

- [A Survey of Data Augmentation Approaches for NLP](https://mlsanity.com/paper/2105.03075/) (2021)
- [To Augment or Not to Augment? A Comparative Study on Text Augmentation Techniques for Low-Resource NLP](https://mlsanity.com/paper/2111.09618/) (2021)
- [GDA: Generative Data Augmentation Techniques for Relation Extraction Tasks](https://mlsanity.com/paper/2305.16663/) (2023)
- [Data Augmentation for Mental Health Classification on Social Media](https://mlsanity.com/paper/2112.10064/) (2021)
- [Bridging the Gap between Decision and Logits in Decision-based Knowledge Distillation for Pre-trained Language Models](https://mlsanity.com/paper/2306.08909/) (2023)
- [Data Augmentation Methods for Anaphoric Zero Pronouns](https://mlsanity.com/paper/2109.09825/) (2021)
- [Mitigating Data Scarcity for Large Language Models](https://mlsanity.com/paper/2302.01806/) (2023)
- [Cyber-Attack Technique Classification Using Two-Stage Trained Large Language Models](https://mlsanity.com/paper/2411.18755/) (2024)

## Abstract (arXiv)

Most NLP datasets are manually labeled, so suffer from inconsistent labeling or limited size. We propose methods for automatically improving datasets by viewing them as graphs with expected semantic properties. We construct a paraphrase graph from the provided sentence pair labels, and create an augmented dataset by directly inferring labels from the original sentence pairs using a transitivity property. We use structural balance theory to identify likely mislabelings in the graph, and flip their labels. We evaluate our methods on paraphrase models trained using these datasets starting from a pretrained BERT model, and find that the automatically-enhanced training sets result in more accurate models.