David Evans, Yangfeng Ji, Hannah Chen
We lifted 4 functions out of this paper's own repositories and ran 3 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| hannahxchen/automatic-paraphrase-dataset-augmentation | — | 3 of 4 |
| Function | Status | Where it lives |
|---|---|---|
| generate_augmented | Ran | hannahxchen/automatic-paraphrase-dataset-augmentation/generate_qqp_datasets.py pointer only (licence: NONE) · get_code("eaf3fa88029b9cfb") |
| infer_non_paraphrases | Ran | hannahxchen/automatic-paraphrase-dataset-augmentation/generate_qqp_datasets.py pointer only (licence: NONE) · get_code("740600e10c3a2d5d") |
| infer_transitive | Ran | hannahxchen/automatic-paraphrase-dataset-augmentation/generate_qqp_datasets.py pointer only (licence: NONE) · get_code("58f0ffdd54cb5c69") |
| find_mislabeled_pairs | Not yet run | hannahxchen/automatic-paraphrase-dataset-augmentation/generate_qqp_datasets.py pointer only (licence: NONE) · get_code("7b8770f6b596d35f") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
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.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2011.01856")
get_code_for_paper("2011.01856")
have("2011.01856")
Connect an agent — have() is free.