Entailment Relation Aware Paraphrase Generation
Authors: Abhilasha Sancheti, Balaji Vasan Srinivasan, Rachel Rudinger11258-11266
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | A combination of automated and human evaluations show that ERAP generates paraphrases conforming to the specified entailment relation and are of good quality as compared to the baselines and uncontrolled paraphrasing systems. Using ERAP for augmenting training data for downstream textual entailment task improves performance over an uncontrolled paraphrasing system, and introduces fewer training artifacts, indicating the benefit of explicit control during paraphrasing. |
| Researcher Affiliation | Collaboration | Abhilasha Sancheti,1,2 Balaji Vasan Srinivasan,2 Rachel Rudinger1 1University of Maryland, College Park 2Adobe Research sancheti@umd.edu, balsrini@adobe.com, rudinger@umd.edu |
| Pseudocode | No | The paper includes a system diagram (Figure 2) but does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper does not provide a direct link to source code or an explicit statement about releasing the code for the described methodology. |
| Open Datasets | Yes | We train the NLI classifier o on existing NLI datasets namely, MNLI (Williams, Nangia, and Bowman 2018), SNLI (Bowman et al. 2015a), SICK (Marelli et al. 2014) as well as diagnostic datasets such as, HANS (Mc Coy, Pavlick, and Linzen 2019), others introduced in Glockner, Shwartz, and Goldberg (2018); Min et al. (2020)... we pre-train the generator on existing large paraphrase corpora e.g., Para Bank (Hu et al. 2019b) or Para NMT (Wieting and Gimpel 2018)... |
| Dataset Splits | Yes | Recasted SICK SICK NLI Split Others E N C Train 1344 684 684 420 1274 2524 641 Dev 196 63 63 43 143 281 71 Test 1386 814 814 494 1404 2790 712 |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'RoBERTa-based' and 'transformer-based' models but does not provide specific version numbers for software libraries or dependencies (e.g., PyTorch, TensorFlow, specific HuggingFace Transformers versions). |
| Experiment Setup | Yes | where α, β, δ, and n are hyperparameters empirically set to 0.4, 0.4, 0.2, and 2, respectively. |