This Repository contains Python scripts and resulting datasets used in my paper for Scintific Writing Course at TU Wien in Witersemester 2023/24. In my paper, I aim to evaluate the capabilities of OpenAI’s API in conducting experiments similar to those presented by Camburu et al. in paper e-SNLI: Natural Language Inference with Natural Language Explanations. We will focus on how effectively the API can generate labels and explanations, as well as perform both tasks simultaneously, using selected sentence pairs from the e-SNLI dataset. My analysis will be constrained to a subset of examples due to the rate limits imposed by the OpenAI API, offering a focused investigation into its NLI performance.
The paper can also be found in the repository, see Enhancing Natural Language Understanding: A Practical Evaluation using e-SNLI and ChatGPT