ExpliCA Dataset This repository contains the resources and code for working with the ExpliCA dataset, which includes natural language explanations (NLEs), evaluators, and tools for automated evaluation. Below is an overview of each folder and its contents:
Folder Structure ch5 Contains resources, code, and data relevant to Chapter 5. This may include analysis scripts, model evaluations, or other chapter-specific files.
ch6 Contains resources, code, and data relevant to Chapter 6. Similar to the ch5 folder, this may include chapter-specific materials for experiments, results, and code examples.
Generated NLEs This folder contains generated Natural Language Explanations (NLEs) created through the ExpliCA model or other related methods. Each file in this directory may correspond to different model runs or parameter settings.
LLMasEvaluators Contains evaluators specifically designed for Large Language Models (LLMs). This may include scripts or configurations for evaluating explanations or model outputs using LLMs.
Original NLEs Stores the original, unmodified Natural Language Explanations (NLEs) from the dataset. This serves as a baseline or reference point against which generated NLEs can be compared.
Triple Extraction and Analysis Use the files in ExpliCA_stats and triple extraction to extract and analyze the linguistic triples or other statistics from explanations.
Requirements Python 3.x Any additional libraries or dependencies (e.g., transformers, nltk, scikit-learn) should be listed in a requirements.txt file.
Contact For questions or feedback, please reach out to Miruna Clinciu
Citation:
Clinciu, M.-A., Hastie, H., & Eshghi, A. (2024). ExpliCA Dataset [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14066155