Update README.md

README.md

@@ -1,3 +1,15 @@
 ---
 license: mit
 ---
+This repository is a public repository of the data used in the paper "CReSE: Enhancing Clinical Trial Design via Contrastive Learning and Rephrasing-based and Clinical Relevance-preserving Sentence Embedding" (under review).
+
+There are three main types of data stored in the repository.
+
+1) Positive-negative EC-title pairs: A dataset that pairs the ECs used in a study with the study's title and other design information. It can be used to train EC recommendation models (binary classification). Different datasets may be available depending on when the trial information was uploaded, the number of ECs in the trial, etc.
+
+2) original-rephrased EC pairs: The original-rephrased EC pairs data used to develop the CReSE model. EC rephrasing was performed using ChatGPT (gpt-3.5-turbo).
+
+3) Clinical relevance data between EC pairs: A dataset evaluating the clinical relevance between different ECs created to evaluate the EC clustering performance of the CReSE model. It was also created using ChatGPT (gpt-3.5-turbo).
+
+Please refer to our paper for more specific data generation conditions and related prompts.
+