Comparative Study and Evaluation of Machine Learning Models for Semantic Textual Similarity

Semantic Textual Similarity (STS) plays a critical role in various natural language processing (NLP) applications such as information retrieval, text summarization, and machine translation. This paper presents a comprehensive comparative study and evaluation of multiple machine learning models for STS. Our study evaluates NLP models on standard STS benchmark datasets, examining their performance based on accuracy and computational efficiency. We specifically emphasize benchmarking using cosine similarity and Root Mean Square Error (RMSE), in contrast to previous research that focused on cosine similarity and Pearson correlation. This study contributes to the ongoing development and optimization of STS benchmarking.