CoT-STS: A Zero Shot Chain-of-Thought Prompting for Semantic Textual Similarity
Musarrat Hussain, Ubaid Ur Rehman, Tri D.T. Nguyen, Sungyoung Lee · 2023
The emergence of Large Language Models (LLMs) have revolutionized the field of Natural Language Processing (NLP) by changing the focus of technical development from features engineering, architecture engineering, and objective engineering to prompt engineering. The main goal of the prompt engineering is to craft clear and concise instructions, known as input prompts, for LLMs to effectively perform the targeted NLP task. Semantic Textual Similarity (STS) is one such significant NLP task, which aims to assess the similarity between the semantic meanings of two input sentences. Numerous approaches have been proposed in the literature, including syntactic similarity evaluations, word-embedding based methods, and dedicated model training. However, these approaches require substantial effort, such as creating extensive annotated datasets and training dedicated STS models.