Is Prompt the Future?

Abstract

A vast amount of unstructured data is being generated in the age of big data. Relation extraction (RE) is the critical way to improve the utility of the data by extracting structured data, which has seen a great evolution in recent years. This paper first introduces five paradigms of RE, namely the rule-based paradigm, the machine learning paradigm, the deep learning model-based paradigm, and the two types of current mainstream methods with pretrained language models. Based on the RE scenario, a comprehensive introduction is made for the currently popular paradigm with prompt learning, which is investigated regarding four aspects. The main contributions of this paper are as follows. Since big models are too large to be easily trained, prompt learning has become a promising research direction for RE, our work is, therefore, a systematic introduction to this paradigm for RE and compared with traditional paradigms. In addition, this paper summarizes the current problems faced by RE tasks and proposes valuable research directions with prompt learning.