Small Sample Intelligence Text Classification Model Based on ChatGPT and Neighboring Prompt

In the field of intelligence analysis, text classification poses a significant challenge, particularly under small sample scenarios. To address this issue, this study introduces a model for small sample intelligence text classification based on ChatGPT and the Prompt mode. The primary objective is to mitigate data scarcity issues and enhance the classification accuracy of intelligence text. Initially, a supervised contrast learning strategy is devised to enhance the text representation retrieval capabilities. Subsequently, a nearest-neighbor Prompt mechanism is developed through the implementation of the nearest-neighbor algorithm. Ultimately, this constructed nearest-neighbor Prompt mechanism is integrated into the ChatGPT model. This mechanism not only imparts pertinent knowledge to the ChatGPT model but also delineates the output format criteria, thereby facilitating the efficient transformation of text information into labels and substantially enhancing the model’s performance in intelligence text classification tasks. Experimental results demonstrate a substantial enhancement in the classification efficacy of small sample information text classification tasks by the proposed model, providing a viable solution for this domain and fostering its advancement.