Search

Over the past few decades, short text classification has emerged as a critical downstream task in natural language processing (NLP). One crucial classification research issue is how to advance semantic understanding considering the short length, feature sparsity, and high ambiguity in short texts. Recently, prompt-tuning has been proposed to insert a template into the input and convert text classification tasks into equivalent cloze-style tasks. However, among most of the previous approaches, either the crafted template methods are time-consuming and labor-intensive, or automatic prompt generation methods cannot achieve satisfied performance. In this paper, we introduce a novel approach called Soft Knowledgeable Prompt-tuning for short text classification. Our method considers both the template generation and classification performance to construct prompts for label prediction. We employ five different strategies to expand the label words space for modifying soft prompts, and the integration of these strategies is used as the final verbalizer. Despite being automatic, experimental results show that our method achieved more desirable performance even than the crafted template methods, outperforming the state-of-the-art by more than 14 Accuracy points on four well-known benchmarks. • Both the automatic prompt generation and text classification performance are achieved. • 5 different strategies are used to construct verbalizer effectively and efficiently. • The comprehensive experiments evaluate the effectiveness of our proposed method. [ABSTRACT FROM AUTHOR]

Abstract: Classification is a supervised learning approach. However, it might be beyond one’s expectation to achieve high accuracy with bad quality of training instances. On the other hand, the classes (categories) for storing the instances were usually determined by related experts in the beginning. After certain period of time, therefore, it is attractive to know whether these existing classes are still suitable for storing new instances or not. In this paper we compute the value of Class Structure Ambiguity (CSA) of one class structure via Dynamic Centroid (DC) to evaluate the ambiguities among the classes. The DC approach was to have distinct centroids for one class from individual instance’s point of view, instead of having only one centroid for that class. To verify the CSA do reveal the ambiguity of class structure, the Pearson’s correlation between the values of the CSA and that of the accuracy achieved by SVM classifier was computed according to variant class structures which were generated manually with the ambiguous degrees under control. Experimental results showed that the values of the Pearson’s correlation were almost −1 as perfect as negative correlation. That meant the CSA did reveal the ambiguous degree of one class structure. [Copyright &y& Elsevier]