Same-Radical Character Association Model in Text Classification

Research has verified that associative learning will improve learning efficiency during the process of recognizing Chinese characters. Therefore, some scholars believe that the associative information guided by radical molecules is the key to ensuring semantic understanding robustness and have incorporated it into their models. However, characters with the same radical are also often associated and widely used as contextual cues to enhance semantic understanding. This indicates that the information of the same radical has a positive effect in learning text. From the perspective of semantic complementarity between characters with the same radical, we drew inspiration and proposed a novel Same-radical Character Association Model (SCAM) for text classification in Chinese. SCAM aims to enhance model learning and training of text classification tasks by integrating same-radical characters. The model can be segmented into three processors, same radical extension processor, critical feature selection processor and cognitive and prediction processor. Experiment results shows the proposed model excels existing system in CNT, TNT and THUC databases, which indicates that integrating same-radical information will improve the Chinese text classification tasks.