The exponential growth of biological literature has led to the accumulation of vast knowledge, encompassing various areas such as protein-protein interactions (PPIs), chemical-to-drug interactions, and drug-to-drug interactions (DDIs). Automatically identifying and categorizing biomedical associations provides significant advantages in diverse biomedical research fields. Over the past decade, notable progress has been made in identifying biomedical relationships. However, traditional models have primarily focused on PPIs and DDIs, often neglecting chemical-to-drug relations due to challenges related to extensive training data and accurate classification models.This study introduces a novel approach for extracting different chemical-to-drug relationships using a hybrid filtered-based text classification model. The proposed model incorporates a new measure of similarity between chemicals and drugs, as well as a maximized kernel learning-based SVM technique, which aims to identify crucial patterns within the training data. Experimental results demonstrate that the presented chemical-to-drug classification model outperforms existing interaction models in terms of efficiency.