Text classification is considered one of the most fundamental and essential problems that deal with automatically classifying textual resources into pre-defined categories. Numerous algorithms, datasets, and evaluation measurements have been proposed to address the task. Within the era of information redundancy, it is challenging and time-consuming to engineering a sizable amount of data in multi-languages manually. However, it is time-consuming to consider all words in a text, but rather several key tokens. In this work, the authors proposed an effective method to classify Vietnamese texts leveraging the TextRank algorithm and Jaccard similarity coefficient. TextRank ranks words and sentences according to their contribution value and extracts the most representative keywords. First, we collected textual sources from a wide range of Vietnamese news websites. We then applied data preprocessing, extracted keywords by TextRank algorithm, measured similarity score by Jaccard distance and predicted categories. The authors have conducted numerous experiments, and the proposed method has achieved an accuracy of 90.07% on real-world datasets. We have proved that it is entirely applicable in practice.