Previous studies show that readability assessment can be useful for text simplification, whereas we argue that text simplification can also play an important role in the evaluation of text readability. In this work, we propose a text graph network to learn the simplified relation between documents. Specifically, we first establish the parallel simplification corpora by preprocessing. Then, we can obtain the simplified relation between documents contained in parallel simplification corpora. We explore our idea by adding edges between documents (nodes) with simplified relation. The extensive experimental results on public datasets demonstrate the effectiveness of our method.