Information extraction is a vital subtask of knowledge graph constructions, where the joint extraction of entities and relations in the form of triples is an essential component. Given the complexity of capturing the correlation between entities and relations, existing methods try to segment the joint extraction task into multiple modules for simplicity and efficiency. However, these methods neglect the entirety of the triple structure, which causes cascading errors and redundancy among the modules. Furthermore, they fail to exploit the rich semantic information in relation labels as relation prior knowledge, which can provide semantic details and guidance to help models better understand entity relations, classify relation labels, and enhance the accuracy of joint entity and relation extraction tasks. In this paper, we propose a single-module joint extraction approach with attention enhanced text embedding, named AttRel. Specifically, a novel single module for triple classification is designed in AttRel to avoid the cascading errors and redundancy, and meanwhile, information of relation labels as an informative source of prior knowledge is incorporated into text embedding vectors to improve the accuracy of extracting triples. Extensive experiments conducted on benchmarking datasets demonstrate that our proposed approach AttRel outperforms the state-of-the-art baselines with higher accuracy.