Graphs offer a natural representation of text structures, capturing connections between words in a sentence or sentences in a paragraph. Exploiting these structures with deep learning models however can be challenging, since their design generally only allows processing data with Euclidean structure. In this talk, I will introduce graph neural networks (GNNs), as a powerful method to address this technical challenge. Later, I will discuss the benefits of this approach in tackling tasks such as relation extraction, link prediction, and knowledge graph construction, in the context of supply chain networks.