Currently, there is a large amount of text being shared through the Internet. These texts are available in different forms—structured, unstructured and semi structured. There are different ways of analyzing texts, but domain experts usually divide this process in some steps such as pre-processing, feature extraction and a final step that could be classification, clustering, summarization, and keyword extraction, depending on the purpose over the text. For this processing, several approaches have been proposed in the literature based on variations of methods like artificial neural network and deep learning. In this paper, we conducted a systematic review of papers dealing with the use of complex networks approaches for the process of analyzing text. The main results showed that complex network topological properties, measures and modeling can capture and identify text structures concerning different purposes such as text analysis, classification, topic and keyword extraction, and summarization. We conclude that complex network topological properties provide promising strategies with respect of processing texts, considering their different aspects and structures.

中文翻译：

---

基于复杂网络的文本结构方法：系统综述

目前，有大量文本通过互联网共享。这些文本有不同的形式——结构化、非结构化和半结构化。分析文本的方法有多种，但领域专家通常将这个过程分为一些步骤，例如预处理、特征提取和最后一步可以是分类、聚类、摘要和关键字提取，具体取决于文本的目的. 对于这种处理，文献中已经提出了几种基于人工神经网络和深度学习等方法的变体的方法。在本文中，我们对处理复杂网络方法在文本分析过程中使用的论文进行了系统回顾。主要结果表明，复杂网络拓扑性质，度量和建模可以捕获和识别涉及不同目的的文本结构，例如文本分析、分类、主题和关键字提取以及摘要。我们得出结论，考虑到文本的不同方面和结构，复杂网络拓扑属性为处理文本提供了有前景的策略。