Text characterization based on recurrence networks

Several complex systems are characterized by presenting intricate characteristics taking place at several scales of time and space. These multiscale characterizations used in various applications, including better understanding diseases, characterizing transportation systems, comparison between cities, among others. In particular, texts also a hierarchical structure that can be approached using multi-scale concepts methods. The properties constitute subject worth further investigation. addition, more effective approaches to text characterization analysis obtained emphasizing words with potentially informational content. present work aims developing these possibilities while focusing on mesoscopic representations networks. More specifically, we adopt an extension the approach represent narratives, which only recurrent relationships tagged parts speech (subject, verb direct object) considered establish connections sequential pieces (e.g., paragraphs). was then achieved considering scale-dependent complementary methods: accessibility, symmetry recurrence signatures. order evaluate potential methods, problem distinguishing literary genres (fiction non-fiction). A set 300 books organized into two were compared aforementioned approaches. All methods capable differentiating some extent genres. accessibility reflected narrative asymmetries, signature provided indication about non-sequential semantic along narrative.