Indices of textual cohesion by lexical repetition based on semantic networks of cliques

Abstract

Cohesion is one of the main defining characteristics of textuality. There is a lack of approaches to textual cohesion benefiting from both high computational implementability, and compatibility with established theories of text organization and comprehension. The main goal of this study was to define indices of textual cohesion based on lexical repetition and semantic networks of cliques that can measure cohesion while being compatible with the aforementioned theories. As specific goals, the study aimed to develop an improved taxonomy for lexical repetition and to analyze the behavior of the proposed indices with a sample of texts from different genres. Six indices were proposed, and 60 texts from six different genres were analyzed. The results showed that the indices can capture different patterns of lexical cohesion that can be explored, for example, in classification tasks applied to texts, in summarization tasks, and in the adaptation of reading material to improve the readability and comprehensibility of texts by less proficient readers. Additionally, the patterns observed within the analysis of full texts were also observed in the analysis of the first 10 to 60 sentences of texts, depending on the index considered, suggesting that the first 10 to 60 sentences of a text are enough to analyze its lexical cohesion