Approximate Entropy in Canonical and Non-Canonical Fiction

Abstract

Computational textual aesthetics aims at studying observable differences between aesthetic categories of text. We use Approximate Entropy to measure the (un)predictability in two aesthetic text categories, i.e., canonical fiction (`classics’) and non-canonical fiction (with lower prestige). Approximate Entropy is determined for series derived from sentence-length values and the distribution of part-of-speech-tags in windows of texts. For comparison, we also include a sample of non-fictional texts. Moreover, we use Shannon Entropy to estimate degrees of (un)predictability due to frequency distributions in the entire text. Our results show that the Approximate Entropy values can better differentiate canonical from non-canonical texts compared with Shannon Entropy, which is not true for the classification of fictional vs. expository prose. Canonical and non-canonical texts thus differ in sequential structure, while inter-genre differences are a matter of the overall distribution of local frequencies. We conclude that canonical fictional texts exhibit a higher degree of (sequential) unpredictability compared with non-canonical texts, corresponding to the popular assumption that they are more `demanding’ and `richer’. In using Approximate Entropy, we propose a new method for text classification in the context of computational textual aesthetics.

Comparative Analysis of Preference in Contemporary and Earlier Texts Using Entropy Measures

“…Is it promotion, network effects, economic or social circumstances—or perhaps the “quality” of the text itself? These questions have recently been addressed in a variety of studies in the field of computational aesthetics, aiming to identify observable correlates of preference in the structure of a text [ 1 , 2 , 3 , 4 , 5 , 6 , 7 ]. In empirical aesthetics the term “preference” is used to capture aesthetic attitudes towards cultural artefacts [ 8 ].…”

“…We deal with texts from two time periods: with the contemporary period, spanning the time between 2000 and 2020, and from an earlier period, covering the time between 1813 and 1922. For each period we (necessarily) use different operationalizations of preference: The earlier texts are divided into canonical and non-canonical texts, using the canon of Western literature [ 13 , 14 ] as a criterion of classification (see also [ 7 , 15 ]). Canonical texts form part of the cultural backbone and historical memory of a society [ 16 ].…”

“…The second major challenge of our work concerns the type of statistic that may be informative with respect to the degree of preference. Previous studies, inspired by research in the domain of vision (for a review, see [ 12 ]), have focused on global statistical properties such as the variability (of observables) in a text [ 15 , 27 , 28 ], long-range correlations [ 6 , 15 , 29 ] and various indicators of predictability or surprise [ 4 , 7 ]. In the present study we use two measures of surprise—Shannon Entropy and Approximate Entropy—as the aesthetic experience has been shown in previous work to be driven by the interplay between expectation and surprise [ 7 , 30 , 31 ] (moreover, see [ 32 , 33 ] for a discussion on “unification” and “diversification” in text).…”

“…As reviewed in more detail in [ 7 ], preference has been operationalized in terms of the scope of the readership in previous studies under different terms, such as “success” [ 27 ], “popularity” [ 34 ], being “professional” vs. “amateur” [ 1 ], or “information-based energy” [ 35 ]. Data from websites and social networks have been used in some previous studies to model readers’ preference, for instance, the download counts from the website of the Gutenberg Project [ 2 ], or ratings of readers on the website Goodreads [ 3 , 5 ].…”

“…In the present study we adopt the approach to textual aesthetics proposed by Mohseni et al [ 7 , 15 ]. We assume that a pleasant reading experience emerges from an interplay of predictability and surprise.…”