Classifying Written Texts Through Rhythmic Features

Abstract

Abstract. Rhythm analysis of written texts focuses on literary analysis and it mainly considers poetry. In this paper we investigate the relevance of rhythmic features for categorizing texts in prosaic form pertaining to different genres. Our contribution is threefold. First, we define a set of rhythmic features for written texts. Second, we extract these features from three corpora, of speeches, essays, and newspaper articles. Third, we perform feature selection by means of statistical analyses, and determine a subset of features which efficiently discriminates between the three genres. We find that using as little as eight rhythmic features, documents can be adequately assigned to a given genre with an accuracy of around 80 %, significantly higher than the 33 % baseline which results from random assignment.

Grammar Rhythm Figures In Literary Prose: Case Study Of S.Thomas’S Popco

“…Beeferman (1996) measures the regularity of lexical stress in English texts and shows a relationship between the probability of word sequences and the amount of rhythm present in them. Balint et al (2016) consider rhythmic features of written texts (speeches, essays and newspaper articles) and statistically determine sets of features for each genre. Golubeva-Monatkina (2017) studies the rhythm in fictional prose and the issues of rhythm translation into another language.…”

The rhetorical nature of rhythm

“…In our previous work [1,9], we adopted a broad view of rhythm that can result from a particular arrangement and seriation of any linguistic items, such as punctuation, phonemes, stresses, syllables, words, parts of speech, n-grams, syntactic structure, or lengths of rhythmic units. We consider that rhythm is achieved through repetition, alternation, or progressive/regressive sequencing of linguistic items, and the model is open to the addition of other devices for rhythm production.…”

“…The studied documents are of significantly different lengths, but they provide a balanced input to the subsequent statistical analyses, in the form of 330 observations for each feature (110 corresponding to each genre), with the feature values already normalized by text length. In previous work [1], we grouped our features into five classes, based on the type of linguistic item that was used as a building block of rhythm: organizational, lexical, grammatical, phonetical, and metrical. For the purposes of this study we introduce a new classification, into (1) generic features, whose values do not depend on a particular segmentation into rhythmic units, and (2) unit-based features, which describe individual properties and interactions of rhythmic units.…”

“…Then, we removed all features which demonstrated non-normality (log values were introduced so that more features would survive this step). Second, we removed the features that did not pass Levene's test for the equality of variances, a step that was not undertaken in our previous work [1] and that would have reduced the accuracy of that model from 81.51% to 68.20%. Third, we assessed multicollinearity based on pair-wise correlations with a correlation coefficient r > .70, and, from each pair of multicollinear features, removed the feature with a weaker effect.…”

“…to a genre), sentiment analysis, and ultimately natural language understanding. In this paper we present a refined and extended set of rhythmic features compared to the set we proposed in previous work [1], which improves the accuracy of the text categorization experiment presented in [1] by 5% (15% if stricter statistical criteria are enforced). The main addition to the model is the usage of elementary discourse units (EDUs), obtained through segmentation in the framework of Rhetorical Structure Theory (RST) [2], as rhythmic units.…”