Towards the Definition of Linguistic Metrics for Evaluating Text Readability

Abstract

The frequency of linguistic units and patterns (linguistic metrics) is specific to each language and varies depending on text complexity. In this study we investigate a number of measures that can contribute to determine text complexity/difficulty and be obtained automatically. The variables considered come primarily from counts of phonological units and patterns (e.g. word size, number of prosodic words and clitics, rare syllable types, exceptional stress patterns and rare segment classes), some of which being also informative of texts’ morphological and syntactic complexity and semantic density. We compared the linguistic metrics in a large sample of Portuguese medicine package leaflets (PL), a sample of PL from the medicines that belong to the three most consumed therapeutic groups, and two samples of more common texts (composed of journalistic and oral texts). For each indicator, oral texts, in particular texts produced by speakers with lower levels of education, showed the lowest values in a difficulty continuum, and frequently PL showed the highest values in that continuum, indicating that PL may be less readable than the other types of texts analysed.