Measuring Text Readability by Lexical Relations Retrieved from Wordnet

Measuring Text Readability by Lexical Relations Retrieved from WordNet Shu-yen Lin, Cheng-chao Su, Yu-da Lai, Li-chin Yang, Shu-kai Hsieh English Department National Taiwan Normal University sylin@ntnu.edu.tw Abstract Current readability formulae have often been criticized for being unstable or not valid. They are mostly computed in regression analysis based on intuitively-chosen variables and graded readings. This study explores the relation between text readability and the conceptual categories proposed in Prototype Theory. These categories form a hierarchy: Basic level words like guitar represent the objects humans interact with most readily. They are acquired by children earlier than their superordinate words (or hypernyms) like stringed instrument and their subordinate words (or hyponyms) like acoustic guitar. Therefore, the readability of a text is presumably associated with the ratio of basic level words it contains. WordNet, a network of meaningfully related words, provides the best online open source database for studying such lexical relations. Our preliminary studies show that a basic level word can be identified by its frequency to form compounds (e.g. chair � armchair) and the length difference from its hyponyms in average. We compared selected high school English textbook readings in terms of their basic level word ratios and their values calculated in several readability formulae. Basic level word ratios turned out to be the only one positively correlated with the text levels. Keywords: Readability, Ontology, Prototype Theory, WordNet, Basic Level Word 1. Introduction Reading process is the core of language education. Teachers now have access to a vast amount of texts extractable from the Internet inter alia, but the materials thus found are rarely classified according to comprehension difficulty. It is not uncommon to see foreign language teachers using texts not compatible with the students’ reading abilities. Traditional methods of measuring text readability typically rely on the counting of sentences, words, syllables, or characters. However, these formulae have been criticized for being unstable and incapable of providing deeper information about the text. Recently, the focus of readability formula formation has shifted to the search for meaningful predictors and stronger association between the variables and the comprehension difficulty. We start our research by assuming in line with Rosch et al.’s Prototype Theory [1] that words form conceptual hierarchies in that words at different hierarchical levels pose different processing difficulties. This processing difficulty is presumably correlated with the reading difficulty of the text containing the words. Putting the logic into templates, the measurement of text readability can be done by calculating the average hierarchical levels at which the words of a text fall. Our study comprises two stages. In the preliminary experiments, we utilized WordNet [2], an online lexical database of English, to identify basic level words. In the subsequent experiment, we compared selected readings in terms of their basic level word ratios and their values calculated in several readability formulae. Basic level word ratios turned out to be the only one positively correlated with the text levels. The remainder of this paper is organized as follows: Section 2 reviews the common indices …

Content truncated to 500 words