Device-Dependent Readability for Improved Text Understanding

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1396–1404, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Device-Dependent Readability for Improved Text Understanding A-Yeong Kim Hyun-Je Song Seong-Bae Park Sang-Jo Lee School of Computer Science and Engineering Kyungpook National University Daegu, 702-701, Korea {aykim,hjsong,sbpark}@sejong.knu.ac.kr, sjlee@knu.ac.kr Abstract Readability is used to provide users with high- quality service in text recommendation or text visualization. With the increasing use of hand- held devices, reading device is regarded as an important factor for readability. There- fore, this paper investigates the relationship between readability and reading devices such as a smart phone, a tablet, and paper. We sug- gest readability factors that are strongly related with the readability of a specific device by showing the correlations between various fac- tors in each device and human-rated readabil- ity. Our experimental results show that each device has its own readability characteristics, and thus different weights should be imposed on readability factors according to the device type. In order to prove the usefulness of the results, we apply the device-dependent read- ability to news article recommendation. 1 Introduction Readability is a function that maps a given text into a readability score by considering “how easily the text is read and understood” (Richards et al., 1992; Zamanian and Heydari, 2012). Normally, the readability score is formulated as a combination of various factors. These factors reflect the easiness and understanding of the text and include text presentation format, font size, av- erage ratio of annotated images, and sentence length (Hasegawa et al., 2008; Kitson, 1927; Ma et al., 2012; ¨Oquist, 2006). Therefore, readability can be used to provide satisfiable services in text recommendation or text visualization. The study on readability has begun in the education field to measure the level of a text. With the success of using readability in education (Franc¸ois and Fairon, 2012; Heilman et al., 2008; Ma et al., 2012), read- ability has been used in a range of domains recently. For example, in document retrieval, readability is used to provide documents to non-expert users so that they can read the retrieved documents easily (Jameel et al., 2012; Yan et al., 2006). In text mining, readability has been employed to analyze the characteristics of text. Especially, Hillbom showed the differences in readabil- ity between broadsheet newspapers and tabloids that share a similar political stance (Hillbom, 2009). There is one important issue of readability that has not been studied in natural language processing. It is a reading device. That is, previous studies focused only on text printed on paper. However, with the increasing use of hand-held devices, people in these days use var- ious reading devices such as a tablet and a smart phone as well as a paper. Readability score can be different according to the device type, because each device has its own idiosyncrasy. For example, assume that a sys- tem recommends the same news article to both user A who reads it in her smart phone and …

Content truncated to 500 words