The Validation of MRCPD Cross-language Expansions on Imageability Ratings

The Validation of MRCPD Cross-language Expansions on Imageability Ratings Ting Liu1, Kit Cho2, Samira Shaikh1, Mehrdad Mirzaei1, Tomek Strzalkowski1 1 University at Albany, State University of New York 2 University of Houston-Downtown E-mail: tliu@albany.edu, kitwcho@gmail.com Abstract In this article, we present a method to validate a multi-lingual (English, Spanish, Russian, and Farsi) corpus on imageability ratings automatically expanded from MRCPD (Liu et al., 2014). We employed the corpus (Brysbaert et al., 2014) on concreteness ratings for our English MRCPD+ validation because of lacking human assessed imageability ratings and high correlation between concreteness ratings and imageability ratings (e.g. r = .83). For the same reason, we built a small corpus with human imageability assessment for the other language corpus validation. The results show that the automatically expanded imageability ratings are highly correlated with human assessment in all four languages, which demonstrate our automatic expansion method is valid and robust. We!believe these new resources can be of significant interest to the research community, particularly in natural language processing and computational sociolinguistics. Keywords: MRC Psycholinguistic Database (MRCPD), wordnet, imageability, validation 1. Introduction Scientists engaged in research using lexical stimuli must take into account a multitude of variables associated with words (e.g. familiarity, frequency of occurrence) that could potentially obscure the interpretation of their results. The Psycholinguistic Database (MRCPD) (Wilson, 1988) provides 26 linguistic and psycholinguistic variables for over 150 thousand English words and has served as a valuable resource for researchers in a variety of disciplines such as linguistics, psychology (Coltheart, 1981), and computational linguistics (Broadwell et al., 2013, Strzalkowski et.al. 2013). However, not all words have ratings for all 26 variables. For example, only 9,240 (6%) of the total words in MRCPD have an imageability value (i.e., how easily and quickly the word evokes a mental image). The traditional way of analyzing the variables requires human analysis and judgments, which is accurate but time consuming and expensive (Higgins, Xi, Zechner, & Williamson, 2011, Brysbaert et al. 2014). Therefore, in Liu et al. (2014), we presented an approach to automatically expand the MRCPD database by adding imageability ratings for an additional 106,911 words and subsequently translating the expanded MRC into Spanish, Russian, and Farsi. Using an automated approach has the advantage of gathering data quickly and with fewer resources. However, as with any other method, validation of the method is required. In the present study, we report two independent validation techniques and 1. assessed the reliability and validity of the expanded results, and 2. tested whether the imageability ratings of English words are retained when they are translated into their foreign-language equivalents. The present study explored the validation by comparing the imageability ratings for English words to their Spanish-, Russian-, and Farsi-translation equivalents. To our knowledge, there are only two corpora of imageability ratings for non-English words (Italian – Della Rosa et al., 2010; Spanish - Sebastián Gallés, 2000) and those corpora are quite limited (417 words for Italian corpus and 6K for Spanish corpus). One reason why imageability ratings are not readily …

Content truncated to 500 words