Affective Lexicon Creation for the Greek Language

Affective Lexicon Creation for the Greek Language Elisavet Palogiannidi1,2, Polychronis Koutsakis4, Elias Iosif 2,3, Alexandros Potamianos2,3 1School of ECE, Technical University of Crete, Chania 73100, Greece 2“Athena” Research and Innovation Center, Maroussi 15125, Athens, Greece 3School of ECE, National Technical University of Athens, Zografou 15780, Greece 4School of Engineering and Information Technology, Murdoch University, Australia epalogiannidi@isc.tuc.gr, p.koutsakis@murdoch.edu.au, {iosife,potam}@central.ntua.gr Abstract Starting from the English affective lexicon ANEW (Bradley and Lang, 1999a) we have created the first Greek affective lexicon. It contains human ratings for the three continuous affective dimensions of valence, arousal and dominance for 1034 words. The Greek affective lexicon is compared with affective lexica in English, Spanish and Portuguese. The lexicon is automatically expanded by selecting a small number of manually annotated words to bootstrap the process of estimating affective ratings of unknown words. We experimented with the parameters of the semantic-affective model in order to investigate their impact to its performance, which reaches 85% binary classification accuracy (positive vs. negative ratings). We share the Greek affective lexicon that consists of 1034 words and the automatically expanded Greek affective lexicon that contains 407K words. Keywords: affective lexicon, affective ratings, valence, arousal, dominance, semantic similarity, sentiment analysis, emotion recognition 1. Introduction Emotions are an integral part of human cognition, helping humans make decisions even in cases when uncertainty and ambiguity are introduced by cognitive analysis (Bechara et al., 2000). Hence, emotion detection attracts the interest of researchers from numerous areas such as psychology, lin- guistics, and engineering. In order to model emotion as it is experienced by hu- mans, resources that contain emotion information in var- ious modalities become a necessity. (Lang et al., 1999) created a picture stimuli resource, known as International Affective Picture System (IAPS). Each stimulus is anno- tated on the affective dimensions Valence, Arousal and Dominance (V,A,D). The International Affective Digitized Sounds (IADS) is the corresponding resource with sound stimuli that was created by (Bradley and Lang, 1999b). (Bradley and Lang, 1999a) in their attempt to provide a set of normative emotional ratings for English words. They created an affective lexicon for English, a.k.a. ANEW (Af- fective Norms for English Words). ANEW contains 1034 words and each word is assigned with one score per af- fective dimension (V,A,D). Since then, numerous affective lexica in many languages have been created. (Redondo et al., 2007) translated the ANEW into Spanish and (Soares et al., 2012) into European Portuguese. (Montefinese et al., 2014) translated the ANEW into Italian and augmented it with more words. Additionally to ANEW, affective lexica with less or more entries can be found for a number of other languages such as German (Kanske and Kotz, 2010) and Dutch (Moors et al., 2013). Bilingual approaches have been reported as well, e.g., (Eilola and Havelka, 2010) created affective norms from English and Finish nouns. (Gilet et al., 2012) provided an affective lexicon of French attributes and investigated the age influence on the ratings collection process. The words of ANEW were also annotated with respect to discrete (categorical) …

Content truncated to 500 words