Developing Bengali WordNet Affect for Analyzing Emotion

This paper reports the process of developing WordNet Affect lists in Bengali from the affect wordlists already available in English. It is organized in six basic emotions such as anger, disgust, fear, joy, sadness and surprise. Emotion or affect words of the six lists are updated using equivalent synsets of the SentiWordNet by keeping their parts-of-speech information unchanged. Another updating process employed especially for the verb entries is carried out on six affect lists with the help of VerbNet. The updated lists are converted into Bengali using the synset based English to Bengali bilingual dictionary. The sense disambiguation task is conducted based on the hints of sense wise separated word groups present in Bengali to English bilingual dictionary. Human translator translates the non-translated emotion word entries into Bengali. The statistical inter-translator agreement between two translators is measured using kappa coefficient (k) that shows a moderate agreement varying from 0.44 to 0.56 for six emotion classes. KeywordsWordNet Affect; SentiWordNet; VerbNet; Bilingual Dictionary; Translation; Sense Disambiguation; Kappa