Segmentation of Weather News Headlines: Sentiment and Tone Shifts in Parallel Bilingual Corpora

Abstract

This paper explores whether emotional tone/sentiment of the text is completely/partially lost, shifted or preserved in parallel bilingual corpora of weather news headlines and to what extent the translation techniques chosen influence this process. To this aim we have carried out both sentiment and tone analysis of weather news headlines, compared and critically evaluated the obtained results. For investigating the influence of translation techniques used on the sentiment/tone of the expression two programs were applied: The IBM Watson™ Tone Analyzer and MonkeyLearn. The first program is used for defining the tone of the text (in our work ˗ the headline), and the second points out its sentiment.