Towards Cross-Language Sentiment Analysis through Universal Star Ratings

The abundance of sentiment-carrying user-generated content renders automated cross-language information monitoring tools crucial for today's businesses. In order to facilitate cross-language sentiment analysis, we propose to compare the sentiment conveyed by unstructured text across languages through universal star ratings for intended sentiment. We demonstrate that the way natural language reveals people's intended sentiment differs across languages. The results of our experiments with respect to modeling this relation for both Dutch and English by means of a monotone increasing step function mainly suggest that language-specific sentiment scores can separate universal classes of intended sentiment from one another to a limited extent