LT3 at SemEval-2020 Task 9: Cross-lingual Embeddings for Sentiment Analysis of Hinglish Social Media Text

This paper describes our contribution to the SemEval-2020 Task 9 on Sentiment\nAnalysis for Code-mixed Social Media Text. We investigated two approaches to\nsolve the task of Hinglish sentiment analysis. The first approach uses\ncross-lingual embeddings resulting from projecting Hinglish and pre-trained\nEnglish FastText word embeddings in the same space. The second approach\nincorporates pre-trained English embeddings that are incrementally retrained\nwith a set of Hinglish tweets. The results show that the second approach\nperforms best, with an F1-score of 70.52% on the held-out test data.\n