Bilingual Sentiment Analysis for a Code-mixed Punjabi English Social Media Text

Abstract

India is a multilingual country and a large portion of its population speaks more than one language. It has been observed that such multilingual speakers switch between the languages while communicating informally. Such code-mixed sentences not only contain words from different languages but also different grammar rules in a single sentence. In this work, we have focused on one such coded-mixed language pair i.e. Punjabi-English. In informal communication and social media, the code-mixed language is very popular, and analyzing the sentiment of such mixed sentences is a challenging task. We have verified that the language models and sentiment analysis methods designed for a single language don't work well in such sentences. And by reviewing the drawbacks, we propose our novel model which is based on LSTM. The proposed model provides quite satisfactory results for the code-mixed data set.