CODE MIXED SENTIMENT ANALYSIS

Abstract

In the areas of social media analytics, Natural Language Processing (NLP) and web mining, Sentiment Analysis has been the most important research area. Code switching or Code mixing is when people change language while conversing this mainly happens where multilingual people reside like in India . In such countries the sentiment analysis becomes difficult as compared to places with a single language as they may differ in semantics and even symbols as in case of Hinglish. Moreover people in different regions have different dialects which is also evident in the way they write hinglish sentences. Therefore there is a huge vacuum for code switched sentiment analysis technology. There are different ways to find the sentiments in code switched data. Some of them include word embedding, classical ML methods and more advanced models like cnn, rnn, bert, ensemble models etc. Social media texts including chats (eg., Facebook and whatsapp messages), blogs and microblogs (eg., twitter) have been constantly evolving. This has led to the creation of new opportunities for language technologies and information access. With this new opportunity comes new challenges making Sentiment Analysis one prime research area in Natural Language Processing (NLP). The primary focus of Current language technologies has been on English. There has been a demand for methods that can also process other languages as Social media is inherently a multilingual environment. People in multilingual environments like India, employ alternating languages in the same utterance to express their views on social media. This phenomenon can be seen around the world. Code-mixing or code-switching is the mixing of languages; one such language is Hinglish. Code- mixing has become a norm in multilingual societies and an important NLP challenge. Moreover this is just the beginning to many more such applications and unimaginably large data which needs to be used so brands and companies can evaluate what people think of their product. The data has increased exponentially and there is a vacuum to use this data. Hence code switched sentiment analysis.