Sentiment Analysis of Marathi–English Code-Mixed Using Ensemble Model

Many languages exist worldwide with different types, identifications, and pronunciations that have been used by different types of people with different age groups for their communication purposes. Many times, peoples use different language to express their views using different medium of language, e.g., one can “kya” using English characters for expressing a word “why”; here both have same meaning written in English language, but spellings are different. This type of conversion or expressing of words is known as “transliteration processing” which means converting a specific word from one language to another using same communication language medium. Nowadays, this transliteration is most widely used by every age group of people in various communication platforms like WhatsApp, Facebook, Twitter, or even in short messaging service (SMS) in mobile application, etc. Transliteration is also called as code-mixed language and is very difficult to identify the language used in each word of code-mixed language text script. As it is the most widely used communication medium by different people using their regional language script, it has some informal and mostly unstructured data in the text written for sharing. Detecting the sentiments from this type of unstructured language is a very difficult task; hence, these challenges attract various researchers. In our research work, we are proposing a model based on deep learning model with n-gram multinomial Naïve Bayes logic for identification of sentiment in Marathi–English language script mixed-code transliteration.