Cross-lingual sentiment analysis for Indian regional languages

Abstract

The process of identifying the opinion and categorization of a given dataset to determine whether the attitude of a writer towards the given data is positive, negative or neutral is called as Sentiment Analysis. Sentiment Analysis involves computational identification of opinion on given dataset. It is also referred as the extraction of the opinion. Cross lingual Sentiment Analysis refers to the generation of the opinions in two languages. One language is highly rich in its resources providing sufficient dataset required for the opinion extraction known as Resource Rich Language. The other languages such as Kannada, Hindi, Marati which are poor in its resources and lacks in the data Wordnet and seeks the help of resource rich languages for the opinion extraction known as Resource Poor Language. Sentiment Analysis may be the opinions of movie reviews and social media responses. Here we have used architecture of auto encoder which helps in the generation of the sentiment analysis in two languages. Sentiment Analysis of two languages can be performed by using the Bilingually Constrained Recursive Auto-encoder (BRAE) model and also with the help of linked Wordnet datasets.