Code Mixed Question Answering Challenge using Deep Learning Methods

In social media, code-mixed language questions(combination of two distinct languages) is turning into the favored method of expression and communication. In twitter people can go over mixed language where individuals frequently utilize their mother tongue along with English. But question answering systems do not support mixed language. They only work on a single language such as English, French and German. This paper presents an online question answering framework for mixed languages. This system analyzes the user's mixed language query and gives the answer and also presents a contrivance that facilitates decode of English mixed with 3 different Indian languages: Hindi, Telugu and Tamil. The code mixed words will be translated into English to reduce data complexity. Later, featurization is performed on them. Deep learning algorithms such as Recurrent Neural Network (RNN), Hierarchical Attention Network (HAN) are used for question classification. Confusion matrix is used as evaluation measure of RNN, HAN. This system is facilitated on the web and in future, it can be used for collecting more code mixed questions and answers data for further improvement.