Publication

Abstract : Stance detection is one step ahead of sentiment analysis where author’s stance for certain topics such as an event, personality, or a government policy is considered. The author’s stance could be in “favor” or “against” the topic under consideration. A myriad amount of data is being accumulated via various social media platforms. This work considers the Kannada–English code-mixed aspect of social media text. The corpus was collected based on various trending topics such as “Bengaluru molestation,” “currency ban”, etc. using particular word phrases. The user comments on social media platform Facebook was used to collect the corpus. The collected dataset was represented using different techniques such as bag of tricks, word embedding like Word2vec and GloVe, and pre-trained embeddings. These representations were further used in combination with various deep learning architectures such as convolutional neural network (CNN), bidirectional long short-term memory (Bi-LSTM). The results for various combinations are listed. © 2019, Springer Nature Singapore Pte Ltd.