End-to-end Aspect Based Sentiment Analysis (E2E-ABSA) focuses on extracting aspects and corresponding sentiments simultaneously. The majority of previous works formulated it as a sequence labeling problem and applied different tagging schemes for the problem. IN this worWe employed a span based strategy for the task rather than a sequence tagging approach. Span based approach identifies the aspect span which is further classified into predefined classes. Our model extends the model proposed in [8] by incorporating linguistic information with a pre-trained bidirectional transformer language model BERT. The suggested approach seeks to determine the span, i.e. start location and end location of the target aspect, which is then utilized to predict the orientation of the span representing the aspect. To strengthen the tagging scheme with the linguistic information, we have combined part of speech (POS) information while learning the contextual representation of words. In this work, we have used BERT to obtain a contextual representation corresponding to the tokens of a sentence. The outcomes of experiments using Semeval-14 datasets are encouraging for our methodology.