Sentiment Analysis is a channel by which automated feedback analysis can be processed effectively and efficiently. The reviews reachable are not only useful for customers buying a product or service but also the manufacturers and industrialists to formulate their production and marketing strategies as well as government organizations to access the views of the citizens. Recommender systems, market researches, and predictions on various social media platforms can be made more practical and comprehensive by sentiment analysis. It has evolved in numerous ways starting from the coarsely grained analysis. But when we need to find intricate details scrunched into a single statement or so, aspect-level sentiment analysis is the way. The uprising of deep learning approaches has opened doors in various applications, including Aspect-based Sentiment Analysis. These networks are computationally strong and can be trained easily. However, since they lag in recognizing semantic complexities, various Natural Language Processing techniques have been joined into the neural models. Researchers have been creative with inventing various novel models, combining a myriad of neural networks and attention mechanisms to improve aspect detection and sentiment polarity identification. The promise that this field provides because of its independent nature compels scientists to delve into the topic. In this work, fine grained analysis is not only processed for the aspect and aspect word detection but also for polarity and its intensity analysis. Our Contributions include the enriched input embedding with token, orientation, grammatical function, field and intensity components in the embedding stage, refined pattern extraction with convolutional kernels and improved performance using attention mechanism in the latter stages. Our experimental results reveal that, our procedural methodology has brought out an optimal enhanced performance compared to the near closer designs.