In this paper, we present an approach for utilising emotions and sentiments for enhancing the context handling capability of Natural Language Processing algorithms. We base our work on the premise that emotions of a person get reflected in the textual content generated by the individual. We thus conceptualise that by making use of existing tools for sentiment analysis and emotion extraction, we can improve capability of machines to understand humans better. We study the effect of sentiment patterns within an utterance and associated emotions on the perceived helpfulness of Amazon reviews. We propose an approach involving use of emotional information as an incremental add-on to transformer architecture for enhanced application-specific textual analytics. In this work we have experimented with various options of enriching BERT embedding with emotional information present in Amazon product reviews. Our results indicate that adding emotional patterns to BERT embedding yields improved performance in predicting helpfulness of a review, thus demonstrating that emotions within text can be utilised in a deliberate and scientific manner for better understanding by machines. The emotion augmented embedding yielded 1.85% better accuracy, 0.5% better precision, 6% better Recall and 3% better F1 Score than original BERT embedding. Our approach does not need fine tuning of BERT which is resource intensive, and can be generalised to an application specific use case.