Context Aware Sentiment Analysis: A Hybrid Framework for Detecting and Analyzing Sentiments based on Entities

Entity-based sentiment analysis is one of the critical tasks in understanding public opinion, especially in the social media platform of Twitter. Most of the current systems focus either on sentence-level sentiment detection or entity recognition but do not integrate both effectively. In this research, we propose a hybrid approach that combines custom-trained spaCy for entity detection with a hybrid of BERT and BiLSTM models for sentiment analysis. It identifies the entities and their sentiments in tweets, thereby providing granular insights into public opinion. Experimental evaluation shows the model's capability to handle informal and context-rich Twitter language and, more importantly, that it achieves accurate sentiment detection for both entities and overall tweets. This approach outperforms traditional methods and has potential applications in customer feedback analysis and trend monitoring.