A Intensified Approach on Aspect Based Sentiment Analysis Using Pre-Trained Deep Learning Model

Sentiment analysis often known as opinion mining, is a technique for determining people's thoughts on a certain product, service, or policy. To better understand the emotion, categorize it according to each component of the document's entity. Even aspect-based sentiment analysis (ABSA) seeks to do this. Previous ABSA research has primarily used models that execute a single task on datasets containing data from a single domain. This creates issues when using domain-independent datasets with ABSA to fulfill the entire task list. In this study, we analyze sentiment using a BERT-LSTM (Bidirectional Encoder Representations with Long Short-Term Memory) model. The CNN layer performs convolutional operations after processing the word embedding and BERT encoder. To extract local textual properties, the model feeds text into a powerful layer of Con-volutional Neural Networks (CNNs). The LSTM layer, which models time series to identify long-term relationships within the text, receives features from the CNN layer. The consequences of the studies revealed that the model described in this study excelled with CNN s in sentiment analysis. This work makes use of the powerful deep learning model BERT Base Uncased, allowing for the experimental consequences of the recommended BERT-LSTM model. The results show that the proposed model outperforms with an accuracy of 95.91 % compared to other deep learning models. Visualization experiments further confirm the rationality of the BERT- LSTM technique.