Sentiment analysis is a fundamental task of natural language processing which involves assessing the emotional tone and intensity expressed in textual data. Sentiment strength detection is an important aspect of sentiment analysis, which involves the measurement of sentiment strength, which enables a deep understanding of sentiment expression. Sentiment strength detection has extensive applications in many domains, including marketing, social media monitoring, and customer review analysis. However, detecting sentiment strength in the text is a challenging task. In this regard, researchers proposed lexicon-based methods and supervised machine learning-based approaches, but they were unable to achieve considerable performance as they had their drawbacks to address. Thereafter, to resolve the issue, long short-term memory-based deep learning methods came into play. In this work, we employ bidirectional long short-term memory (BiLSTM) along with an attention mechanism by incorporating multi-channel CNN to effectively address the challenges of sentiment strength detection. We performed a comparative analysis with Bidirectional Encoder Representations from Transformers (BERT) and other models using different word embeddings. The proposed method for sentiment strength detection is assessed with mean absolute error(MAE) and predictive \(R^{2} (pR^{2})\) on the two publicly available datasets to show its effectiveness.