Publication

Abstract : Sentiment analysis in colloquial data has distinct issues because of the informal and context-sensitive character of the language employed, especially in messaging services such as WhatsApp. In order to analyze sentiment in WhatsApp chats, this study suggests an advanced deep learning architecture that combines encoder-decoder models and bidirectional long short-term memory (Bi-LSTM) networks with attention mechanism. By using Bi-LSTM networks, the suggested method takes into account the sequential and contextual elements of speech and allows the model to record word associations and dependencies in both forward and backward directions. During sentiment analysis, the attention method is incorporated to improve the model’s concentration on pertinent segments of the loaded sequence and enable it to dynamically determine the relative relevance of various words. Moreover, an encoder-decoder architecture is presented to comprehend the links between messages and the context of the discussion better. Sentiment-aware representations are produced by the decoder, after the encoder has processed the input messages. This process helps to capture the complex relationships and subtleties seen in WhatsApp conversations. A labeled dataset of WhatsApp chats with annotated sentiment labels, covering a range of emotions and sentiments frequently stated in casual talks, is used to train the model. The assessment shows that the suggested method outperforms conventional techniques in reliably assessing sentiment in WhatsApp chats and exhibits resilience in managing the complexities of informal language and contextual dependencies. The findings imply that the effectiveness of sentiment analysis in WhatsApp conversations is much improved by integrating Bi-LSTM networks with attention mechanisms and encoder-decoder models. By improving sentiment analysis approaches that are suited to the special qualities of conversational data in messaging systems, this research opens the door to a better comprehension of user sentiments in real-world communication environments.