A Novel Deep Learning Language Model with Hybrid-GFX Embedding and Hyperband Search for Opinion Analysis

Policies, legislation, surveillance, monitoring, direction, and enforcement, are heavily influenced by public opinion or emotion. Due to the increase in electronic data generation, it has been forced to do an automatic analysis of this opinion or feelings termed as opinion analysis. To process massive volumes of data, deep learning is now trending. Word embeddings serve an essential role of feature representatives in deep understanding. The present paper offers a novel deep learning architecture that represents hybrid embedding that deals with polysemy, semantic, and syntactic problems in a language representation. The effectiveness of a deep learning model is extremely sensitive to using hyperparameters. Here, the proposed a novel Hybrid-GFX–Attention–BiGRU–CNN model with a hyperband language model. Hyperband search is used to find optimal values for the model's hyperparameters. To justify classification results, statistical and graphical approaches have been used. We analyzed the model's efficacy using the MR and Hate speech data sets. The model’s performance is quite promising compared with existing state-of-the-art architectures.