Katydid Inspired BERT Optimization for Improved Sentence Semantics in Text Classification

The proposed Katydid-Inspired BERT Optimization for Improved Sentence Semantics in Text Classification (KyBERT), aims to optimize the ability of the model to identify more complex linguistic patterns along with long-range dependencies. Bio-inspired optimization on KyBERT, where the inspiration is drawn from the katydid's sensory capabilities, enhances better handling with context-rich and resource-dense language. Generally these models enhance the values of key metrics like classification accuracy, F-measure, and Fowlkes-Mallows Index (FMI) since sentence semantics was poorly represented in simple models. This proposed model KyBERT has also confirmed that useful in creating a much-refined framework, that rarely gives false predictions and significantly improves text classification rate. The evaluation based on the Amazon Reviews dataset with 476,001 unique IDs established a high performance of text classification with 64.29%, F-measure of 64.67% and FMI of 64.68%.