A Novel Long and Short-Term Memory Network-Based Krill Herd Algorithm for Explainable Art Sentiment Analysis in Interior Decoration Environment

Aiming at the problem that most existing models of art sentiment analysis only consider text encoding from word level, this paper proposes a novel long and short-term memory network-based krill herd algorithm for explainable in interior decoration environment. Firstly, multi-scale convolution is used to capture local correlation different granularity, so as obtain more semantic information levels form richer representation. Then, gating mechanism introduced control path flowing aggregation layer. An improved swarm based on cosine factor Cauchy proposed solve model. Finally, full connection layer argmax function are achieve classification. The experimental results show compared with other advanced models, model can improve accuracy emotion classification by 2.3% 0.8% respectively two public data sets IMDB Yelp2014, minimum root mean square error (RMSE).