Information Retrieval for Japanese Character Database based on LeNet-5-Saliency Algorithm
Tao Feng · 2023
With the increasing number of Internet character database size, the efficient analysis regarding the information is becoming the principle task for the intelligent analysis, hence, in this manuscript, information retrieval for Japanese character database based on LeNet-5-Saliency algorithm is studied. In the designed algorithm, the Hash model is considered, in order to learn more robust hash codes, the feature fusion operation is performed on global features and fine-grained features in the feature space, and the attention algorithm is used to fuse features for hash code learning. Furthermore, this study performs binary tree hash encoding on the searched first word string, that is, the first word detection and binary tree hash encoding run in parallel, which can save a lot of retrieval for text retrieval time. Through above operations with combination of the feature selection, the model is implemented and the comparison simulation is then conducted. The designed model has the Map value of better that 0.9 in most cases, that mean, the accuracy is higher than most of the state-of-the-art algorithms.