Enhancing computer-aided translation system with BiLSTM and convolutional neural network using a knowledge graph approach

This research aims to develop platform integration, service discovery, and recommendation system within the Internet of Things (IoT) industry. It accomplishes this endeavor through the integration of improved Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) techniques to construct a Capability Knowledge Graph (CKG) for the IoT and a computer-aided translation (CAT) system. Initially, the distinctive characteristics of IoT are analyzed, leading to the proposition of a Bidirectional Long Short-Term Memory (BiLSTM) model designed to extract specific vocabulary from textual corpora. Subsequently, the BiLSTM-Convolutional Neural Networks (BiLSTM-CNN) model is introduced. Finally, extensive experiments are conducted to evaluate the effectiveness of the proposed model. The results demonstrate the superior performance of the model in entity labeling within sentences. Specifically, it achieves an accuracy of 85.1% for the top 100 entities and 80.0% for the top 200 entities. Additionally, this model surpasses current state-of-the-art methods in relation extraction, attaining the highest extraction results on the dataset. The constructed translation system exhibits an accuracy of 60.22% and a remarkable recall rate of 59.64%. The proposed system significantly contributes valuable data support for the development of IoT CKGs and CAT systems.