Translation and Generative Optimization Strategies in English Question Answering Systems based on BERT and Generative Adversarial Networks

Translation and Generative Optimization Strategies in English Question Answering Systems based on BERT and Generative Adversarial Networks

연구 분야: Artificial Intelligence

학회: 2025 International Conference on Intelligent Systems and Computational Networks (ICISCN)

With the rapid development of globalization, the demand for cross-linguistic communication is becoming increasingly urgent. However, traditional English question answering systems often face issues such as low translation accuracy and poor answer generation quality when handling cross-language tasks. This article proposes an optimization strategy for an English question answering system based on BERT (Bidirectional Encoder Representations from Transformers) and generative adversarial networks (GANs). Through deep semantic understanding and generative adversarial mechanisms, GANs models are applied into the question answering system, and the competition mechanism between generators and discriminators is utilized to continuously optimize the answer generation process. The findings demonstrate that after the optimization strategy with BERT and GANs, the translation accuracy is also maintained at a high level of over 80% in each iteration, and the maximum even reaches 92%. The method generates more accurate and smooth translations and answers when dealing with cross-language question and answer (Q&A) tasks.