Multi-Keyword Automatic Recognition of Low-Resource Continuous English Text for Content Filtering

In low-resource environments, traditional neural networks often struggle to accurately and efficiently recognize multiple keywords in continuous English text with grammatical errors, leading to limitations in content filtering and information retrieval. To address these challenges, this paper proposes a novel automatic keyword recognition model by integrating the Transformer structure with a Generative Adversarial Network (GAN). The Transformer, enhanced by a self-attention mechanism, captures subtle semantic relationships in text, while the GAN further optimizes the model's ability to identify and correct grammatical errors. Experimental results demonstrate that the proposed model achieves a detection accuracy of 0.96, stability of 0.97, and a low mean square error of 0.01, significantly outperforming traditional methods such as CNN and LSTM-Transformer. The model not only improves the accuracy and efficiency of keyword recognition but also provides a practical solution for content filtering in low-resource settings, offering potential applications in natural language processing and educational technology.