Automated Translation Model Based on Deep Learning Algorithm

With the increasing exchanges between China and neighboring countries in industries, economy, culture, and other aspects, improving the level of machine translation between China and foreign countries has become a key factor in promoting international cooperation. However, existing machine translation systems often suffer from poor translation quality and high rates of entity word mistranslation when dealing with small-scale, small sample, and multilingual texts. This article explores a new method for solving the problem of neural network translation under low resource conditions, which utilizes data augmentation techniques and language models to improve the accuracy and readability of the translation, while also enhancing the translation accuracy of intra sentence vocabulary. By using vocabulary restriction method and constructing a large-scale entity dictionary, we have achieved accurate translation of entity words, effectively improving the quality of the translation. This article studies the construction of an English machine translation model based on deep learning algorithms, mainly using an encoding decoding structure and utilizing Gated Recirculation Unit (GRU) as its core language model. This model was trained and tested on a Chinese English translation corpus, optimizing performance by adjusting the dimensions of word embeddings and training batch sizes. The experimental results showed that as the text size increased, the BLEU (Bilateral Evaluation Understudy) score of the model gradually improved, reaching a maximum of 86 points; meanwhile, the model has high translation efficiency and excellent accuracy in translating entity words when dealing with large-scale texts and complex sentence structures. In addition, through the optimization of deep learning algorithms, our model exhibits high translation accuracy and quality in different text types. When dealing with different text sizes, the model exhibits gradually improving BLEU scores, reaching high levels of Chinese to English and English to Chinese respectively, reflecting its effectiveness in semantic conversion and grammar processing. These results not only reveal the superiority of the model in multilingual translation tasks, but also provide empirical support for further optimizing the application of deep learning algorithms in the field of machine translation.