Integration of Machine Translation Systems and Language Models Based on Big Data

Early language modeling and machine translation were limited by artificially constructed rules, and rule-based methods often required the establishment of a large number of rules to express their features due to their complexity and variability. However, their maintenance and complexity often depend on the experience and knowledge of experts, and cannot fully describe language phenomena. In order to compensate for this drawback, statistical based machine translation research has gradually emerged, with the main research approach being to extract essential features of a language from massive labeled samples. However, the cost of this process is enormous. At present, machine translation based on artificial neural networks can effectively overcome the shortcomings of traditional statistical methods. Compared to traditional algorithms such as Transformer neural network algorithms, due to the large number of network parameters it contains, it often requires a large amount of data for parameter estimation, resulting in lower translation efficiency. Based on the above issues, this article explored a machine translation system based on the combination of BP (back propagation) neural network algorithm and GA (Genetic Algorithm) algorithm, and verified through experiments that the efficiency of the machine translation system constructed by this algorithm was higher than other algorithms. The experimental results demonstrated that the accuracy, recall, and F1 of the BP-GA algorithm were 0.8835,0.8775, and 0.8284, respectively, which were higher than the accuracy, rec