Article Preview
Top1. Introduction
In current Machine Translation, word order model is very important. There are many problems in translation models based on hierarchical phrases, such as cross syntax translation errors, word loss phenomena, and so on. (Tong & Zhu, 2016) is also very serious. Since the reordering effect of hierarchical phrase translation model mainly depends on the selection of hierarchical rules, and hierarchical rules are also the most essential feature of hierarchical phrase translation model different from other models, in this paper, we propose a joint selection method of hierarchical rules, which comprehensively utilizes various features generated in the process of translation, and solves the problem of correct selection of hierarchical rules to a certain extent, And it effectively solves part of the reordering problem based on hierarchical phrase translation model(Yang et al., 2017). Nowadays, the difficulty of statistical machine translation lies in the low syntactic and semantic components contained in the model (Bhadoria & Chaudhari, 2019). Therefore, when dealing with language pairs with large syntactic differences, such as Chinese to English translation, there will be a long distance reordering problem. Sometimes the result of translation can't be understood even though word to word is correct. It can be said that the current mainstream statistical machine translation system is still at the bottom of the machine translation Pyramid (Li et al., 2017).
Machine translation refers to the automatic translation of text or speech from one language to another by computer, which belongs to the interdisciplinary research field of computer science and linguistics. In today's social life characterized by knowledge economy, the increasingly frequent international exchanges and the accelerating process of globalization make the total amount of cross language information exchange increase rapidly, and the natural language barrier between different countries and regions becomes more and more prominent. With the help of computer technology, people are eager to break down language barriers and realize smooth interpersonal and inter lingual communication. Therefore, machine translation research has great application value. Gradually improved machine translation technology has more and more extensive application in human social and economic life, and plays an extremely important role in the deeper and faster international information dissemination. Experts at home and abroad have also made a lot of research on hierarchical phrase machine translation (Liu et al., 2017).
Xiao Tong takes hierarchical phrase as the basic model and applies it to tree to string model reinforcement, which increases the space for translation derivation. In this framework, he studies the principles of machine translation and translation decoding technology (Buckley et al., 2019). In his experiment, the efficiency and performance of Chinese-English translation of machine translation have been greatly improved (Gulen et al., 2016) Jiangming introduces the shortcomings of machine translation at present. Aiming at these shortcomings, it uses Japanese case grammar and combines it with machine translation to mark different case information, which makes the language to be translated more obvious and easy to be analyzed. In addition, this method uses Japanese model, so it can achieve better results in Japanese translation, It proposes to use this framework in Japanese phrase translation Yegang hopes to integrate valuable syntax into machine translation, so that after translation, it can be more in line with the grammar and context of the country, and will not cause misunderstanding. Therefore, he puts forward three strategies to integrate bilingual into the machine translation, which integrates the phrases at the grammatical level in the way of hard constraints. The research finds that after the integration, the translated phrases and sentences are more perfect, and in complex long sentences, they are also more perfect It can play a certain role (Chen, 2019). These studies have certain reference value for this paper, but due to the lack of relevant research samples, there is no universal value, the research methods are not rigorous, and there are too few translation reference languages, so it can only improve certain theoretical reference.