基于Django的汉维地址翻译网络服务的设计与实现  

Design and implementation of Chinese-Uyghur Addresses Translation Network Service Based on Django

作　　者：斯拉吉艾合麦提·如则麦麦提 艾山·吾买尔[1,2] 张济民 汪烈军 刘胜全[1,2] Sirajahmat Ruzmamat;Aishan Wumaier;ZHANG Ji-min;WANG Lie-jun;LIU Sheng-quan(College of Information Science and Engineering,Xinjiang University,Urumqi 830046;Xinjiang Laboratory of Multi-Language Information Technology,Xinjiang University,Urumqi 830046)

机构地区：[1]新疆大学信息科学与工程学院,乌鲁木齐830046 [2]新疆大学新疆多语种信息技术实验室,乌鲁木齐830046

出　　处：《现代计算机》2020年第19期99-104,共6页Modern Computer

基　　金：国家自然科学基金项目(No.61662077、61262060);新疆多语种信息技术实验室开放课题(No.2016D03023);自治区高校科研计划自然科学项目(No.XJUEDU2019I004)。

摘　　要：命名实体是文本信息的主要承载者,识别与翻译命名实体是目前自然语言处理领域的研究热点。地址翻译在学术上属于命名实体翻译任务,地址表达文本中的关键信息,在机器翻译系统中,地址的正确翻译对译文整体翻译质量具有十分重要的影响。使用PyTorch构建基于神经网络的地址翻译模型,设计与实现基于Django的汉语-维吾尔语地址在线翻译网络服务,并使用UWSGI以及Nginx服务器来负载均衡。实验结果表明,所设计的系统具有良好的并发性以及高鲁棒性。Named entity is the main carrier of text information.At present,recognition and translation of named entity is a research hotspot in the field of natural language processing.Address translation belongs to the task of named entity translation.Address is used to express the key information in the text.In machine translation system,the correct translation of address has a very important impact on the overall translation quality.In this paper,we use PyTorch to build an address translation model based on neural network,design and implement an online Chinese Uyghur address translation network service based on Django,and use UWSGI and Nginx servers to balance the load.Experimental results show that the system designed in this paper has good concurrency and high robustness.

关 键 词：命名实体 地址 神经机器翻译 DJANGO 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]