机器翻译辅助开发平台的设计与实现

Design and Implementation of a Platform of Computer-assistant MT

介绍了一个机器翻译辅助开发平台MTDevStudio中的部分关键技术,包括模型定义、句法树操作的实现、语料库规则提取与统计分析、翻译引擎的评测等,重点对基于规则的机器翻译的辅助开发实现进行了介绍。该平台有利于减少机器翻译过程中的大量重复性工作,使研究者致力于自然语言理解的核心理论研究。与其它辅助工具相比,该平台致力于集成机器翻译涉及到的全面工作,用户可以自定义机器翻译模型,可扩展性好,并由机器翻译专家实际参与使用、评价,取得了不错的效果。

The paper introduces several key technologies of assistant development platform for machine translation called MTDevStudio including model definition, operations of syntax trees, rule retrieving from corpus and rule statistics, MT engines＇ marking, etc. It focuses on rule-based MT. This platform can reduce a lot of duplicated works for researchers and allow researchers focus on core MT researches. Compared with other MT assistant tools, MTDevStudio is inclined to integrate all of MT involved works. Researchers can define their own MT models. MTDevStudio is extensible in assistant function modules and MT models. Several MT experts involved the platform development and used it in research and gave a good mark to MTDevStudio.