Patent-pending Generative AI-based translation technology from memoQ

memoQ正在申请专利的基于人工智能的生成式翻译技术

2023-09-06 21:00 MemoQ

本文共323个字,阅读需4分钟

阅读模式 切换至中文

Budapest, September 6, 2023 – memoQ, developer of the memoQ Translation Management System (memoQ TMS), has filed a patent application for a new Generative AI-based translation automation solution. memoQ Adaptive Generative Translation, or memoQ AGT in short, is a novel approach to translation automation. With memoQ AGT, translations are generated by a Large Language Model (LLM), similar to the model powering the popular ChatGPT service. memoQ AGT provides instant domain adaptation to generate translations tailored to the customer’s existing language resources and achieves this without retraining or fine-tuning the model itself. As a result, both the language data and the control over the translation process remain entirely with the customer. “Translation automation is at its best when it is based on the user’s existing domain data”, explains Gábor Ugray, founder of memoQ and inventor of memoQ AGT, “and an LLM can incorporate this data instantly and on a segment-by-segment basis. memoQ AGT offers the best of both worlds when it combines the generative power of an LLM with the high-quality linguistic data that users of memoQ TMS already have.” “The data needed for this is right there in the translation management system,” Ugray adds, “because that’s where the user’s assets are stored—in translation memories, term bases, and corpora. In fact, the key to good generative translations is bilingual data management and data hygiene, which are the core functionalities of a TMS.” memoQ AGT builds on memoQ’s existing LiveDocs and TM+ technologies, which allow users to add content to memoQ’s resources in large amounts and in a wide variety of formats and make them ready for reuse instantly. memoQ also partners with Microsoft Azure, both to provide for the infrastructure needs of memoQ TMS and the large language models to power memoQ AGT. memoQ AGT is currently being tested and integrated into memoQ’s other technologies and will be released to users later in 2023.
布达佩斯,2023年9月6日——memoQ翻译管理系统(memoQ TMS)的开发商memoQ已经为一种新的基于生成式人工智能的翻译自动化解决方案提交了专利申请。 memoQ自适应生成翻译,简称memoQ AGT,是一种新的翻译自动化方法。使用memoQ AGT,翻译由大型语言模型(LLM)生成,类似于支持流行的ChatGPT服务的模型。memoQ AGT提供即时领域适应,以生成适合客户现有语言资源的翻译,并且无需重新训练或微调模型本身即可实现这一点。因此,语言数据和对翻译过程的控制权都完全属于客户。 memoQ的创始人和memoQ AGT的发明者G á bor Ugray解释说:“当翻译自动化基于用户现有的领域数据时,它处于最佳状态,LLM可以即时地、逐段地整合这些数据。当memoQ AGT将LLM的生成能力与memoQ TMS用户已经拥有的高质量语言数据结合在一起时,它提供了两个世界的最佳选择。” “所需的数据就在翻译管理系统中,”Ugray补充道,“因为这是用户资产存储的地方——在翻译记忆库、术语库和语料库中。事实上,良好的生成式翻译的关键是双语数据管理和数据卫生,这是TMS的核心功能。” memoQ AGT建立在memoQ现有的LiveDocs和TM+技术的基础上,允许用户以各种格式向memoQ的资源中大量添加内容,并使它们可以立即重用。 memoQ还与微软Azure合作,提供memoQ TMS的基础设施需求和支持memoQ AGT的大型语言模型。 memoQ AGT目前正在测试并集成到memoQ的其他技术中,将于2023年晚些时候向用户发布。

以上中文文本为机器翻译,存在不同程度偏差和错误,请理解并参考英文原文阅读。

阅读原文