Augment Translators with AI: A LocWorld Recap


2020-08-29 04:50 Lilt


阅读模式 切换至中文

Alessandra Binazzi has seen what strong localization programs can accomplish and knows what it takes to get there. After all, as the Head of Global Localization at ASICS Digital, she understands how to grow a global brand and expand into new locales and regions.  Recently, Binazzi sat down with Lilt’s Co-founder and Chief Scientist John DeNero at LocWorldWide42, the newest chapter in LocWorld’s conference series. This year’s virtual conference focused on engaging global users and the importance of the user’s experience.    Moving Beyond Acceptable Quality To kickoff the conversation, DeNero talked about the current landscape of machine translation and localization, what it’s missing, and how it’s evolving for the future. And while machine translation has come a long way, its raw output still leaves room for error.  That’s why many companies turn to Machine Translation Post-Editing (MTPE), but while a human linguist is now involved in the process, there are still quality issues. The linguist is only fixing what is already broken and, as a result, the translation sounds like the original machine translation. Ultimately, companies aren’t looking for acceptable quality. They’re certainly not looking for delayed projects due to those quality issues.  The human-in-the-loop machine translation system, on the other hand, actually improves over time with each translated word. The linguist accepts or rejects a suggestion, which trains the system and improves its accuracy. The outcome of this process is increased translator efficiency (not having to fix the same mistakes over and over) and higher translation quality over time.   Human-in-the-Loop: In Action Binazzi saw the results firsthand when she decided to bring Lilt onboard to help with ASICS Digital’s localization efforts. The digital ecosystem, in charge of e-commerce, fitness applications, global digital marketing, and more, was spending too much time translating product descriptions.  To continue growth, ASICS aimed to add two new locales: Polish and Portuguese. To get these languages up to speed, they needed to localize four seasons worth of product descriptions, or about 4x the normal amount of content. For a traditional localization process, that wouldn’t be possible - they were already lagging on copy, so how could they quadruple the normal output? Instead, they decided to partner with Lilt, and the results were immediately clear: a 70% cost savings, a 60% improvement in turnaround time speed, all while maintaining the same quality as the human translators.    Looking Towards Innovation The team at ASICS Digital is still working towards the future and is making sure that they don’t remain complacent. They’re looking to continue optimizing processes internally so they can focus on improved response times and smoother product launches, and turning to more automation and innovation so they don’t have to worry about the little things.  “Lilt checked a very important innovation checkbox for us, and we’re very proud of that,” Binazzi said. “We want to continue to innovate, and we look to Lilt to provide more of those solutions in the future.” • • • If you want to watch the full panel from LocWorldWide42, check out the discussion on our YouTube channel here. Want higher quality, consistency, and deliver speeds? Augment your translators with an adaptive, human-in-the-loop localization workflow by requesting a demo today.
亚历山德拉·比纳齐(Alessandra Binazzi)已经看到了大本地化项目能够实现的目标,并且知道要达到这一目标需要做些什么。毕竟,作为ASICS Digital的全球本地化主管,她十分清楚如何去发展一个全球品牌,并将其拓展到新区域和地区。 最近,比纳齐与Lilt的联合创始人兼首席科学家约翰·迪内罗(John DeNero)一起参加了LocWorldWide42的座谈会,这是LocWorld会议系列的最新章节。今年的虚拟会议侧重于吸引全球用户和强调用户体验的重要性。 超越可接受的翻译质量 会议的开始,迪内罗谈到了机器翻译和本地化的现状,二者有什么缺点,以及未来应如何发展。虽然机器翻译已经取得了很大的进步,但它的原始输出仍留有出错的空间。 这就是为什么许多公司转向机器翻译后期编辑(MTPE),但尽管现在有一个人类语言学家参与到这个过程中,机器翻译仍存在质量问题。语言学家只是修正机器翻译译错的地方,这样看来,译后编辑似乎与原始的机器翻译没有什么区别。归根结底,企业并不是在寻求可接受的质量。他们当然不会寻找那些由于质量问题而延期的项目。 另一方面,随着时间的推移,每翻译一个单词,人为参与的机器翻译系统就会不断改进。语言学家接受或拒绝一个机器给出的建议,就是在训练系统并提高其准确性。这一过程的结果是提高了翻译效率(不必一次又一次地修正相同的错误),并逐渐提高了翻译质量。 利用人为参与的机器翻译 当比纳齐决定使用Lilt来帮助ASICS Digital的本地化成果时,她亲眼目睹了结果。负责电子商务、健身应用、全球数字营销等业务的数字生态系统在翻译产品描述方面所花费的时间太长了。 为了推进本地化进程,ASICS打算增加两个新区域:波兰和葡萄牙。为了让这两种语言跟上速度,他们需要本地化四个季度的产品描述,或者本地化四倍左右的内容。对于一个传统的本地化过程来说,这是不可能的。他们在复制上就已经落后了,那么他们怎样能把正常的输出提高四倍呢? 相反,他们决定与Lilt合作,结果马上就很显而易见:节省了70%的成本,提高了60%的周转时间和速度,同时还保证质量与人工翻译相等。 走向创新 ASICS Digital的团队仍在为未来而努力,并保证他们不会自满。他们正在寻求继续优化内部流程,以便能够专注于提高响应时间和更平稳地发布产品,实现更多的自动化和创新,从而不必担心细节。 比纳齐说:“Lilt给我们指明了一条非常重要的创新道路,我们为此感到非常自豪。”“我们希望继续创新,我们期待Lilt在未来提供更多这样的解决方案。” …… 如果您想从LocWorldWide42观看完整的座谈会,请查看我们YouTube频道的讨论。想要更高的质量,完美的一致性和更快的交付速度?通过申请演示,用一个自适应的,人为参与的本地化工作流来强化你的翻译人员吧。