Google Says There’s a Better Way to Create High-Quality Training Data for AI Translation

In an October 14, 2024 paper, Google researchers highlighted the potential of AI translations refined by humans or human translations refined by large language models (LLMs) as alternatives to traditional human-only references.

Talking to Slator, Zhongtao Liu, a Software Engineer at Google, explained that their study addresses a growing challenge in the translation industry: scaling the collection of high-quality data needed for fine-tuning and testing machine translation (MT) systems. 

With translation demand expanding across multiple languages, domains, and use cases, traditional methods that rely solely on human translators have become increasingly expensive, time-consuming, and hard to scale.

To address this challenge, the researchers explored more efficient approaches to collect high-quality translation data. They compared 11 different approaches — including human-only, machine-only, and hybrid methods — to determine the most effective and cost-efficient one.

Human-only workflows involved either a single human translation step or included an additional one or two human review steps. Machine-only workflows ranged from single-step AI translations using top AI systems — MT systems or LLMs — to more complex workflows, where AI translations were refined by an LLM. Hybrid workflows combined human expertise and AI efficiency; in some cases, AI translations were refined by humans (i.e., post-editors), while in others, human translations were refined by LLMs.

They found that combining human expertise and AI efficiency can achieve translation quality comparable to, or even better than, traditional human-only workflows — all while significantly reducing costs. “Our findings demonstrate that human-machine collaboration can match or even exceed human-only translation quality while being more cost-efficient,” the researchers said.

The best combination of quality and cost appears to be human post-editing of AI translations. This approach delivered top-tier quality at only 60% of the cost of traditional human-only methods, while maintaining the same level of quality.

“This indicates that human-machine collaboration can be a faster, more cost-efficient alternative to traditional collection of translations from humans, optimizing both quality and resource allocation by leveraging the strengths of both humans and machines,” they noted.

The researchers emphasized that the quality improvements stem from the complementary strengths of human and AI collaboration, rather than from the superior capability of either the AI or the human (post-editor) alone, underscoring the importance of leveraging both human and AI strengths to achieve optimal translation quality.

They noted that LLMs were less effective than human post-editors at identifying and correcting errors in AI-generated translations. On the other hand, human reviewers tended to make fewer changes when reviewing human-generated translations, possibly overlooking certain errors. Interestingly, even additional rounds of human review did not substantially improve the quality. This observation supports the argument for human-machine collaboration, where each component helps address the other’s blind spots, according to the researchers.

“These findings highlight the complementary strengths of human and machine post-editing methods, indicating that a hybrid method is likely the most effective strategy,” they said.

Authors: Zhongtao Liu, Parker Riley, Daniel Deutsch, Alison Lui, Mengmeng Niu, Apu Shah, and Markus Freitag

Smartcat's Series C Funding, YouTube Dubs Launch, Viva Translate Closes Down

Slator- Language Industry Intelligence

Florian and Esther discuss the language industry news of the week, where they give their impressions from SlatorCon Silicon Valley and touch upon the findings from the 2024 ALC Industry Survey.

In a significant funding update, Esther reports that Smartcat raised USD 43m in a Series C round, bringing their total funding to USD 70m. This funding will support product innovation in AI translation and multilingual content generation.

Florian talks about YouTube’s potential launch of AI dubbing, a feature in testing that aims to generate translated audio tracks for videos, significantly enhancing content accessibility and engagement.

In Esther’s M&A corner, Cloudbreak, now rebranded as Equiti, acquired its competitor Voyce and brought on a new private equity partner, Heritage Group. Meanwhile, EasyTranslate acquired World Translation, expanding its reach in the Nordic and DACH regions.

The duo bid farewell to publicly traded Keywords Studios, which is delisting after being acquired by private equity firm EQT. They also note the shutdown of Viva Translate, a speech-to-speech translation company that will open-source its tools as it winds down.

Can AI Agents Execute Complete Translation Workflows?

Slator- Language Industry Intelligence

The Evolution of Translation

Translation has come a long way from the days of bilingual dictionaries and phrasebooks. The need to bridge language barriers has driven innovation, bringing us to the age of digital translation tools and now, AI agents. But can AI truly handle the complexity of complete translation workflows?

Can AI Agents Execute Complete Translation Workflows?

The Rise of AI in Translation

Artificial Intelligence (AI) has revolutionized many industries and translation is no exception. The question isn't just about AI performing translations but about AI agents managing entire translation workflows. Let's dive deeper into this fascinating development.

Understanding AI Agents

What are AI Agents?

AI agents are autonomous entities designed to perform specific tasks. These tasks range from simple commands to complex problem-solving activities, all without human intervention. In the context of translation, AI agents can automate processes, ensuring efficiency and consistency.

How AI Agents Work

AI agents operate through machine learning algorithms, constantly evolving by processing new data. They analyze patterns, learn from previous translations, and improve their accuracy over time. Their ability to handle repetitive tasks makes them invaluable in translation workflows.

The Role of AI in Translation

AI vs. Human Translators

While human translators bring cultural sensitivity and contextual understanding, AI offers speed and consistency. The debate often centers on whether AI can match the nuanced understanding of a human. However, AI's rapid advancements suggest a complementary relationship rather than a competitive one.

Advantages of AI in Translation

AI excels in handling large volumes of text quickly, making it ideal for businesses needing fast turnaround times. It also reduces costs and ensures uniformity in translations, essential for maintaining brand voice across different languages.

Components of a Translation Workflow

Pre-Translation Processes

Before translation begins, tasks such as data preparation, terminology management, and content analysis are crucial. These steps set the foundation for accurate translations.

Translation Phase

This is the core of the workflow, where text is translated into the target language. AI agents use machine learning and natural language processing (NLP) to perform this task.

Post-Translation Processes

Quality assurance, editing, and proofreading ensure the final product meets the desired standards. This phase is critical for catching any errors and refining the translation.

AI in Pre-Translation

Data Preparation

AI agents can efficiently sort and prepare data, identifying relevant content and discarding unnecessary information. This streamlines the workflow and sets the stage for accurate translations.

TMSs at a Crossroads

The production side of language services has heavily relied on the tried and true features of translation management systems (TMSs) since the 1990s. Until neural machine translation entered the localization process, the general structure of TMSs underwent little change. 

Things are very different in July 2024. Machine translation (MT), now enabled by AI, is but a small component of the translation and localization cycle, and the management aspects of the process can all now be highly automated and integrated using AI. 

While a few of the well-established TMSs have incorporated some level of automation, new products continue to enter the market, at the same time driving localization buyer expectations. A look at AI orchestration for localization, for example, can alone serve as an example of what is now possible.

We asked readers if they are happy with their TMS, and most responders (48.0%) said “not really, needs improvement.” Over a third (36.0%) believe their current choice does the job, and the rest are content (16.0%) with it.

Terminology Management

Consistency in terminology is vital, especially for technical documents. AI agents manage glossaries and ensure that specific terms are used consistently throughout the translation.

AI in the Translation Phase

Machine Translation Engines

At the heart of AI translation are machine translation engines like Google Translate and DeepL. These engines have evolved to provide more accurate and contextually relevant translations.

Contextual Understanding

AI agents analyze context to avoid literal translations that miss the mark. By understanding the context, they can deliver translations that make sense in the target language.

AI in Post-Translation

Quality Assurance

AI-driven quality assurance tools check for consistency, grammar, and style. They can flag potential issues, ensuring the final translation meets quality standards.

Editing and Proofreading

While AI handles the bulk of translation, human editors often step in for final proofreading. This hybrid approach combines the efficiency of AI with the finesse of human touch.

Challenges in AI-Driven Translation Workflows

Language Nuances and Context

Languages are full of nuances and idiomatic expressions that AI might not fully grasp. This is a significant challenge in achieving high-quality translations.

Cultural Sensitivity

Cultural differences influence language use. AI must be trained to recognize and respect these differences to avoid misinterpretations.

Overcoming Challenges with AI

Continuous Learning Algorithms

AI agents continuously learn from their mistakes and successes. This ongoing learning process helps them adapt to language nuances and cultural sensitivities.

Human-AI Collaboration

Combining AI's efficiency with human translators' expertise creates a robust translation workflow. Humans provide context and cultural insight, while AI handles repetitive tasks.

Future of AI in Translation

Innovations on the Horizon

AI technology is constantly evolving. Future innovations promise even more accurate and contextually aware translations.

Long-term Impacts

The long-term impact of AI on the translation industry includes greater efficiency, reduced costs, and the potential for AI to handle increasingly complex tasks.

Ethical Considerations

Data Privacy

Ensuring data privacy is paramount in AI-driven translation workflows. AI agents must handle sensitive information securely to maintain trust.

Bias in AI Models

AI models can inadvertently reflect biases present in training data. Addressing and mitigating these biases is crucial for fair and accurate translations.

Comparing AI Translation Tools

Popular Tools and Their Features

Comparing popular AI translation tools like Google Translate, DeepL, and Microsoft Translator helps users choose the best tool for their needs.

Performance Metrics

Evaluating tools based on accuracy, speed, and user satisfaction provides a comprehensive view of their performance.

User Adoption and Acceptance

Training and Onboarding

Proper training and onboarding are essential for users to maximize the benefits of AI translation tools.

User Feedback and Adaptation

User feedback is crucial for continuous improvement. AI agents must adapt based on user experiences to enhance their performance.

Conclusion

Summary of Key Points

AI agents are transforming translation workflows by offering speed, efficiency, and consistency. While challenges remain, continuous learning and human collaboration are paving the way for more accurate translations.

The Road Ahead for AI in Translation

The future of AI in translation looks promising, with ongoing innovations and increasing integration into workflows. The balance between AI and human translators will continue to evolve, creating more robust and reliable translation solutions.

 Landexx, a language services provider based in Germany, has filed for bankruptcy.

According to a court filing reported by several German legal aggregation sites, the language services provider Landexx has filed for bankruptcy.

Information on the LSP is now difficult to find, as Landexx's website appears to be down. Additionally, as a private company, Landexx’s annual financial reports are not publicly accessible.

German Language Services Provider Landexx Files for Bankruptcy

Landexx, led by Managing Director Christel Stemmer, provided various services, including translation, interpreting, language training, and desktop publishing.

Language professionals have informed colleagues of the news. For example, German-English translator Jill Sommer advised freelancers in a blog post to contact the bankruptcy trustee, Stephan Höltershinken, about any unpaid invoices. She also warned others not to accept translation assignments from the company.

For at least a few years before the LSP's bankruptcy filing, there had been rumblings about Landexx among freelancers online.

“They were an excellent client, always very professional, EXCEPT that I always had to chase late payments. They often paid 3 to 4 months late,” read a November 2020 complaint by one translator who said she began working with Landexx in 2009.

The freelancer explained that she had to "chase payments by email and phone " and wait a year, even hiring a lawyer, to receive payment for six outstanding invoices.

"[T]hey have told my lawyer that they will no longer be sending me any work, even though I did nothing wrong," she added. "I believe other freelance translators must be in the same position."

Another translator wrote on Reddit in May 2024 that they were pursuing legal action against Landexx through a court order for payment "[a]fter countless reminder e-mails that have been ignored."

Landexx also had a mixed reputation on the job board ProZ, where the LSP’s "Blue Board affiliation" — based on freelancers’ willingness to work with the company again — stood at two out of five stars. A December 2022 staff note indicated that "[t]his outsourcer has been banned from posting jobs at ProZ.com." Details regarding Landexx’s specific case were not disclosed.

“Use of the site dishonestly or fraudulently will result in termination of site use and associated privileges,” states ProZ’s termination policy, which also specifies that ProZ "reserves the right to refuse access to this site to any party without giving a reason."