Is Your Translation Really Accurate? CATER v2, the Translation Quality 'Visualization' Tool, Officially Launches Today

ErudAite Inc. (Headquarters: Ginza, Chuo-ku, Tokyo; CEO: Zou Do Iida; hereinafter "the Company") officially launched its translation quality diagnostic service "CATER v2" today. The basic mode is available free of charge. At the end of this page, an 80% off coupon is available for the paid advanced mode, which offers detailed evaluation. Please take advantage of it.

Service URL: https://cater.erudaite.ai

Accessible using accounts from Stand By Me, our personal translation tool.

Why Has Translation Quality Been "Unclear"?

When companies translate documents or individuals use translation services, there is a fundamental problem: if you don't understand the target language, you have no way to verify whether the resulting translation is correct.

Whether you hire a human translator or use machine translation, ultimately, you can only "accept it on faith." Even if you have quality concerns, you cannot pinpoint exactly what is wrong or how to improve it. You also lack the evidence to request corrections. As a result, an information asymmetry has always existed between translation clients and providers.

CATER v2 was developed to change this structure.

Diagnose Translations Simply by Inputting Source and Target Texts

The usage is simple. Input the source and translated text, then press the "Evaluate" button. That’s all it takes. The translation quality is assessed across six evaluation axes, and a detailed report with a 100-point score is generated.

Being told a translation is "correct" doesn’t feel concrete. But being told "grammatical accuracy is 95, but stylistic appropriateness is only 68, with three expressions too casual for a business document" makes it clear exactly what and how to improve.

What CATER provides is the "visualization" of translation quality.

Evaluation is possible with just source and translated text (the target of evaluation). Accuracy improves further when the translation intent (brief) is also provided.

Evaluation Without a "Reference Translation"

Traditional translation quality evaluation has primarily relied on comparing translations against a human-created "correct" reference translation. However, in reality, most situations lack such a reference, making evaluation impossible. If a correct translation already existed, there would be no need to evaluate.

CATER v2 does not require a reference translation (a "correct" version). Evaluation is completed using only the source and translated text. This enables anyone to diagnose any translation at any time.

Value for Translation Clients

CATER v2 serves as a negotiation and decision-making tool for translation clients.

You can objectively assess the quality of translations delivered by external providers, even if you don't understand the target language. Problematic areas are highlighted in the report with improvement suggestions. You can compare the quality of multiple translators or translation services on the same standardized criteria.

Move from decisions like "I'm vaguely uneasy, so I'll hire an expensive translation company" to decisions like "I'll choose the appropriate service based on quality scores." CATER rationalizes translation procurement.

100-Point Evaluation Across Six Axes

Value for Translation Providers

For translators and localization companies, CATER serves as a means to prove quality.

By submitting the CATER evaluation report alongside deliverables, translation quality can be demonstrated using a third-party standard. Being able to state "this translation scored over 80 points on all six evaluation axes" directly builds client trust.

Translations scoring over 80 points on CATER are considered to meet sufficient quality standards for commercial use.

On the other hand, many AI and machine translations actually score between 40 and 90, showing significant variability, and frequently fall below 80. This is likely due to major misunderstandings, translations that stray from context, or hallucinations (fabrication of sentences, insertion of unwarranted additions).

Using CATER, translators can confidently argue, "Use human translators, not AI translation."

Evaluation Screen

Improved Accuracy by Communicating "Translation Intent"

CATER includes a function to input the translation brief (purpose and intent of the translation).

For example, when translating a passage from a novel, if the brief specifies "prioritize literary expression," then free translation and creative expressions will be scored higher. Without such a brief, the same translation might be penalized for deviating from the original.

This mechanism allows CATER to evaluate not just "accurate" translations, but also "purpose-appropriate" translations. After all, what makes a good translation depends on its purpose. We believe this functionality is particularly important for creative content.

Reports Available in 59 Languages

Evaluation reports can be output in 59 languages, regardless of the source or target language. For instance, you can verify the result of translating a Japanese original into English using a French report. This supports quality management within global teams.

Born from the Development of a High-Precision Translation Engine

CATER originated as an internal tool developed by our company to improve the quality of our proprietary translation engine. It was created to automatically diagnose engine outputs and drive continuous improvement cycles.

After being published as an academic paper in December 2024 (arXiv: 2412.11261), the framework was significantly expanded into version 2. Following a pilot operation as a SaaS starting in June, and after enhancing the reporting functions, it is now officially launching today.

Comment from CEO Zou Do Iida

"The biggest problem in translation isn't the quality of the translation itself. It's that clients cannot verify the quality. If you don't understand the language, you have no choice but to accept whatever is delivered. This paradigm has created structural inefficiencies in the translation industry. CATER provides a common language that allows everyone involved in translation to discuss quality on the same objective criteria. For translators, it becomes a means to objectively prove the value of their work. Making translation a matter of trust and evidence—that is the world CATER aims to create."

[Service Overview]

Service Name: CATER v2 (Comprehensive AI-assisted Translation Edit Ratio)

Official Launch: July 3, 2026

Delivery Model: SaaS (accessible via web browser)

Usage Modes: Basic (free, guest access available) / Advanced (requires SBM account)

Six Evaluation Axes: (GP / SI / FC / TC / DC / CSA)

Report Languages: 59 languages

Service URL: https://cater.erudaite.ai

[80% Off Coupon]

For the first three months, exclusively for the STATER plan (980 JPY/month): BOOSTER1

[Company Overview]

Company Name: ErudAite Inc. (ErudAite)

Address: 1-22-11 Ginza, Chuo-ku, Tokyo

Representative Director: Zou Do Iida

Established: July 2024

Business: Development and provision of AI translation engines

Corporate Website URL: https://about.erudaite.ai