How We Translate at Honyaku Cloud

1. Read through the Source Text in its entirety

This sounds like a no-brainer, and I’m sure we all do this already, but reading through the source text in its entirety can speed up the translation process exponentially, especially when generating ideas via machine translation, like ChatGPT, because you have a better knowledge of the context surrounding it.

2. Use AI tools to brainstorm ideas

We often leverage ChatGPT using specific MTPE-related prompts

⚠️ We use AI tools like ChatGPT as brainstorming tools and ideation assistants, not as a translation solution. Below are some of our favorite prompts when using AI during th translation process

If using ChatGPT, please make sure that “Chat history & training” is turned off for your account before using the prompts below.

• Different types of prompts
  • gpt-translator
‍I want you to act as an English translator, editor, and copywriter. I will speak to you in any language and you will detect the language, translate it into English. Keep the meaning the same, but make the translation literary. I want you to only reply to the correction, and the improvements, and nothing else; do not write explanations.‍
  • gpt-multi
‍Translate into English in several different ways that maintain accuracy and retain nuance. The translations should also be eloquent and natural-sounding to a native English speaker and should not add any extra information or explanations.‍
  • gpt-match
‍Please review the Original Translation for accuracy and grammar/spelling against the Original Japanese. Only make edits when necessary for readability. Display the results, one sentence per row, in a four-column table showing Original Japanese, Original Translation, Edited Translation, and a fourth column explaining if the original translation matches the original Japanese. Make a new row each time you see this character: 。 or if there is a line break and use bolding to indicate what has changed between the Original Translation and the Edited Translation.
    Original Translation:
    Original Japanese:‍
  • gpt-revise
‍Try to revise in the same language based on the provided paragraph. You should follow the writing style of the provided paragraph. You should not change the writing style, such as making a formal paragraph casual. You should only improve the user’s grammar and vocabulary and make sure it sounds natural.

• Install and leverage Espanso to save time with keyboard shortcuts!
  • Espanso
    ‍https://espanso.org/

• GPT Strategies
  • Larger chunks of text for better flow
  • Utilizing GPT to organize data/information (two-column table)
  • Always take it with a grain of salt! (Often wrong)

3. Leave meaningful comments that are mindful (and actionable) by the proofreader

• The type of comments we appreciate:
  • References
    • For cross-check / avoid research duplication
  • Questions that you have a potential solution for
    • Something is wrong in the TM/TB and proposing an alternative
  • Questions that don’t have a solution
    • Needs to be raised to proofreader or client for confirmation, having done research on a topic

4. Leveraging the QA function inside Phrase

The Quality Assurance function in Phrase is extremely handy, and we require all translators (ourselves included) to run a QA analysis at the end of every stage to catch grammar and spelling errors, discrepancies with the TM/TB, and other issues. Read more about the QA inside Phrase at the link below.

Quality Assurance - QA (TMS)

5. Lastly, give it the Grammarly treatment before hitting “Complete”

Using either the free or paid versions of Grammarly helps review spelling, grammar, punctuation, clarity, engagement, and delivery mistakes in English texts, detects plagiarism, and suggests replacements for the identified errors.

• Helps catch many potential spelling/grammatical errors
• Another (albeit machine) brain for style — take with a grain of salt

6. Next Steps: Proofreading & Revision

• Proofreaders will take all of the above into account and proofread on the basis that it is completely grammatical and edited for accuracy and tone.
• Please note that we employ a Language Quality Assessment (LQA) rubric in accordance with the framework outlined at the Phrase website below:LQA Assessments (TMS)

We have customized this rubric (see "Honyaku Cloud LQA Rubric" below) for use between the lead translator and proofreader, who have been trained to maintain an objective standard according to the rubric.

• Once the proofread is complete, you will receive a notification and a link to revise your draft. Please refer to the comments and LQA inside Phrase for more information.
  • General comments, references, and suggestions will be located in Comments
  • Any mistakes, errors, or serious issues will be addressed in the LQA (indicated by a flag next to comments in the Phrase editor). Please refer to specific examples of LQA issues in the table below.

‍

‍

Reference to theHonyaku Cloud LQA Rubric

1. ACCURACY

• Accuracy
• Mistranslation
• Omission
• Over-translation

2. FLUENCY

• Grammar
• Inconsistency
• Punctuation
• Spelling

3. TERMINOLOGY

• Inconsistent use of terminology
• Inconsistent with termbase
• Terminology

4. STYLE

• Awkward
• Style

5. LOCALE CONVENTION

• Address format
• Date format
• Locale convention

6. DESIGN

• Design
• Length

7. Other

• Other