by David Grunwald
ChatGPT can be used to review and proofread translations by feeding the source and target into the tool.
In the past, I have written some blog posts on using Google Translate to review translations using back translation. Here is one of these posts.
The concept of back translation is hardly new.
A “back-translation” is a translation of a translated text back into the language of the original text, made without reference to the original text. Comparison of a back-translation with the original text is sometimes used as a check on the accuracy of the original translation, much as the accuracy of a mathematical operation is sometimes checked by reversing the operation.
Even if you don’t have the budget or time to do a human back translation, it is a good idea to stick the translation into Google (or another MT) and then compare the result with the original file. It should be noted that machine translation is far from perfect. Which means that you may see some discrepancies between the source document and the back translation. But even if this is the case, the back translation will give you a feel for the general quality of the translation. The back translation will also allow you to check for missing translations and other errors that may occur in a human translation process.
The Future Trend: Using AI as your reviewer
I tried this a few times and was amazed at the result. Take a translation that you received from a professional translator and insert that and the source text into ChatGPT. Tell it something like: “please compare this English to Spanish translation” and then copy-paste in the source and target. Wait a few moments and get the result.
(Please note that I used ChatGPT 4 which is a paid subscription. I am not sure if this feature is supported in earlier versions.
Here are some samples from real jobs:
Example 1: English to Chinese translation
The English to Chinese translation of the CSR (Corporate Social Responsibility) Significance Survey is generally well done, accurately conveying the intent and content of the original English text. Here are some observations and suggestions:
- Consistency in Terminology: The translation consistently uses the term “企业社会责任” for “Corporate Social Responsibility,” which is appropriate.
- Translation of Survey Introduction: The translation of the survey introduction is clear and conveys the original message effectively. It properly invites participants to the survey and explains its purpose.
- Survey Questions Translation: The survey questions are translated accurately. The options provided (e.g., for levels of confidence or agreement) are appropriately rendered in Chinese.
- Translation of Specific Phrases:
- “Corporate Social Responsibility Manager” is translated as “企业社会责任经理,” which is correct.
- The specific focus areas of URUS’ CSR efforts, like ending hunger and protecting the planet, are well translated, maintaining the original intent and detail.
- Clarity and Fluency: The translation reads fluently in Chinese, and the sentence structures are suitable for a formal survey context.
- Survey Closing and Reminder Emails: These sections are translated effectively, with a polite and formal tone that encourages participation.
One minor suggestion is to ensure that any specific names or terminologies unique to the company or industry (such as “URUS,” “AgSource,” “Alta,” etc.) are kept consistent throughout the translation. This consistency appears to be maintained well in your translation.
Overall, the translation is of high quality, accurately conveying the information and tone of the original English text in a way that is appropriate for a Chinese-speaking audience.
In this case, ChatGPT gave the translation a rave review.
Example 2: Dutch to English translation
The translation from Dutch to English is generally accurate and conveys the original content effectively. However, there are some minor issues and areas for improvement in terms of language fluency and precision. Below, I will highlight these areas with suggested revisions:
- Section 1: Registration on the transplant waiting list
- Original: “…we must pass on all necessary information to Eurotransplant.”
- Suggestion: “…we are required to transmit all necessary information to Eurotransplant.”
- Section 2: Your medical data, photographic and audiovisual recordings
- Original: “We may use your medical data, images and recordings of your diagnosis, treatment and follow-up that form part of your medical file for the following purposes:”
- Suggestion: “We may utilize your medical data, images, and recordings related to your diagnosis, treatment, and follow-up, which are part of your medical record, for the following purposes:”
- General Note on Data Protection Regulation
- Original: “European General Data Protection Regulation (AVG).”
- Suggestion: “European General Data Protection Regulation (GDPR).”
- Note: “AVG” is the Dutch acronym for GDPR. For an English audience, using “GDPR” is more recognizable.
- Section 3: The use of residual material
- Original: “…we may be able to use residual tissue or body material (including blood, urine, biopsies, etc.)…”
- Suggestion: “…we may utilize residual tissue or bodily materials (including blood, urine, biopsies, etc.)…”
- Consent Sections (A, B, C, D, E)
- Original: “I have been informed that my general practitioner and other specialists involved in my medical treatment will be informed about my transplant.”
- Suggestion: “I am aware that my general practitioner and other specialists involved in my care will be notified about my transplant.”
- Section E: Patient’s resistance to use of residual tissue
- Original: “I do not agree that the collected biological samples (tissue, blood, urine, genetic material), which are not necessary for my diagnosis or medical treatment, may be used (residual material) and stored for 20 years for scientific research, on provided that this is limited to purposes as described in this information brochure.”
- Suggestion: “I do not consent to the use of collected biological samples (tissue, blood, urine, genetic material), which are not required for my diagnosis or medical treatment (referred to as residual material), being used and stored for 20 years for scientific research, provided that such use is limited to the purposes as described in this information brochure.”
Overall, the translation is quite good, but attention to these details can enhance its clarity and precision.
In this example, we sent the comments to the translator who revised the work based on ChatGPT’s comments.
This is without a doubt the future of the translation industry. Once AI can be trained to automatically translate and review text, the quality of machine translation will improve to a level of human translation quality.