Comparing the accuracy of machine translation tools has been a topic of discussion for quite some time now. In a recent experiment conducted by a writer, the Portuguese language was used to compare the outputs generated by Google Translate and ChatGPT.
The writer wanted to translate their blog post into Portuguese and decided to use this opportunity to compare how the two translation tools fared against each other. The primary aim was to find out which tool offered faster and more accurate translation and to use the content generated for another article.
It was observed that the original text had to be adjusted to fit Google Translate’s limitations of 80 characters per line, leading to sentence breaks. In contrast, the text could be fed as is into ChatGPT without any preparation.
The experiment was then carried out to compare the translation output from both tools.
While there were some stark differences in output, both tools registered numerous hits and misses. Still, ChatGPT seemed to perform better overall. For example, the tool translated ‘working’ correctly as a gerund, while Google failed to comprehend ‘singular they.’ Additionally, ChatGPT appeared more flexible, allowing the writer to alter the prompt and change the tone of the text to a more formal one.
The writer emphasized that both tools provided an excellent starting point for a translation; however, a final review by a human translator was highly recommended.
The author concluded that ChatGPT appeared to get more context from the overall text, and it required no preparation or text adjustment. Therefore, if the writer had to pick one tool, they would choose ChatGPT.
Ultimately, the experiment highlights the fact that machine translation technology has not yet reached perfection, and it is still necessary to be cautious when dealing with translated content.
