标签归档 上海后花园2021

How popular is ChatGPT recently? Only one week after its launch, it has exceeded 200 million users, 89% of American students use ChatGPT to do their homework, and some students get full marks for the papers written by ChatGPT. What’s even more surprising is that when you talk to ChatGPT, you enter Zhihu style-a novel that starts with "You don’t really think"? ChatGPT fed back a short story in a few seconds.

Although the story is very old-fashioned, the overall structure is complete. In the field of translation, ChatGPT is omnipotent. You can ask him to act as an interpreter and give him some requirements.

ChatGPT’s translation gives you the feeling that it is no longer a cold machine, but a caring assistant, which can generate a translation that meets your requirements better and even help you do some inspection, interpretation and polishing. For example, you can ask ChatGPT to act as an interpreter and make some demands:

You can also ask ChatGPT to act as a quality supervisor and help us check the low-level mistakes in the translation:

Of course, ChatGPT can even help to polish the machine-translated text with a strong "machine-turned flavor":

After the overall experience, it is found that the flavor of "machine translation" is no longer so strong whether it is text or translation generated by ChatGPT. However, it is worth mentioning that there are still some differences between ChatGPT and mainstream commercial rollover systems.

ChatGPT is a dialogue model. In order to ensure the diversity of generated answers, each result is "Sample" from the model, while most commercial machine translation engines are decoded by "beam search", which is the optimal solution of the model.

The traditional evaluation of machine translation is mostly from the sentence level, but obviously ChatGPT should have a stronger understanding of documents and contexts. If the translation task at the document level is used for evaluation, ChatGPT should have a better effect on context consistency and texts that rely on contextual information for understanding.

With the increasing number of products such as machine translation and ChatGPT, many people are beginning to think that machine translation will eventually replace human translation. Yayan Translation has written several articles about machine translation before, and the unified answer is that machine translation can’t completely replace human translation, and human translation can’t "dry up" machine translation.

For such an era of artificial intelligence, the emergence of products such as machine translation and ChatGPT should not be rejected because of fear, but should embrace them and understand them, making them favorable tools to improve their work efficiency.

At the end of this article, I want to tell you that in the foreseeable future, artificial intelligence will be an important stage of development and machine translation will also become one of the important projects. But in any case, human translation will not be forced to "be laid off" because of this, and the demand for human translation in the market will always be just needed.