更懂电商语境!阿里推翻译大模型,实测比GPT 4o、Google、DeepL好

作者子川

来源AI先锋官


AI翻译领域居然冲出一匹黑马!效果比Google、GPT-4o、DeepL还要好?

让我们来看看是怎么个事!

近日,阿里国际推出最新研发的Marco翻译大模型。

目前已在阿里国际AI官网Aidge上发布,向全球用户开放使用。

一经推出,备受好评!

数据层面上:

Marco翻译大模型在BLEU自动评测指标上,领先于市面上的其他翻译产品,如Google翻译、DeepL、GPT-4等,成为行业第一。

图片


比较的特点是,以往很多翻译产品,会把一些电商用语翻译的奇奇怪怪!

比如把“你的宝贝正在路上”,翻译成“Your baby is on the way”;把“尊嘟喜欢”翻译为“Zundu likes it”。

图片


而在Marco大模型中,这些都能很地道的表达出原本的意思—— 把“你的宝贝在路上”,翻译成“Your package is on the way”;把“尊嘟喜欢,”翻译成“l like it"。

的确,在电商语境中,很多语言的表达对翻译产品是一个极大的挑战。

出于这个好奇心,小编忍不住实测了下。

小编用传统翻译软件、AI翻译产品,及Marco模型进行对比,分别是谷歌翻译、DeepL翻译、GPT-4o和Marco模型。


第一弹,首先我们派出一个简单的词“光腿神器”。

谷歌翻译:Bare-legged artifact

Deepl翻译:Bare Legs Magic

Marco模型:The magic of bare legs

似乎都翻译的挺好的,不过像Deepl翻译和Marco翻译都翻译成光腿的魔法,只有GPT-4o的翻译带有裤子这个词,似乎效果更好一点。

第二弹上上难度

图片


谷歌翻译:Shore holy water cup

Deepl翻译:The Holy Water Chalice of Disembarkation

GPT-4o:Land Ashore Holy Water Cup

Marco模型:Shore Saint Water Cup

额。。。。

貌似大差不差的都翻译成“岸上神圣的水杯”。

好家伙,全军覆没,都没有表达出上岸的具体寓意!只能说中文博大精深

第三弹,比较简单的句子,“真的是可盐可甜,穿着显瘦,拍照简直不要好看呀,太上镜了”!

谷歌翻译:It's really salty and sweet, dressed thinly, it's not good to look good in photos, it's too photogenic

Deepl翻译Really can be salt and sweet, wearing thin, photos simply do not look good ah, too photogenic!

GPT-4o: It's truly versatile and charming, making you look slim when worn. It looks absolutely stunning in photos, extremely photogenic.

Marco模型:Really sweet and cute, looks slim when wearing it, looks amazing in photos, sophotogenic

谷歌和DeepL都把“盐”翻译成“salty”,GPT-4o也没有把“可盐可甜”翻译出来。只有Marco模型,翻译成“Really sweet and cute”。胜出!


第四弹,“泰美辣!你好好看啊”

谷歌翻译是:Thai and spicy! You look so good.

Deepl翻译:Tammy Spice! You look great.

GPT-4o: Thai beautiful and spicy! You look amazing!

Marco模型:Beautiful! You look so good!

谷歌和Deepl一样,翻译成“泰”和“辣”,真的笑不活了。只有Marco模型翻译准确,GPT-4o翻译准确一半,“漂亮又辣”。

第五弹,“焦糖苔原风穿搭”

图片

谷歌翻译是:Caramel tundra style 

Deepl翻译:Caramel Tundra Style Caramel tundra style
GPT-4o:Caramel tundra style outfit
Marco模型:caramel tundra style outfit

貌似都没有翻译出来,谷歌和Deepl只翻译出焦糖风格,GPT-4o和Marc效果会更好一点,不过也翻译不全,最后的穿搭两个字翻译成了衣服。


总的来说,在中文语境中,似乎Marco模型效果会更好一点,而其他的模型多少还是会有些偏差。


 .END.

往期文章回顾