Hacker News Reader

Comment by alvatar | Hacker News Reader

Just in case you didn't know, and you have plans to actually use Google Translate for translating web pages into Spanish... please don't. I haven't seen a single sentence with more than 3-4 words translated properly from English to Spanish. Even the examples shown in the README make no sense in Spanish.

View on HN

Really? From my experience English to Spanish translation works pretty well in Google Translate. I've been using it a lot when I started to live in Spain and my Spanish skills were not too great yet: people were actually complementing me for how good my (written) Spanish is and were in huge shock when they met me in person and it turned out that I barely can formulate any sentence. True story.

It depends... most of what's translated is understandable for a native speaker, but it will be clear that it wasn't written by one. If you need to communicate with a Spanish speaker, it will work quite well. If you need to write a brochure or something formal... better get someone to translate for you.

Well, that's quite obvious, Google Translate is not perfect, e.g. overuses "Usted/Ustedes" or often gets lost when there's a lot of pronouns and missed "yo", "tu", "el", but:

a) with a little big of knowledge about the language and bit of editing you can overcome most of those stuff

b) even without editing it makes most of the stuff readable/understandable

c) works much better that competitive products (babblefish).

"my Spanish skills were not too great yet"

"From my experience English to Spanish translation works pretty well"

I think there's a connection.

You're wrong. "Not too great" doesn't mean I had completely no idea about the language, I was able to spot most obvious mistakes or strange constructions in the translation . Plus, 1,5 year later my Spanish is much better and I still think GT does amazingly good job for a machine translation. I also can compare it with how it works when translating from/to Spanish or English to/from Polish and really, English <-> Spanish is much, much better.

I'm fluent in English and Spanish. Yes, GT's English -> Spanish output is better than gibberish and you can often get the gist despite the glaring errors, especially if you know both languages.

But my point was, don't you think there's a bit of cognitive dissonance to somebody using machine translation because they don't speak a language very well, yet confidently asserting that the translations are good?

It might be like that but the point of my original post was that the correctness of the GT job was verified by native Spanish speakers (one might say it's a sort of a Turing test), not that my Spanish was bad so I thought translations are done good.

"Very cool gem!".to_spanish -> "Muy fresco joya!" This sounds like it would in English: "Fresh very jewel!". Makes no sense in any way, neither south american nor european Spanish. I'm sure there are many cases that it helps, and it actually did help me in many cases, even for learning German. But eventually, for writing something I always have to check the grammar with a native german, because grammar changes the whole meaning or sense. For a web page, I think it doesn't help reach the spanish community.

I assume Spanish people are quite polite.

They are but there is a world of difference between "don't worry, your Spanish is OK, I understand what you mean" and "I thought you know the language because your e-mails were written in perfect Spanish".

A few years back an ex-girlfriend of mine did an erasmus year over in Spain and I ended up making friends with a whole bunch of people that I could barely converse with. They've always been very impressed at far my Spanish has come in IM, thank you babel fish.

My dissertation research involved using (and, most importantly, evaluating) GT for certain medical text problems (going from EN->ES), and I found that its performance is highly domain-dependent (i.e., the quality of its translations depends a lot on what kind of input text you're feeding it). Furthermore, I found that whether or not it's "good enough" depends heavily on the user, specifically that user's level of proficiency in English.

GT is better than a lot of machine translation systems, but it still falls victim to a number of common problems that face such systems. Specifically, I've run into a lot of major word-sense-disambiguation issues, especially when working with the sort of short snippets of text that the gem in question uses as examples. Basically, that sort of short bit of text is a worst-case scenario for statistical machine translation, since there's so little context. Google generally does much better with longer runs of text than it does with short phrases.

Reply on news.ycombinator.com