Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
Is it a good idea? ;-)
Is it legal to do so? Is it legal to release such a dataset to public?
Say I have a language X for which I want to create a dataset for translation to/from English, for which I have no online sources to scrape out the data from. But our giant Google Translate somehow has the support for language X.
Can we just dump the translations from English datasets for my language X using Google Translate API and create a rough dataset to start with?
It depends. Google Translate works very good with some pairs of languages, and not good on others. Based on my personal experience, translating from North-European languages (Dutch, Danish, Swedish) to English worked almost perfectly, while English-Italian translation lead to bad results.
You can find a Spanish-English dataset here, this is a very official source.
Additionally, it seems DeepL is a great German <-> English online translator around.
In terms of using the data for initial training of a machine learning model, it is a very good idea, because Google Translate produces exceptional results, which will give yo very good ideas about how to go about building and improving your machine learning model and save you time in the long run.
As for releasing the dataset to the public, Google has a website with information and from my understanding they just want attribution.
However, if you are dumping a dataset larger than 500k characters, because that is the tier of free usage per month, it seems to be more likely to be a violation of their Acceptable Use Policy that does not allow us to disable, interfere with or circumvent any aspect of the Services.
From my reading of the website Google does not provide any direct answer. But they even have a special form for asking about publishing Google Cloud Translation so I encourage using that form to ask them directly.
I don't think it's a very good idea, because it boils down to training a model based on the predictions from another model. So no matter how good your model would be, the best it could achieve is the performance obtained by Google Translate. In other words, it will learn the errors made by Google Translate and add its own in the mix. But I guess it's better than nothing, if you don't have any other choice.
I don't know about the legal implications, but last time I checked one had to pay in order to perform a high number of translation queries automatically (beyond a certain number allowed for free).
