The translation service of Google seems to do a good service in research projects. At least it has been mentioned repeatedly during the CLEF 2009 wokshop in Corfu. An alternative is Microsoft's translate web-service. Both can be accessed via a REST-ful web-service API.
With a few lines of code I managed to include both interfaces into a Java project. I started a batch translation of documents and summarize a few observations about the two:
- Speed: Google seems to respond slightly quicker, while Mircosoft's translation takes longer. I did not track the response times precisely, but there is definitly a difference.
- Text length: Google translates text up to a length of 5000 characters. Everything beyond that results in an error messages. Also Microsoft puts some restriction on the length of the text. I looked briefly at the API reference, but did not find a description of where the limit is. However, when trying to translate longer texts I got error messages.
- Reponse format: Google delivers the translations in the JSON format. Microsoft gives a plan text format. For both I am not entirely sure, whether there are options for configuring the results.
- Registration: Microsoft requires an application ID with each request. The ID can be generated online. Google encourages the use of an application key, but does not require it.
Resumee: Well, so far I did not compare the quality of the translations, neither did I look in all the possibilities of the APIs. However, Googles faster response and more clearly expressed limitations for text length seem to favour this solution.
Links:
Google Language API
Microsoft Translate API
No comments:
Post a Comment