Get hands-on experience with 20+ free Google Cloud products and $300 in free credit for new customers.

Cloud translations using glossaries trigger double values

MLC
Bronze 1
Bronze 1

We're encountering an issue where we use glossaries for translation and whenever these glossaries trigger on a 'from' value, the 'to' value appears twice consecutively in the translated output.

For example:

Input(Dutch): Plafond vuren 
Output(English): Spruce Spruce
Dictionary: Vuren --> Spruce

As you can see the output value triggers twice instead of once and the word 'plafond' (ceiling in dutch) doesn't trigger at all. Any help with this issue would be very much appreciated!

0 2 254
2 REPLIES 2


@MLC wrote:

We're encountering an issue where we use glossaries for translation and whenever these glossaries trigger on a 'from' value, the 'to' value appears twice consecutively in the translated output.

For example:

Input(Dutch): Plafond vuren 
Output(English): Spruce Spruce
Dictionary: Vuren --> Spruce

As you can see the output value triggers twice instead of once and the word 'plafond' (ceiling in dutch) doesn't trigger at all. Any help with this issue would be very much appreciated!


You can review your translation pipeline's logic for processing glossary entries. Ensure that each entry is correctly matched only once and that the translation output doesn't suffer from duplication. Also, investigate why certain words like "plafond" might not be triggering the glossary at all; it could be due to case sensitivity, formatting, or other factors in your processing.

Thanks for your insight! 

What stands out is that the word 'vuren' changes position in the text where it goes wrong. If the word 'vuren' stays in the same position as in the original text, it seems to translate correctly. 

In our code the following is returned by Google as we were trying to understand what exactly is being output:

translations { translated_text: "Spruce ceiling for Allure modular outdoor accommodation" } glossary_translations { translated_text: "Spruce spruce for Allure modular garden building" glossary_config { glossary: "projects/685601193296/locations/us-central1/glossaries/gld-650428a0-d8ba-4f04-9bd1-9217da88a82f" }

 It appears that Google Translate first translates the text to 'Spruce ceiling for Allure modular outdoor accommodation'. Then, it seems to do something with the glossary translation, resulting in the final translation 'Spruce spruce for Allure modular garden building' displayed in the application. What exactly is done on the side of Google Translate is unclear to me, is this something you know more about?