Hello @ywmo
Thanks for reaching out to us, yes, it is possible to use a dictionary-based approach to handle inflection in Azure Custom Translator. One way to do this is to create a glossary that includes the base form of each word, along with its inflected forms and their corresponding translations.
For example, in your case, you could create a glossary that includes the base form of each Korean word, along with its inflected forms and their corresponding translations in Polish. You could then use this glossary to look up the appropriate translation for each inflected form of a word.
Here is an example of what the glossary might look like:
<glossary version="3.0" languageFrom="ko" languageTo="pl">
<glossaryEntry>
<source>발화</source>
<translations>
<translation>pożar</translation>
<translation>ogień</translation>
<translation>palenie</translation>
<translation>strzelanie</translation>
<translation>strzał</translation>
<translation>odpalenie</translation>
<translation>rozpalenie</translation>
</translations>
</glossaryEntry>
...
</glossary>
In this example, the glossary includes the base form of the Korean word "발화" (which means "fire"), along with its inflected forms and their corresponding translations in Polish. For example, the inflected form "발화하는" (which means "firing") is translated to "strzelanie" in Polish.
When training your custom translation model in Azure Custom Translator, you can upload this glossary as a resource. The glossary will then be used to look up the appropriate translation for each inflected form of a word during the translation process.
Keep in mind that this approach may not be perfect, as inflection can be complex and there may be cases where the appropriate translation is not included in the glossary. However, it can be a useful starting point for handling inflection in translation. I hope this helps!
Regards,
Yutong
-Please kindly accept the answer if you feel helpful to support the community, thank a lot.