This is intended to be a discussion and idea formation.
Enable multilingual support in Phileas. Each filter strategy (for each filter) should have a list of language codes. If the text being processed belongs to one of those languages, the filter strategy will be applied.
Ideally, the list of language codes would be associated with the filters and not the filter strategies. This is better for the PhEye filter because it would only need to be called for that particular language. If the list of languages is on the filter strategies, each filter has to be applied and that would mean calling each PhEye service even if the model is not for that language.
The problem with the ideal approach is the filters in the policy are one-to-one, meaning filters that could apply to multiple languages are limited to a single instance. Perhaps just the filters that could apply to multiple languages should be able to be lists with a languages property.
Phileas would perform language detection if a language code is not provided, and then apply the appropriate filters to the text.
This is intended to be a discussion and idea formation.
Enable multilingual support in Phileas. Each filter strategy (for each filter) should have a list of language codes. If the text being processed belongs to one of those languages, the filter strategy will be applied.
Ideally, the list of language codes would be associated with the filters and not the filter strategies. This is better for the PhEye filter because it would only need to be called for that particular language. If the list of languages is on the filter strategies, each filter has to be applied and that would mean calling each PhEye service even if the model is not for that language.
The problem with the ideal approach is the filters in the policy are one-to-one, meaning filters that could apply to multiple languages are limited to a single instance. Perhaps just the filters that could apply to multiple languages should be able to be lists with a
languagesproperty.Phileas would perform language detection if a language code is not provided, and then apply the appropriate filters to the text.