DLME Internationalization overview

jacobthill commented 5 years ago

Description

This is a significant feature with many moving parts. This ticket summarizes all of the steps at a high level. The individual steps have been broken out into smaller tickets and linked here. I have also created a milestone with the name 'DLME Internationalization' in each repo and assigned the smaller tickets to that milestone.

Requirements:

translate Spotlight UI labels into Arabic (data manager) https://github.com/sul-dlss/dlme/issues/541
translate text (prose and headings) into Arabic (data manager) https://github.com/sul-dlss/dlme/issues/542
translate metadata field values for cho_edm_type, cho_has_type, date, holding institution, source institution, and language (if possible) (data manager & Infrastructure team) https://github.com/sul-dlss/dlme-traject/issues/110, https://github.com/sul-dlss/dlme-transform/issues/172, https://github.com/sul-dlss/dlme-transform/issues/173, https://github.com/sul-dlss/dlme-transform/issues/171, https://github.com/sul-dlss/dlme-transform/issues/174, https://github.com/sul-dlss/dlme-transform/issues/155
redesign the DLME intermediate representation (IR) (Infrastructure team) https://github.com/sul-dlss/dlme-transform/issues/173
update each config file so that they map to the updated IR (data manager & Infrastructure team) https://github.com/sul-dlss/dlme-traject/issues/112
design work (Access team) https://github.com/sul-dlss/dlme/issues/544

Examples of multi-lingual metadata records:

QNL has separate records for Arabic and English. The ID is the same except the ar or en endings. The English can be mapped to the Arabic by matching id.
AUB has mixed records. Some fields are always in Arabic; others are in English (numbers mainly)
AUC has mostly English with a few Arabic fields.
SSM has modern Turkish
Princeton has transliterated Arabic and English for some fields (e.g. description)
Harvard has multiple languages including French, English, and transliterated Arabic
BnF has French

jacobthill commented 5 years ago

@jcoyne @justinlittman How does this revision look? @ggeisler has the information he needs to begin thinking through the design. Are steps 7-10 doable? That is the area that we need to pin down. Basically, I need your help thinking through the transformation issue (e.g. how difficult would it be to write a macro that would enable me to specify the language of each metadata field with a literal?) and the use of the IR to determine display direction and indexing rules. We may encounter other issues like periods, parenthesis, and quotation marks appearing in the wrong place with Arabic text but we will have to take those as they come. The tech lead from the Access team should probably weigh in on this as well @anarchivist?

justinlittman commented 5 years ago

@jacobthill Thanks for the details. Would also be helpful if you could attach screenshots to make sure that we understand the parts of the app you are referring to.

jacobthill commented 5 years ago

@justinlittman which numbers do yo need screenshots for?

justinlittman commented 5 years ago

1, 3, 5 -- basically anything that is rendered in the app.

ggeisler commented 5 years ago

I think @jacobthill knows this from our recent discussion, but just to be clear to all involved, items 1 through 4 should not require any additional application work, just translation gathering from Jacob and then using Spotlight's existing multilingual features to enter translations via the Spotlight UI.

jcoyne commented 5 years ago

5 is the one I'm most concerned with. What does it look like if we have these records in the system:

(R1) Arabic and English (R2) English only (R3) English and Persian (R4) French only

mejackreed commented 5 years ago

For the IR, perhaps look at using the IIIF approach to this: https://iiif.io/api/presentation/3.0/#44-language-of-property-values

ggeisler commented 5 years ago

Closing since all of the issues seem to be covered by more actionable tickets, such as #544, #601, and #589.

sul-dlss / dlme