[HOLD] add language tag from cocina to sqs message #1428
Conversation
|
|
||
| def stt | ||
| @stt ||= Dor::TextExtraction::SpeechToText.new(cocina_object:) | ||
| end |
There was a problem hiding this comment.
cache this in the object so it can be used in our updated media method without re-instantiating it.
| @@ -45,7 +45,7 @@ | |||
|
|
|||
| context 'when the object has no files that can be STTed' do | |||
| let(:first_fileset_structural) { instance_double(Cocina::Models::FileSetStructural, contains: [text_file]) } | |||
| let(:second_fileset_structural) { instance_double(Cocina::Models::FileSetStructural, contains: [text_file, text_file]) } | |||
| let(:second_fileset_structural) { instance_double(Cocina::Models::FileSetStructural, contains: [text_file, text_file2]) } | |||
There was a problem hiding this comment.
small typo from a previous PR, turned out to have no impact on the test, but just fixing because
| @@ -71,6 +71,11 @@ def output_location | |||
| "#{job_id}/output" | |||
| end | |||
|
|
|||
| # look in resource structural metadata for a given filename and return it's cocina language (or nil if no language is set) | |||
| def language_tag(filename) | |||
| stt_files.find { |file| file.filename == filename }&.languageTag | |||
There was a problem hiding this comment.
question ... do we want the default value to pass to whisper to be nil or en ? this makes it nil, but we could make it en
There was a problem hiding this comment.
Preferably there would be no language parameter passed to the service if it is not known.
There was a problem hiding this comment.
Let me see if I can make that adjustment.
There was a problem hiding this comment.
changed - so ready for testing when the new container is ready
peetucket
force-pushed
the
1427-language-from-cocina
branch
from
0e2520d
to
62da03b
Compare
| @@ -71,6 +71,11 @@ def output_location | |||
| "#{job_id}/output" | |||
| end | |||
|
|
|||
| # look in resource structural metadata for a given filename and return it's cocina language (or nil if no language is set) | |||
| def language_tag(filename) | |||
| stt_files.find { |file| file.filename == filename }&.languageTag | |||
There was a problem hiding this comment.
Preferably there would be no language parameter passed to the service if it is not known.
Why was this change made? 🤔
Fixes #1427 - adds the
languageTagfrom cocina to the SQS message for the object so it can be passed through to Whisper. The languageTag can be set in pre-assembly via the file manifest.Note that this change must be coordinated with work done for sul-dlss/speech-to-text#51 . Since the format of the message has changed, the whisper container needs to know how to process and deal with this new format.
How was this change tested? 🤨
Specs