let the user select speech to text for media objects

app/controllers/batch_contexts_controller.rb

@@ -46,6 +46,7 @@ def batch_contexts_params
           .permit(:project_name, :content_structure, :staging_style_symlink,
                   :processing_configuration, :staging_location, :all_files_public,
                   :run_ocr, :manually_corrected_ocr, :ocr_available,
+                  :run_stt, :manually_corrected_stt, :stt_available,
                   :using_file_manifest, job_runs_attributes: [:job_type], ocr_languages: [])
           .merge(user: current_user)
   end

app/javascript/controllers/caption_controller.js

@@ -1,9 +1,9 @@
 import { Controller } from '@hotwired/stimulus'
 
 export default class extends Controller {
-  static targets = ['contentStructure', 'ocrSettings', 'ocrAvailable', 'sttSettings',
-    'manuallyCorrectedOcr', 'runOcr', 'ocrLanguages', 'ocrDropdown', 'runOcrDocumentNotes',
-    'runOcrImageNotes', 'selectedLanguages', 'languageWarning', 'dropdownContent', 'ocrLanguageWrapper']
+  static targets = ['contentStructure', 'ocrSettings', 'ocrAvailable', 'sttSettings', 'sttAvailable', 'runStt',
+    'manuallyCorrectedOcr', 'manuallyCorrectedStt', 'runOcr', 'ocrLanguages', 'ocrDropdown', 'runOcrDocumentNotes',
+    'runOcrImageNotes', 'runSttNotes', 'selectedLanguages', 'languageWarning', 'dropdownContent', 'ocrLanguageWrapper']
 
   static values = { languages: Array }
 
@@ -54,6 +54,10 @@ export default class extends Controller {
     return 'Do the PDF documents comply with accessibility standards? More info: <a target=_blank href="https://uit.stanford.edu/accessibility/guides/pdf">PDF Accessibility</a>.'
   }
 
+  labelMediaManuallyCorrected () {
+    return 'Have the caption/transcript files been corrected to comply with accessibility standards? More info: <a target=_blank href="https://www.w3.org/WAI/media/av/captions/">Captions</a> and <a target=_blank href="https://www.w3.org/WAI/media/av/transcripts/">Transcripts</a>.'
+  }
+
   labelRunOcr () {
     return `Do you want to auto-generate OCR files for the ${this.ocrFileTypeLabel()}?`
   }
@@ -65,6 +69,7 @@ export default class extends Controller {
   contentStructureChanged () {
     // Hide the OCR and speech to text settings by default; we will show them if the content structure allows them to
     this.ocrSettingsTarget.hidden = true
+    this.sttSettingsTarget.hidden = true
 
     if (this.ocrAvailable()) {
       this.showOcrControls()
@@ -102,6 +107,22 @@ export default class extends Controller {
   // Show the Speech to text settings and controls
   showSttControls () {
     this.sttSettingsTarget.hidden = false
+
+    this.manuallyCorrectedSttTarget.querySelector('legend').innerHTML = this.labelMediaManuallyCorrected()
+  }
+
+  // if the user indicates they have speech to text available, show/hide the manually corrected and run stt options (for media)
+  sttAvailableChanged () {
+    const sttAvailable = this.sttAvailableTarget.querySelector('input[type="radio"]:checked').value === 'true'
+    this.manuallyCorrectedSttTarget.hidden = !sttAvailable
+    this.runSttTarget.hidden = sttAvailable
+    this.runSttChanged()
+  }
+
+  // if the user indicates they want to run SDR speech to text, show any relevant notes/warnings
+  runSttChanged () {
+    const runstt = this.runSttTarget.querySelector('input[type="radio"]:checked').value === 'true'
+    this.runSttNotesTarget.hidden = !runstt
   }
 
   // if the user indicates they have ocr available, show/hide the manually corrected and run OCR option (for images/books)

app/lib/pre_assembly/batch.rb

@@ -24,6 +24,7 @@ class Batch
              :project_name,
              :staging_style_symlink,
              :ocr_available,
+             :stt_available,
              to: :batch_context
 
     def initialize(job_run, file_manifest: nil)

app/lib/pre_assembly/digital_object.rb

@@ -12,6 +12,7 @@ class DigitalObject
              :processing_configuration,
              :content_structure,
              :ocr_available,
+             :stt_available,
              :project_name,
              :file_manifest,
              to: :batch
@@ -121,6 +122,7 @@ def build_structural
         build_from_staging_location(objects: object_files.sort,
                                     processing_configuration:,
                                     ocr_available:,
+                                    stt_available:,
                                     reading_order:)
       end
     end
@@ -161,13 +163,14 @@ def update_structural_metadata
       object_client.update(params: updated_cocina)
     end
 
-    def build_from_staging_location(objects:, processing_configuration:, reading_order:, ocr_available:)
-      filesets = FromStagingLocation::FileSetBuilder.build(processing_configuration:, ocr_available:, objects:, style: content_md_creation_style)
+    def build_from_staging_location(objects:, processing_configuration:, reading_order:, ocr_available:, stt_available:)
+      filesets = FromStagingLocation::FileSetBuilder.build(processing_configuration:, ocr_available:, stt_available:, objects:, style: content_md_creation_style)
       FromStagingLocation::StructuralBuilder.build(cocina_dro: existing_cocina_object,
                                                    filesets:,
                                                    all_files_public: batch.batch_context.all_files_public?,
                                                    reading_order:,
-                                                   manually_corrected_ocr: batch.batch_context.manually_corrected_ocr)
+                                                   manually_corrected_ocr: batch.batch_context.manually_corrected_ocr,
+                                                   manually_corrected_stt: batch.batch_context.manually_corrected_stt)
     end
 
     # The reading order for books is determined by what the user set when registering the object.

app/lib/pre_assembly/from_staging_location/file.rb

@@ -37,10 +37,17 @@ class File
         'application/xml' => { preserve: 'yes', shelve: 'yes', publish: 'yes', role: 'transcription', corrected_for_accessibility: true }
       }.freeze
 
+      # TODO: which files represent the speech to text files when provided and thus need to the transcription role?
+      # if the user tells us they are providing speech to text, we will set the transcription role and possibly also corrected_for_accessibility
+      ATTRIBUTES_FOR_TYPE_WITH_STT = {
+        'application/xml' => { preserve: 'yes', shelve: 'yes', publish: 'yes', role: 'transcription', corrected_for_accessibility: true }
+      }.freeze
+
       # @param [Assembly::ObjectFile] file
-      def initialize(file:, ocr_available:)
+      def initialize(file:, ocr_available:, stt_available:)
         @file = file
         @ocr_available = ocr_available
+        @stt_available = stt_available
       end
 
       delegate :sha1, :md5, :provider_md5, :mimetype, :filesize, :relative_path, to: :file
@@ -51,10 +58,11 @@ def file_attributes
 
       private
 
-      attr_reader :file, :ocr_available
+      attr_reader :file, :ocr_available, :stt_available
 
       def file_attributes_for_mimetype(mimetype)
         return ATTRIBUTES_FOR_TYPE_WITH_OCR[mimetype] if ocr_available && ATTRIBUTES_FOR_TYPE_WITH_OCR.key?(mimetype)
+        return ATTRIBUTES_FOR_TYPE_WITH_STT[mimetype] if stt_available && ATTRIBUTES_FOR_TYPE_WITH_STT.key?(mimetype)
 
         ATTRIBUTES_FOR_TYPE[mimetype] || ATTRIBUTES_FOR_TYPE['default']
       end

app/lib/pre_assembly/from_staging_location/file_set.rb

@@ -9,10 +9,12 @@ class FileSet
       # @param [Array<Assembly::ObjectFile>] resource_files
       # @param [Symbol] style one of: :simple_image, :file, :simple_book, :book_as_image, :book_with_pdf, :map, :geo, or :'3d'
       # @param [Boolean] ocr_available
-      def initialize(resource_files:, style:, ocr_available:)
+      # @param [Boolean] stt_available
+      def initialize(resource_files:, style:, ocr_available:, stt_available:)
         @resource_files = resource_files
         @style = style
         @ocr_available = ocr_available
+        @stt_available = stt_available
       end
 
       # otherwise look at the style to determine the resource_type_description
@@ -25,12 +27,12 @@ def label_from_file(default:)
       end
 
       def files
-        resource_files.map { |file| File.new(file:, ocr_available:) }
+        resource_files.map { |file| File.new(file:, ocr_available:, stt_available:) }
       end
 
       private
 
-      attr_reader :resource_files, :style, :ocr_available
+      attr_reader :resource_files, :style, :ocr_available, :stt_available
 
       # rubocop:disable Metrics/AbcSize
       # rubocop:disable Metrics/CyclomaticComplexity

app/lib/pre_assembly/from_staging_location/file_set_builder.rb

@@ -7,15 +7,16 @@ class FileSetBuilder
       # @param [Symbol] processing_configuration one of: :default or :filename
       # @param [Array<Assembly::ObjectFile>] objects
       # @param [Symbol] style one of: :simple_image, :file, :simple_book, :book_as_image, :book_with_pdf, :map, :geo, or :'3d'
-      def self.build(processing_configuration:, objects:, style:, ocr_available:)
-        new(processing_configuration:, objects:, style:, ocr_available:).build
+      def self.build(processing_configuration:, objects:, style:, ocr_available:, stt_available:)
+        new(processing_configuration:, objects:, style:, ocr_available:, stt_available:).build
       end
 
-      def initialize(processing_configuration:, objects:, style:, ocr_available:)
+      def initialize(processing_configuration:, objects:, style:, ocr_available:, stt_available:)
         @processing_configuration = processing_configuration.to_sym
         @objects = objects
         @style = style
         @ocr_available = ocr_available
+        @stt_available = stt_available
       end
 
       # @return [Array<FileSet>] a list of filesets in the object
@@ -24,7 +25,7 @@ def build
 
         case processing_configuration
         when :default # one resource per object
-          objects.collect { |obj| FileSet.new(resource_files: [obj], style:, ocr_available:) }
+          objects.collect { |obj| FileSet.new(resource_files: [obj], style:, ocr_available:, stt_available:) }
         when :filename, :filename_with_ocr # one resource per distinct filename (excluding extension)
           build_for_filename
         else
@@ -34,7 +35,7 @@ def build
 
       private
 
-      attr_reader :processing_configuration, :objects, :style
+      attr_reader :processing_configuration, :objects, :style, :stt_available
 
       # until the new OCR settings are available, we have to look in the processing configuration
       def ocr_available
@@ -52,7 +53,8 @@ def build_for_filename
         distinct_filenames.map do |distinct_filename|
           FileSet.new(resource_files: objects.collect { |obj| obj if obj.filename_without_ext == distinct_filename }.compact,
                       style:,
-                      ocr_available:)
+                      ocr_available:,
+                      stt_available:)
         end
       end
     end

app/lib/pre_assembly/from_staging_location/structural_builder.rb

@@ -10,25 +10,32 @@ class StructuralBuilder
       # @param [String] reading_order
       # @param [Boolean] all_files_public
       # @param [Boolean] manually_corrected_ocr set by user when creating the job
-      def self.build(filesets:, cocina_dro:, all_files_public:, reading_order: nil, manually_corrected_ocr: false, ocr_available: false)
+      # @param [Boolean] manually_corrected_stt set by user when creating the job
+      # @param [Boolean] ocr_available set by user when creating the job
+      # @param [Boolean] stt_available set by user when creating the job
+      def self.build(filesets:, cocina_dro:, all_files_public:, reading_order: nil, manually_corrected_ocr: false, ocr_available: false, manually_corrected_stt: false, stt_available: false)
         new(filesets:,
             cocina_dro:,
             reading_order:,
             all_files_public:,
             manually_corrected_ocr:,
-            ocr_available:).build
+            ocr_available:,
+            stt_available:,
+            manually_corrected_stt:).build
       end
 
-      def initialize(filesets:, cocina_dro:, all_files_public:, reading_order:, manually_corrected_ocr:, ocr_available:)
+      def initialize(filesets:, cocina_dro:, all_files_public:, reading_order:, manually_corrected_ocr:, ocr_available:, manually_corrected_stt:, stt_available:)
         @filesets = filesets
         @cocina_dro = cocina_dro
         @reading_order = reading_order
         @all_files_public = all_files_public
         @manually_corrected_ocr = manually_corrected_ocr
         @ocr_available = ocr_available
+        @manually_corrected_stt = manually_corrected_stt
+        @stt_available = stt_available
       end
       # rubocop:enable Metrics/ParameterLists
-      attr_reader :filesets, :cocina_dro, :reading_order, :all_files_public, :manually_corrected_ocr, :ocr_available
+      attr_reader :filesets, :cocina_dro, :reading_order, :all_files_public, :manually_corrected_ocr, :ocr_available, :manually_corrected_stt, :stt_available
 
       # rubocop:disable Metrics/AbcSize
       # rubocop:disable Metrics/MethodLength

app/models/batch_context.rb

@@ -180,7 +180,7 @@ def verify_output_dir_no_exists
   def verify_file_manifest_selected_for_media
     return unless content_structure == 'media' && !using_file_manifest
 
-    errors.add(:content_structure, 'requires a file manifest.  Please select the checkbox and ensure a file manifest is present.')
+    errors.add(:content_structure, 'requires a file manifest.  Please indicate you have a file manifest and ensure a file manifest is present.')
   end
 
   def verify_staging_location

app/services/start_accession.rb

@@ -3,17 +3,21 @@
 # Starts the accession workflow by calling dor-services-app
 # See https://sul-dlss.github.io/dor-services-app/#operation/objects#accession
 class StartAccession
+  # rubocop:disable Metrics/AbcSize
   def self.run(druid:, batch_context:, workflow: nil)
     object_client = Dor::Services::Client.object(druid)
 
     workflow_context = {}
-    workflow_context[:runOCR] = batch_context.run_ocr
-    workflow_context[:manuallyCorrectedOCR] = batch_context.manually_corrected_ocr
+    workflow_context[:runOCR] = batch_context.run_ocr if batch_context.run_ocr
+    workflow_context[:manuallyCorrectedOCR] = batch_context.manually_corrected_ocr if batch_context.manually_corrected_ocr
+    workflow_context[:runSpeechToText] = batch_context.run_stt if batch_context.run_stt
+    workflow_context[:manuallyCorrectedStt] = batch_context.manually_corrected_stt if batch_context.manually_corrected_stt
     workflow_context[:ocrLanguages] = batch_context.ocr_languages unless batch_context.ocr_languages.empty?
 
     params = { description: 'pre-assembly re-accession', opening_user_name: batch_context.user.sunet_id, workflow: }
     params[:context] = workflow_context if workflow_context.present?
 
     object_client.accession.start(params)
   end
+  # rubocop:enable Metrics/AbcSize
 end

app/views/batch_contexts/_new_bc_form.erb

@@ -40,7 +40,7 @@
                 collection: [['Yes', true], ['No', false]] %>
         </div>
         <div data-caption-target="runOcr" hidden>
-            <%= form.input :run_ocr, input_html: { data: { action: "change->caption#runOcrChanged", "ocr-target": "runOcrOptions" } },
+            <%= form.input :run_ocr, input_html: { data: { action: "change->caption#runOcrChanged", "caption-target": "runOcrOptions" } },
                 as: :radio_buttons,
                 collection: [['Yes', true], ['No', false]] %>
             <div data-caption-target="runOcrImageNotes" hidden>
@@ -82,19 +82,24 @@
         </div>
     </div>
 
+    <% unless Settings.ocr.enabled # if OCR is not enabled, keep the processing configuration menu %>
+       <%= form.input :processing_configuration, collection: processing_configuration %>
+    <% end %>
+
     <div data-caption-target="sttSettings" hidden>
-        <div data-caption-target="sttAvailable" hidden>
+        <div data-caption-target="sttAvailable">
             <%= form.input :stt_available, label: 'Do you have caption/transcript files for the media?', input_html: { data: { action: "change->caption#sttAvailableChanged" } },
                 as: :radio_buttons,
                 collection: [['Yes', true], ['No', false]] %>
         </div>
         <div data-caption-target="manuallyCorrectedStt" hidden>
-            <%= form.input :manually_corrected_stt, input_html: { data: { action: "change->caption#manuallyCorrectedSttChanged" } },
+            <%= form.input :manually_corrected_stt,
                 as: :radio_buttons,
                 collection: [['Yes', true], ['No', false]] %>
         </div>
-        <div data-caption-target="runStt" hidden>
-            <%= form.input :run_stt, input_html: { data: { action: "change->caption#runSttChanged", "caption-target": "runSttOptions" } },
+        <div data-caption-target="runStt">
+            <%= form.input :run_stt, label: 'Would you like to auto-generate caption/transcript files for the media?',
+                input_html: { data: { action: "change->caption#runSttChanged" } },
                 as: :radio_buttons,
                 collection: [['Yes', true], ['No', false]] %>
             <div data-caption-target="runSttNotes" hidden>

db/migrate/20241011190309_stt.rb

@@ -0,0 +1,7 @@
+class Stt < ActiveRecord::Migration[7.1]
+  def change
+    add_column :batch_contexts, :run_stt, :boolean, default: false
+    add_column :batch_contexts, :manually_corrected_stt, :boolean, default: false
+    add_column :batch_contexts, :stt_available, :boolean, default: false
+  end
+end

db/schema.rb

@@ -10,7 +10,7 @@
 #
 # It's strongly recommended that you check this file into your version control system.
 
-ActiveRecord::Schema[7.1].define(version: 2024_05_29_215747) do
+ActiveRecord::Schema[7.1].define(version: 2024_10_11_190309) do
   # These are extensions that must be enabled in order to support this database
   enable_extension "plpgsql"
 
@@ -40,6 +40,9 @@
     t.boolean "manually_corrected_ocr", default: false
     t.jsonb "ocr_languages", default: []
     t.boolean "ocr_available", default: false
+    t.boolean "run_stt", default: false
+    t.boolean "manually_corrected_stt", default: false
+    t.boolean "stt_available", default: false
     t.index ["user_id", "project_name"], name: "index_batch_contexts_on_user_id_and_project_name", unique: true
     t.index ["user_id"], name: "index_batch_contexts_on_user_id"
   end

spec/features/preassembly_run/dark_file_object_spec.rb

@@ -63,7 +63,8 @@
             filesets: Array,
             all_files_public: false,
             reading_order: nil,
-            manually_corrected_ocr: false)
+            manually_corrected_ocr: false,
+            manually_corrected_stt: false)
     expect(dsc_object).to have_received(:update).with(params: item)
     expect(StartAccession).to have_received(:run).with(druid: "druid:#{bare_druid}", batch_context: BatchContext.last, workflow: 'assemblyWF')
   end

spec/features/preassembly_run/document_spec.rb

@@ -64,7 +64,8 @@
               filesets: Array,
               all_files_public: false,
               reading_order: nil,
-              manually_corrected_ocr: false)
+              manually_corrected_ocr: false,
+              manually_corrected_stt: false)
       expect(dsc_object).to have_received(:update).with(params: item)
       expect(StartAccession).to have_received(:run).with(druid: "druid:#{bare_druid}", batch_context: BatchContext.last, workflow: 'assemblyWF')
     end

spec/features/preassembly_run/geo_spec.rb

@@ -61,7 +61,8 @@
             filesets: [], # a blank array for any geo object
             all_files_public: false,
             reading_order: nil,
-            manually_corrected_ocr: false)
+            manually_corrected_ocr: false,
+            manually_corrected_stt: false)
     expect(dsc_object).to have_received(:update).with(params: item)
     expect(StartAccession).to have_received(:run).with(druid: "druid:#{bare_druid}", batch_context: BatchContext.last, workflow: 'gisAssemblyWF')
   end

spec/features/preassembly_run/hierarchical_files_spec.rb

@@ -76,7 +76,8 @@
               filesets: Array,
               all_files_public: false,
               reading_order: nil,
-              manually_corrected_ocr: false)
+              manually_corrected_ocr: false,
+              manually_corrected_stt: false)
       expect(dsc_object).to have_received(:update).with(params: item)
       expect(StartAccession).to have_received(:run).with(druid: "druid:#{bare_druid}", batch_context: BatchContext.last, workflow: 'assemblyWF')
     end