Changeset 35252 for main/trunk/model-interfaces-dev/atea

Timestamp:

2021-08-05T09:57:47+12:00 (3 years ago)

Author:

davidb

Message:

Implement auto-chunking of transcription requests, and display more error information to the user.

Location:

main/trunk/model-interfaces-dev/atea

Files:

• js/asr/TranscribeService.js (modified) (2 diffs)
• js/asr/asr-controller.js (modified) (3 diffs)
• style/asr.css (modified) (1 diff)
• transform/pages/asr.xsl (modified) (1 diff)

main/trunk/model-interfaces-dev/atea/js/asr/TranscribeService.js

@@ -5 +5 @@
  * @property {String} file_name The name of the file that was transcribed.
  * @property {String} log A note of how the transcription was processed.
- * @property {{char: String, start_time: Number}[]} metadata The character metadata.
+ * @property {{char: String, confidence: Number, start_time: Number}[]} metadata The character metadata.
  * @property {boolean} success A value indicating if the transcription was successful or not.
  * @property {String} transcription The transcription.
@@ -37 +37 @@
 
     /**
+     * Performs chunked queries to transcribe the given audio files, returning the data in iterations.
+     * Data is chunked according to which ever occurs first:
+     * A maximum of three files per request, or;
+     * A maximum of 5 MiB per request.
+     * 
+     * @param {FileList} files The files to upload
+     * @returns {AsyncGenerator<TranscriptionModel[]>} The transcribed audio files.
+     */
+    async* chunkTranscribeFiles(files)
+    {
+        let filesToSubmit = [];
+        let fileCounter = 0;
+        let byteCounter = 0;
+
+        for (let file of files) {
+            if (fileCounter == 3 || byteCounter > 5242880) // 5 MiB
+            {
+                yield await this.transcribeFiles(filesToSubmit);
+                filesToSubmit = [];
+                byteCounter = 0;
+                fileCounter = 0;
+            }
+
+            filesToSubmit[fileCounter++] = file;
+            byteCounter += file.size;
+        }
+
+        if (filesToSubmit.length > 0) {
+            yield await this.transcribeFiles(filesToSubmit);
+        }
+    }
+
+    /**
      * Performs a query to transcribe the given audio files.
      * 
-     * @param {FileList} files The files to upload.
-     * @returns {Promise<TranscriptionModel[]>} The transcribed aufio file.
+     * @param {FileList | File[]} files The files to upload.
+     * @returns {Promise<TranscriptionModel[]>} The transcribed audio file.
      */
     async transcribeFiles(files)

main/trunk/model-interfaces-dev/atea/js/asr/asr-controller.js

@@ -49 +49 @@
     const files = $(FILE_UPLOAD_INPUT_NAME)[0].files;
     const transcribeService = new TranscribeService();
-    
+
     $(FILE_UPLOAD_PROGRESS_CONTAINER_NAME).removeClass("asr-hidden");
 
@@ -66 +66 @@
     try
     {
-        const transcriptions = await transcribeService.transcribeFiles(files);
-        for (const t of transcriptions)
+        for await (const transcriptionBundle of transcribeService.chunkTranscribeFiles(files))
         {
-            if (!t.success) {
-                insertError(t.file_name, t.log);
-            }
-            else {
-                insertTranscription(t);
+            for (const t of transcriptionBundle)
+            {
+                if (!t.success) {
+                    insertError(t.file_name, t.log);
+                }
+                else {
+                    insertTranscription(t);
+                }
             }
         }
@@ -79 +81 @@
     catch (e)
     {
+        console.error("Failed to transcribe files: " + e);
         insertError("all", e.statusMessage);
     }

main/trunk/model-interfaces-dev/atea/style/asr.css

@@ -19 +19 @@
 }
 
-.transcription-list-item .spaced-block {
+.transcription-list-item .spaced-block, .error-list-item .spaced-block {
   margin-bottom: 8px;
 }
 
-.transcription-list-item .spaced-block:last-child {
+.transcription-list-item .spaced-block:last-child, .error-list-item .spaced-block:last-child {
   margin-bottom: 0;
 }

main/trunk/model-interfaces-dev/atea/transform/pages/asr.xsl

@@ -84 +84 @@
                 <li class="error-list-item">
                     <div>
-                        Transcription failed: <span></span><br />
-                        <b>File: </b><span></span>
+                        <div class="spaced-block">
+                            <b>Transcription Failed: </b><span></span>
+                        </div>
+                        <div class="spaced-block">
+                            <b>File: </b><span></span>
+                        </div>
                     </div>
                 </li>