source: main/trunk/model-interfaces-dev/atea/korero-maori-asr/src/js/TranscribeModule.js@ 35456

/**
 * @file Defines components that are used to interact with the transcription proxy servlet.
 * @author Carl Stephens
 * @module
 */

/* eslint-disable camelcase */

export class TranscriptionMetadata {
    /**
     * A metadata object held by a {@link TranscriptionModel}.
     *
     * @param {String} char The character.
     * @param {Number} confidence The confidence that the character is correct. Smaller values are more confident.
     * @param {Number} start_time The timestamp in the audio at which this character is spoken.
     */
    constructor(char, confidence, start_time) {
        this.char = char;
        this.confidence = confidence;
        this.start_time = start_time;
    }
}

export class TranscriptionModel {
    /**
     * The transcription object returned from our transcription endpoint.
     *
     * @param {String} file_name The name of the file that was transcribed.
     * @param {String} log A note of how the transcription was processed.
     * @param {TranscriptionMetadata[]} metadata The character metadata.
     * @param {boolean} success A value indicating if the transcription was successful or not.
     * @param {String} transcription The transcription.
     */
    constructor(file_name, log, metadata, success, transcription) {
        this.file_name = file_name;
        this.log = log;
        this.metadata = metadata;
        this.success = success;
        this.transcription = transcription;
    }
}

/* eslint-enable camelcase */

export class TranscriptionError extends Error {
    /**
     * Initialises a new instance of the {@link TranscriptionError} object.
     *
     * @param {Number | undefined} statusCode The HTTP status code of the error.
     * @param {String | undefined} message The status message.
     * @param {String | null} fileName The file on which the transcription failed.
     */
    constructor(message = undefined, fileName = null, statusCode = -1) {
        super(message);

        /** @type {String | null} The name of the file that the transcription error occured on. */
        this.fileName = fileName;

        /** @type {Number | undefined} The status code returned by the API when the erro was generated. */
        this.statusCode = statusCode;
    }
}

/**
 * A service that uploads audio files in a multipart request to the Korero Maori API Interface servlet, and interprets the result.
 */
export default class TranscribeService {
    constructor() {
        /* eslint-disable quotes */
        this.TEST_JSON = '{"file_name":"akl_mi_pk_0002.wav","log":"Took 0.6s to transcribe 8.7s audio file /tmp/transcribe-aw7ujpu8-clean.wav","metadata":[{"char":"k","prob":7.0E-6,"start_time":0.24},{"char":"o","prob":8.3E-5,"start_time":0.26},{"char":"t","prob":0.995578,"start_time":0.42},{"char":"a","prob":0.997089,"start_time":0.44},{"char":"h","prob":3.0E-6,"start_time":0.54},{"char":"i","prob":8.0E-6,"start_time":0.56},{"char":" ","prob":2.2E-5,"start_time":0.6},{"char":"a","prob":0.982987,"start_time":0.64},{"char":"n","prob":5.0E-6,"start_time":0.8},{"char":"ō","prob":1.82E-4,"start_time":0.84},{"char":" ","prob":0.996944,"start_time":1.08},{"char":"t","prob":1.23E-4,"start_time":1.14},{"char":"e","prob":6.3E-5,"start_time":1.16},{"char":" ","prob":0.587595,"start_time":1.28},{"char":"t","prob":4.91E-4,"start_time":1.34},{"char":"u","prob":0.010654,"start_time":1.36},{"char":"p","prob":5.9E-5,"start_time":1.52},{"char":"u","prob":1.26E-4,"start_time":1.54},{"char":"n","prob":0.0,"start_time":1.7},{"char":"a","prob":0.99997,"start_time":1.72},{"char":" ","prob":0.993724,"start_time":2.14},{"char":"o","prob":4.3E-5,"start_time":2.22},{"char":" ","prob":0.99696,"start_time":2.32},{"char":"t","prob":1.4E-5,"start_time":2.36},{"char":"e","prob":3.8E-5,"start_time":2.38},{"char":" ","prob":0.996606,"start_time":2.48},{"char":"t","prob":0.0,"start_time":2.52},{"char":"a","prob":0.994873,"start_time":2.54},{"char":"ŋ","prob":2.9E-5,"start_time":2.64},{"char":"a","prob":0.994639,"start_time":2.66},{"char":"t","prob":3.0E-6,"start_time":2.8},{"char":"a","prob":0.99995,"start_time":2.82},{"char":" ","prob":0.0,"start_time":2.9},{"char":"m","prob":0.999606,"start_time":2.98},{"char":"ā","prob":0.010628,"start_time":3.0},{"char":"o","prob":0.996415,"start_time":3.2},{"char":"r","prob":0.0,"start_time":3.28},{"char":"i","prob":0.999698,"start_time":3.32},{"char":" ","prob":0.997927,"start_time":4.5},{"char":"k","prob":0.0,"start_time":4.58},{"char":"o","prob":3.2E-5,"start_time":4.6},{"char":" ","prob":0.956949,"start_time":4.7},{"char":"r","prob":2.5E-5,"start_time":4.74},{"char":"a","prob":0.997649,"start_time":4.76},{"char":"ŋ","prob":0.0,"start_time":4.9},{"char":"i","prob":1.63E-4,"start_time":4.92},{"char":"n","prob":1.3E-5,"start_time":5.28},{"char":"u","prob":4.0E-6,"start_time":5.3},{"char":"i","prob":0.99997,"start_time":5.42},{"char":" ","prob":0.994305,"start_time":5.78},{"char":"e","prob":3.1E-5,"start_time":5.84},{"char":" ","prob":0.966097,"start_time":6.04},{"char":"t","prob":2.0E-6,"start_time":6.1},{"char":"ū","prob":5.4E-5,"start_time":6.12},{"char":" ","prob":0.837368,"start_time":6.34},{"char":"n","prob":0.0,"start_time":6.38},{"char":"e","prob":0.004238,"start_time":6.4},{"char":"i","prob":0.99966,"start_time":6.5},{"char":" ","prob":0.99728,"start_time":7.28},{"char":"o","prob":6.28E-4,"start_time":7.36},{"char":" ","prob":0.961049,"start_time":7.48},{"char":"p","prob":5.0E-6,"start_time":7.52},{"char":"a","prob":0.984707,"start_time":7.54},{"char":"p","prob":9.0E-6,"start_time":7.66},{"char":"a","prob":0.998754,"start_time":7.68},{"char":"t","prob":0.002356,"start_time":7.88},{"char":"ū","prob":1.24E-4,"start_time":7.9},{"char":"ā","prob":0.883958,"start_time":8.0},{"char":"n","prob":0.985871,"start_time":8.24},{"char":"u","prob":0.907369,"start_time":8.26},{"char":"k","prob":0.999045,"start_time":8.38},{"char":"u","prob":0.995632,"start_time":8.4}],"success":true,"transcription":"kotahi anō te tupuna o te tangata māori ko ranginui e tū nei o papatūānuku"}';
        /* eslint-enable quotes */

        /** @type {String} The URL to which query POST requests should be made. */
        this.queryUrl = "//localhost:8383/gs3-koreromaori/transcribe";

        /** @type {Number} The maximum number of files which can be transcribed in one request to the API. */
        this.MAX_BATCH_COUNT = 3;

        /** @type {Number} The soft upper limit on how many bytes can be submitted in one request. */
        this.BATCH_BYTE_LIMIT = 5242880; // 5 MiB
    }

    /**
     * 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 before chunking.
     *
     * @param {FileList | File[]} files The files to upload
     * @returns {AsyncGenerator<TranscriptionModel[]>} The transcribed audio files.
     * @throws {TranscriptionError} When the transcription request fails to complete.
     */
    async* batchTranscribeFiles(files) {
        let filesToSubmit = [];
        let fileCounter = 0;
        let byteCounter = 0;

        for (const file of files) {
            if (fileCounter === this.MAX_BATCH_COUNT || byteCounter > this.BATCH_BYTE_LIMIT) { // 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 | File[]} files The files to upload.
     * @returns {Promise<TranscriptionModel[]>} The transcribed audio file.
     * @throws {TranscriptionError} When the transcription request fails to complete.
     */
    async transcribeFiles(files) {
        const objects = [];

        for (const file of files) {
            const object = JSON.parse(this.TEST_JSON);
            object.file_name = file.name;
            objects.push(object);
        }

        return objects;
    }

    /**
     * Performs a query to transcribe the given audio files.
     *
     * @param {FileList | File[]} files The files to upload.
     * @returns {Promise<TranscriptionModel[]>} The transcribed audio file.
     * @throws {TranscriptionError} When the transcription request fails to complete.
     */
    async transcribeFilesActual(files) {
        const that = this;
        const formData = new FormData();

        let audioFileKeys = "";
        for (let i = 0; i < files.length; i++) {
            const f = files[i];
            const key = "audioFile" + i;

            formData.append(key, f, f.name);
            audioFileKeys += key + "|";
        }
        formData.append("audioFileKeys", audioFileKeys);

        try {
            const response = await fetch(
                that.queryUrl,
                {
                    method: "POST",
                    body: formData
                }
            );

            if (!response.ok) {
                console.error(`Transcription API failed with status ${response.status} and message ${response.statusText}`);
                throw new TranscriptionError(response.statusText, undefined, response.status);
            }

            return await response.json();
        }
        catch (e) {
            console.error(`Transcription failed with reason ${e}`);
            throw new TranscriptionError(undefined, "Unknown");
        }
    }
}