mirror of
https://github.com/jambonz/batch-speech-utils.git
synced 2026-01-25 02:08:27 +00:00
lib update with functional style code
This commit is contained in:
@@ -1,66 +0,0 @@
|
||||
// deepgramTranscriber.js
|
||||
const Transcriber = require('./transcriber');
|
||||
const { createClient } = require('@deepgram/sdk');
|
||||
const fs = require('fs');
|
||||
|
||||
class DeepgramTranscriber extends Transcriber {
|
||||
constructor(apiKey, audioFilePath) {
|
||||
super(audioFilePath);
|
||||
this.deepgram = createClient(apiKey);
|
||||
}
|
||||
|
||||
async transcribeFile(filePath) {
|
||||
const { result } = await this.deepgram.listen.prerecorded.transcribeFile(
|
||||
fs.readFileSync(filePath),
|
||||
{ model: 'nova-2', smart_format: true, detect_entities: true }
|
||||
);
|
||||
return result;
|
||||
}
|
||||
|
||||
async redactFile(filePath) {
|
||||
const { result } = await this.deepgram.listen.prerecorded.transcribeFile(
|
||||
fs.readFileSync(filePath),
|
||||
{ model: 'nova-2', smart_format: true, redact: 'pii' }
|
||||
);
|
||||
return result;
|
||||
}
|
||||
|
||||
async analyzeText(text) {
|
||||
const { result } = await this.deepgram.read.analyzeText(
|
||||
{ text },
|
||||
{ language: 'en', sentiment: true, intents: true, summarize: true }
|
||||
);
|
||||
return result;
|
||||
}
|
||||
|
||||
async processAudio(filePath = this.audioDir) {
|
||||
try {
|
||||
const transcription = await this.transcribeFile(filePath);
|
||||
const redaction = await this.redactFile(filePath);
|
||||
const transcript = transcription.results.channels[0].alternatives[0].transcript;
|
||||
const timestamps = transcription.results.channels[0].alternatives[0].words;
|
||||
const redactionTimestamps = redaction.results.channels[0].alternatives[0].words;
|
||||
const redacted = redaction.results.channels[0].alternatives[0].transcript;
|
||||
const entities = transcription.results.channels[0].alternatives[0].entities;
|
||||
|
||||
const analysisResult = await this.analyzeText(transcript);
|
||||
const sentimentSegment = analysisResult.results.sentiments.segments[0];
|
||||
const sentiment = sentimentSegment.sentiment;
|
||||
const sentimentScore = sentimentSegment.sentiment_score;
|
||||
|
||||
return {
|
||||
transcript,
|
||||
timestamps,
|
||||
redactionTimestamps,
|
||||
redacted,
|
||||
sentiment,
|
||||
sentimentScore,
|
||||
entities
|
||||
};
|
||||
} catch (error) {
|
||||
throw error;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
module.exports = DeepgramTranscriber;
|
||||
59
lib/get-transcription-deepgram.js
Normal file
59
lib/get-transcription-deepgram.js
Normal file
@@ -0,0 +1,59 @@
|
||||
const fs = require('fs');
|
||||
|
||||
async function transcribeFile(deepgram, filePath) {
|
||||
const fileBuffer = fs.readFileSync(filePath);
|
||||
const options = { model: 'nova-2', smart_format: true, detect_entities: true };
|
||||
const { result } = await deepgram.listen.prerecorded.transcribeFile(fileBuffer, options);
|
||||
return result;
|
||||
}
|
||||
|
||||
async function redactFile(deepgram, filePath) {
|
||||
const fileBuffer = fs.readFileSync(filePath);
|
||||
const options = { model: 'nova-2', smart_format: true, redact: 'pii' };
|
||||
const { result } = await deepgram.listen.prerecorded.transcribeFile(fileBuffer, options);
|
||||
return result;
|
||||
}
|
||||
|
||||
async function analyzeText(deepgram, text) {
|
||||
const options = { language: 'en', sentiment: true, intents: true, summarize: true };
|
||||
const { result } = await deepgram.read.analyzeText({ text }, options);
|
||||
return result;
|
||||
}
|
||||
|
||||
|
||||
async function processAudio(deepgram, filePath) {
|
||||
try {
|
||||
const transcription = await transcribeFile(deepgram, filePath);
|
||||
const redaction = await redactFile(deepgram, filePath);
|
||||
const transcript = transcription.results.channels[0].alternatives[0].transcript;
|
||||
const timestamps = transcription.results.channels[0].alternatives[0].words;
|
||||
const redactionTimestamps = redaction.results.channels[0].alternatives[0].words;
|
||||
const redacted = redaction.results.channels[0].alternatives[0].transcript;
|
||||
const entities = transcription.results.channels[0].alternatives[0].entities;
|
||||
|
||||
const analysisResult = await analyzeText(deepgram, transcript);
|
||||
const sentimentSegment = analysisResult.results.sentiments.segments[0];
|
||||
const sentiment = sentimentSegment.sentiment;
|
||||
const sentimentScore = sentimentSegment.sentiment_score;
|
||||
|
||||
return {
|
||||
transcript,
|
||||
timestamps,
|
||||
redactionTimestamps,
|
||||
redacted,
|
||||
sentiment,
|
||||
sentimentScore,
|
||||
entities
|
||||
};
|
||||
} catch (error) {
|
||||
console.error('Error processing audio:', error);
|
||||
throw error;
|
||||
}
|
||||
}
|
||||
|
||||
module.exports = {
|
||||
transcribeFile,
|
||||
analyzeText,
|
||||
processAudio,
|
||||
redactFile
|
||||
};
|
||||
@@ -1,15 +1,15 @@
|
||||
const Redactor = require('./redactor');
|
||||
const ffmpeg = require('fluent-ffmpeg');
|
||||
|
||||
class DeepGramRedactor extends Redactor {
|
||||
async redactAudio(transcriptionData, audioPath, audioOutputPath, { delta = 0.05 } = {}) {
|
||||
|
||||
async function redactAudioDeepgram(transcriptionData, audioPath, audioOutputPath, { delta = 0.05 } = {}) {
|
||||
return new Promise((resolve, reject) => {
|
||||
const command = ffmpeg(audioPath)
|
||||
.outputFormat('wav'); // Ensure output format is WAV
|
||||
|
||||
// Iterate over transcription data to apply audio filters
|
||||
for (let i = 0; i < transcriptionData.length; i++) {
|
||||
const {word, start} = transcriptionData[i];
|
||||
let end = transcriptionData[i].end; // Default end time
|
||||
transcriptionData.forEach((data, i) => {
|
||||
const { word, start } = data;
|
||||
let end = data.end; // Default end time
|
||||
|
||||
// Check if the word needs redaction
|
||||
if (word.startsWith('[') && word.endsWith(']')) {
|
||||
@@ -28,15 +28,19 @@ class DeepGramRedactor extends Redactor {
|
||||
// Log the redacted segments
|
||||
console.log(`Redacting from ${start}s to ${end}s: "${word}"`);
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
// Handlers for command execution
|
||||
command.on('end', () => {
|
||||
console.log(`Redacted audio saved at ${audioOutputPath}`);
|
||||
resolve(); // Resolve the promise on successful completion
|
||||
}).on('error', (err, stdout, stderr) => {
|
||||
console.error('Error processing audio file:', err.message);
|
||||
console.error('ffmpeg stdout:', stdout);
|
||||
console.error('ffmpeg stderr:', stderr);
|
||||
reject(err); // Reject the promise on error
|
||||
}).saveToFile(audioOutputPath);
|
||||
}
|
||||
});
|
||||
}
|
||||
module.exports = DeepGramRedactor;
|
||||
|
||||
module.exports = { redactAudioDeepgram };
|
||||
16
lib/redact.js
Normal file
16
lib/redact.js
Normal file
@@ -0,0 +1,16 @@
|
||||
const { redactAudioDeepgram } = require('./make-redact-audio-deepgram');
|
||||
|
||||
const assert = require('assert');
|
||||
|
||||
function getRedactedAudio(vendor, transcriptionData, audioPath, audioOutputPath) {
|
||||
assert.ok(['deepgram', 'otherVendor'].includes(vendor), 'vendor not supported');
|
||||
|
||||
if (vendor === 'deepgram') {
|
||||
return redactAudioDeepgram(transcriptionData, audioPath, audioOutputPath);
|
||||
}
|
||||
else {
|
||||
throw new Error(`Unsupported vendor: ${vendor}`);
|
||||
}
|
||||
}
|
||||
|
||||
module.exports = { getRedactedAudio };
|
||||
@@ -1,13 +0,0 @@
|
||||
class Redactor {
|
||||
constructor() {
|
||||
if (new.target === Redactor) {
|
||||
throw new TypeError('Cannot construct Redactor instances directly');
|
||||
}
|
||||
}
|
||||
|
||||
async redactAudio(transcriptionData, audioPath, audioOutputPath, options) {
|
||||
throw new Error("Method 'redactAudio' must be implemented.");
|
||||
}
|
||||
}
|
||||
|
||||
module.exports = Redactor;
|
||||
19
lib/transcribe.js
Normal file
19
lib/transcribe.js
Normal file
@@ -0,0 +1,19 @@
|
||||
const { createDeepGramClient } = require('./utils');
|
||||
const fs = require('fs');
|
||||
const { processAudio } = require('./get-transcription-deepgram');
|
||||
const assert = require('assert');
|
||||
|
||||
function getTranscription(vendor, apiKey, filePath) {
|
||||
assert.ok(['deepgram'].includes(vendor), 'vendor not supported');
|
||||
if (vendor === 'deepgram') {
|
||||
const deepgramClient = createDeepGramClient(apiKey);
|
||||
assert.ok(deepgramClient, 'Invalid Deepgram API key');
|
||||
return processAudio(deepgramClient, filePath);
|
||||
}
|
||||
else {
|
||||
throw new Error(`Unsupported vendor: ${vendor}`);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
module.exports = {getTranscription};
|
||||
@@ -1,27 +0,0 @@
|
||||
// transcriber.js
|
||||
class Transcriber {
|
||||
constructor(audioFilePath) {
|
||||
if (new.target === Transcriber) {
|
||||
throw new TypeError('Cannot construct Transcriber instances directly');
|
||||
}
|
||||
this.audioDir = audioFilePath;
|
||||
}
|
||||
|
||||
async transcribeFile(filePath) {
|
||||
throw new Error("Method 'transcribeFile' must be implemented.");
|
||||
}
|
||||
|
||||
async redactFile(filePath) {
|
||||
throw new Error("Method 'redactFile' must be implemented.");
|
||||
}
|
||||
|
||||
async analyzeText(text) {
|
||||
throw new Error("Method 'analyzeText' must be implemented.");
|
||||
}
|
||||
|
||||
async processAudio(filePath) {
|
||||
throw new Error("Method 'processAudio' must be implemented.");
|
||||
}
|
||||
}
|
||||
|
||||
module.exports = Transcriber;
|
||||
27
lib/utils.js
27
lib/utils.js
@@ -1,26 +1,9 @@
|
||||
const DeepgramTranscriber = require('./deepgramTranscriber');
|
||||
const DeepgramRedactor = require('./deepgramRedactor');
|
||||
|
||||
class AudioProcessing {
|
||||
static getTranscriber(serviceType,DEEPGRAM_API_KEY, audioFilePath) {
|
||||
switch (serviceType) {
|
||||
case 'deepgram':
|
||||
return new DeepgramTranscriber(DEEPGRAM_API_KEY, audioFilePath);
|
||||
const { createClient } = require('@deepgram/sdk');
|
||||
|
||||
default:
|
||||
throw new Error('Unknown transcription service');
|
||||
}
|
||||
}
|
||||
|
||||
static getRedactor(serviceType, audioFilePath) {
|
||||
switch (serviceType) {
|
||||
case 'deepgram':
|
||||
return new DeepgramRedactor(audioFilePath);
|
||||
|
||||
default:
|
||||
throw new Error('Unknown redaction service');
|
||||
}
|
||||
}
|
||||
function createDeepGramClient(DEEPGRAM_API_KEY) {
|
||||
return createClient(DEEPGRAM_API_KEY);
|
||||
}
|
||||
|
||||
module.exports = AudioProcessing;
|
||||
|
||||
module.exports = {createDeepGramClient};
|
||||
|
||||
Reference in New Issue
Block a user