File size: 2,116 Bytes
6dd9bad | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 | import { prisma } from './prisma';
import { calculateWER } from '../utils/metrics';
import { normalizeWolof } from '../scripts/normalizeWolof';
export class TrainingService {
static async getPendingAudios(page: number, limit: number) {
const [pending, total] = await Promise.all([
prisma.trainingData.findMany({
where: { status: 'PENDING' },
orderBy: { createdAt: 'desc' },
skip: (page - 1) * limit,
take: limit,
}),
prisma.trainingData.count({ where: { status: 'PENDING' } })
]);
return { pending, total, page, limit };
}
static async submitCorrection(id: string | undefined, audioUrl: string, transcription: string, manualCorrection: string) {
const normResult = normalizeWolof(transcription);
const rawWER = calculateWER(manualCorrection, transcription);
const normalizedWER = calculateWER(manualCorrection, normResult.normalizedText);
// Analyze missing words
const manualWords = manualCorrection.toLowerCase().replace(/[.,/#!$%^&*;:{}=\-_`~()]/g, "").split(/\s+/).filter(w => w);
const normWords = normResult.normalizedText.toLowerCase().replace(/[.,/#!$%^&*;:{}=\-_`~()]/g, "").split(/\s+/).filter((w: string) => w);
const missingWords = Array.from(new Set(manualWords.filter((w: string) => !normWords.includes(w))));
const data = await prisma.trainingData.upsert({
where: { id: id || '00000000-0000-0000-0000-000000000000' },
update: { manualCorrection, rawWER, normalizedWER, status: 'REVIEWED' },
create: { audioUrl, transcription, manualCorrection, rawWER, normalizedWER, status: 'REVIEWED' }
});
return { data, missingWords, rawWER, normalizedWER };
}
static async createTrainingRecord(audioUrl: string, transcription: string) {
return prisma.trainingData.create({
data: {
audioUrl,
transcription: transcription || '',
status: 'PENDING'
}
});
}
}
|