thanks! primarily classical. we designed an ensemble that considers both lexical & semantic similarity and trained on large datasets of labeled text.
we train our own 'out-of-the-box' models (like the intent classification, IAB, O*NET-SOC, NAICS) or you can create a custom model by just defining what labels you want.
