Models for MeMAD language identification pipeline

A collection of models for MeMAD spoken language identification pipeline. Zip contains four models:

1. An xvector embedding model trained on 67 languages using the lidbox toolkit.
2. A scikit-learn StandardScaler for standardizing the embedding model output before Naîve Bayes classification.
3. A probabilistic linear discriminant analysis (PLDA) model for reducing the dimensions of the embedding vectors.
4. A scikit-learn Naïve Bayes model for classifying embedding vectors to six categories: de, en, fi, fr, sv, x-nolang