metadata

task_categories:
  - audio-classification
language:
  - de
tags:
  - intent
  - intent-classification
  - audio-classification
  - audio
base_model:
  - facebook/wav2vec2-xls-r-300m
datasets:
  - FBK-MT/Speech-MASSIVE
library_name: transformers
license: apache-2.0

wav2vec 2.0 XLS-R 128 (300m) fine-tuned on Speech-MASSIVE - de-DE

Speech-MASSIVE is a multilingual Spoken Language Understanding (SLU) dataset comprising the speech counterpart for a portion of the MASSIVE textual corpus. Speech-MASSIVE covers 12 languages. It includes spoken and written utterances and is annotated with 60 intents. The dataset is available on HuggingFace Hub.

This is the facebook/wav2vec2-xls-r-300m model fine-tuned on the de-DE language.

It achieves the following results on the test set:

Accuracy: 0.681
F1: 0.584

Usage

You can use the model directly in the following manner:

import torch
import librosa
from transformers import AutoModelForAudioClassification, AutoFeatureExtractor

## Load an audio file
audio_array, sr = librosa.load("path_to_audio.wav", sr=16000)

## Load model and feature extractor
model = AutoModelForAudioClassification.from_pretrained("alkiskoudounas/xls-r-128-speechmassive-de-DE")
feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-xls-r-300m")

## Extract features
inputs = feature_extractor(audio_array.squeeze(), sampling_rate=feature_extractor.sampling_rate, padding=True, return_tensors="pt")

## Compute logits
logits = model(**inputs).logits

Framework versions

Datasets 3.2.0
Pytorch 2.1.2
Tokenizers 0.20.3
Transformers 4.45.2

BibTeX entry and citation info

@inproceedings{koudounas2025unlearning,
  title={"Alexa, can you forget me?" Machine Unlearning Benchmark in Spoken Language Understanding},
  author={Koudounas, Alkis and Savelli, Claudio and Giobergia, Flavio and Baralis, Elena},
  booktitle={Proc. Interspeech 2025}, 
  year={2025},
}