--- license: apache-2.0 base_model: - facebook/hubert-base-ls960 tags: - intent-classification - slu - audio-classification metrics: - accuracy - f1 model-index: - name: hubert-base-unslurp-gold results: [] datasets: - unslurp language: - en pipeline_tag: audio-classification library_name: transformers --- # HuBERT-base-UNSLURP-GOLD (Retain Set) This model is a fine-tuned version of [facebook/hubert-base-ls960](https://huggingface.co/facebook/hubert-base-ls960) on the UNSLURP dataset (retain set) for the intent classification task. SLURP does not provide speaker-independent splits, which are, however, required by Machine Unlearning techniques to be effective. In fact, the identities present in the retain, forget, and test sets must be exclusive to successfully apply and evaluate unlearning methods. To address this, we propose new speaker-independent splits. In the following, we refer to the new dataset as SLURP*, or UNSLURP. It achieves the following results on the test set: - Accuracy: 0.826 - F1: 0.704 ## Model description The base [Facebook's Hubert](https://ai.facebook.com/blog/hubert-self-supervised-representation-learning-for-speech-recognition-generation-and-compression) model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. ## Task and dataset description Intent Classification (IC) classifies utterances into predefined classes to determine the intent of speakers. The dataset used here is [(UN)SLURP](https://arxiv.org/abs/2011.13205), where each utterance is tagged with two intent labels: action and scenario. ## Usage examples You can use the model directly in the following manner: ```python 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/hubert-base-unslurp-gold") feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/hubert-base-ls960") ## 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 ```bibtex @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}, } ```