Edit model card

Wav2Vec2-Large for Intent Classification

Model description

This is a ported version of S3PRL's Wav2Vec2 for the SUPERB Intent Classification task.

The base model is wav2vec2-large-lv60, which is pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.

For more information refer to SUPERB: Speech processing Universal PERformance Benchmark

Task and dataset description

Intent Classification (IC) classifies utterances into predefined classes to determine the intent of speakers. SUPERB uses the Fluent Speech Commands dataset, where each utterance is tagged with three intent labels: action, object, and location.

For the original model's training and evaluation instructions refer to the S3PRL downstream task README.

Usage examples

You can use the model directly like so:

import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForSequenceClassification, Wav2Vec2FeatureExtractor

def map_to_array(example):
    speech, _ = librosa.load(example["file"], sr=16000, mono=True)
    example["speech"] = speech
    return example

# load a demo dataset and read audio files
dataset = load_dataset("anton-l/superb_demo", "ic", split="test")
dataset = dataset.map(map_to_array)

model = Wav2Vec2ForSequenceClassification.from_pretrained("superb/wav2vec2-large-superb-ic")
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained("superb/wav2vec2-large-superb-ic")

# compute attention masks and normalize the waveform if needed
inputs = feature_extractor(dataset[:4]["speech"], sampling_rate=16000, padding=True, return_tensors="pt")

logits = model(**inputs).logits

action_ids = torch.argmax(logits[:, :6], dim=-1).tolist()
action_labels = [model.config.id2label[_id] for _id in action_ids]

object_ids = torch.argmax(logits[:, 6:20], dim=-1).tolist()
object_labels = [model.config.id2label[_id + 6] for _id in object_ids]

location_ids = torch.argmax(logits[:, 20:24], dim=-1).tolist()
location_labels = [model.config.id2label[_id + 20] for _id in location_ids]

Eval results

The evaluation metric is accuracy.

s3prl transformers
test 0.9528 N/A

BibTeX entry and citation info

@article{yang2021superb,
  title={SUPERB: Speech processing Universal PERformance Benchmark},
  author={Yang, Shu-wen and Chi, Po-Han and Chuang, Yung-Sung and Lai, Cheng-I Jeff and Lakhotia, Kushal and Lin, Yist Y and Liu, Andy T and Shi, Jiatong and Chang, Xuankai and Lin, Guan-Ting and others},
  journal={arXiv preprint arXiv:2105.01051},
  year={2021}
}
Downloads last month
1,858

Dataset used to train superb/wav2vec2-large-superb-ic