Open-source model wav2vec2-xls-r-300m-ftspeech - Accurately implement automatic Danish speech recognition

Wav2vec2 Xls R 300m Ftspeech

Developed by saattrupdan

A Danish automatic speech recognition model fine-tuned on Danish parliamentary speech dataset based on facebook/wav2vec2-xls-r-300m

Speech Recognition

Transformers

OtherOpen Source License:Other #Danish speech recognition #Parliament scenario optimization #Low word error rate

Downloads 1.3M

Release Time : 3/4/2022

Model Overview

This is a model for Danish automatic speech recognition (ASR), based on the wav2vec2-xls-r-300m architecture, fine-tuned on 1,800 hours of Danish parliamentary speech data.

Model Features

Danish language optimization

Specifically optimized for Danish speech recognition, fine-tuned on Danish parliamentary speech dataset

High performance

Achieved low word error rates on the Common Voice 8.0 Danish subset and Alvenir test set

Language model support

Supports using 5-gram language model to further improve recognition accuracy

Model Capabilities

Danish speech recognition

Speech-to-text

5-gram language model enhancement support

Use Cases

Speech transcription

Parliamentary speech transcription

Automatically transcribe Danish parliamentary speeches into text

Trained on parliamentary datasets, suitable for such content recognition

General speech recognition

Daily speech transcription

Convert Danish daily conversations into text

Achieved WER of 17.91 on Common Voice dataset

🚀 XLS-R-300m-FTSpeech

This is a fine - tuned model based on the facebook/wav2vec2 - xls - r - 300m, trained on the FTSpeech dataset, aiming to provide high - quality automatic speech recognition for Danish.

🚀 Quick Start

This model is designed for automatic speech recognition tasks. You can use it in your projects by referring to the Hugging Face's model usage guidelines.

✨ Features

Fine - tuned on Danish Data: Trained on the FTSpeech dataset, which consists of 1,800 hours of transcribed speeches from the Danish parliament, making it well - adapted to Danish language characteristics.
Good Performance: Achieves relatively low Word Error Rate (WER) scores on multiple Danish datasets, indicating high recognition accuracy.

📚 Documentation

Model description

This model is a fine - tuned version of [facebook/wav2vec2 - xls - r - 300m](https://huggingface.co/facebook/wav2vec2 - xls - r - 300m) on the FTSpeech dataset, being a dataset of 1,800 hours of transcribed speeches from the Danish parliament.

Performance

The model achieves the following WER scores (lower is better):

Dataset	WER without LM	WER with 5 - gram LM
[Danish part of Common Voice 8.0](https://huggingface.co/datasets/mozilla - foundation/common_voice_8_0/viewer/da/train)	20.48	17.91
Alvenir test set	15.46	13.84

Information Table

Property	Details
Language	Danish
License	Other
Datasets	FTSpeech
Metrics	WER
Tasks	Automatic Speech Recognition
Base Model	facebook/wav2vec2 - xls - r - 300m

Model Index

Name: wav2vec2 - xls - r - 300m - ftspeech
- Results:
  - Task:
    - Type: automatic - speech - recognition
  - Dataset:
    - Name: Danish Common Voice 8.0
    - Type: mozilla - foundation/common_voice_8_0
    - Args: da
  - Metrics:
    - Type: wer
    - Value: 17.91
  - Task:
    - Type: automatic - speech - recognition
  - Dataset:
    - Name: Alvenir ASR test dataset
    - Type: Alvenir/alvenir_asr_da_eval
  - Metrics:
    - Type: wer
    - Value: 13.84

📄 License

The use of this model needs to adhere to [this license from the Danish Parliament](https://www.ft.dk/da/aktuelt/tv - fra - folketinget/deling - og - rettigheder).

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご