W

Wav2vec2 Xls R Parlaspeech Hr Lm

Developed by classla
A Croatian automatic speech recognition model fine-tuned from facebook/wav2vec2-xls-r-300m, trained on the ParlaSpeech-HR v1.0 dataset
Downloads 18
Release Time : 4/28/2022

Model Overview

This model is an automatic speech recognition system for Croatian, based on the wav2vec2-xls-r architecture, specifically designed to convert Croatian speech into text

Model Features

High-precision recognition
Achieves 3.63% character error rate and 9.85% word error rate on the ParlaSpeech-HR test set
Parliamentary speech optimization
Specially trained and optimized for Croatian parliamentary speech data
Language model enhancement
Incorporates a language model (LM) for decoding to improve recognition accuracy

Model Capabilities

Croatian speech recognition
Parliamentary speech transcription
Real-time speech-to-text

Use Cases

Government agencies
Parliament meeting minutes
Automatically transcribe Croatian parliamentary meeting content
Improves meeting minute efficiency and reduces manual transcription costs
Speech transcription services
Croatian speech transcription
Provides speech-to-text services for Croatian speakers
Achieves over 90% word recognition accuracy
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase