# Multi-dialect Support

XiYanSQL-QwenCoder-3B-2504 is the latest SQL generation model released by XGenerationLab. Optimized through fine-tuning and GRPO training, it supports multiple dialects and delivers efficient and accurate SQL generation capabilities.

A fine-tuned SQL generation model based on QwenCoder, supporting multiple dialects with excellent performance

This model is fine-tuned from MMS-300m for Arabic dialect speech recognition, capable of identifying Modern Standard Arabic and four major Arabic dialects.

A speech recognition model fine-tuned on Telugu audio datasets based on OpenAI Whisper-large-v2

A high-quality Arabic speech synthesis model fine-tuned based on F5-TTS, supporting diverse pronunciations and accents from different regions

AudioX is a multilingual automatic speech recognition model developed by Jivi AI, specifically optimized for South Indian languages, supporting Tamil, Telugu, Kannada, and Malayalam.

A 7-billion-parameter model fine-tuned on CodeLlama, specifically designed for natural language to SQL tasks, supporting multiple SQL dialects and 16k context length processing

IndicWhisper is a cutting-edge speech recognition model optimized for Indian languages, excelling in various benchmarks for Indian languages.

An Arabic speech recognition model based on Whisper-base, fine-tuned on multiple Arabic datasets, specializing in Arabic speech-to-text tasks

This model is trained on Arabic dialect datasets for translating Arabic dialects into Modern Standard Arabic (MSA).

A Chinese text-to-speech model fine-tuned on the fleurs dataset based on microsoft/speecht5_tts

A French automatic speech recognition model fine-tuned based on openai/whisper-small, trained on the Common Voice 11.0 French dataset, supporting case sensitivity and punctuation prediction.

A Telugu automatic speech recognition (ASR) model fine-tuned based on OpenAI Whisper-base, trained on multiple public Telugu datasets

A speech recognition model fine-tuned on Arabic datasets based on openai/whisper-medium

A Northern Sami speech recognition model fine-tuned on Whisper-large-v2, achieving a word error rate of 24.91% on the test set

This is a neural machine translation model for Arabic to English translation, part of the OPUS-MT project.

This model is a speech recognition model fine-tuned on the Common Voice dataset based on facebook/wav2vec2-xls-r-300m, supporting Hindi.

This model is an automatic speech recognition model fine-tuned on the Romansh-Sursilvan dialect dataset based on facebook/wav2vec2-xls-r-1b, achieving a word error rate (WER) of 13.82% on the Common Voice 8 test set.

A Hausa speech recognition model fine-tuned on the Common Voice dataset based on facebook/wav2vec2-xls-r-300m

A sentiment analysis model trained based on the BETO Spanish BERT model, supporting POS/NEG/NEU three-class sentiment classification

A BERT model optimized for Arabic dialects and tweets, pre-trained on 60 million Arabic tweets with MLM tasks, with added support for emojis and common vocabulary.

AraBERTv0.2-Twitter is a pre-trained language model optimized for Arabic dialects and tweets, developed based on the BERT architecture, with added support for emojis and common vocabulary.

Arabic pretrained version of ALBERT large model, trained on approximately 4.4 billion words of Arabic corpus

An automatic speech recognition model based on XLS-R architecture, optimized for Cantonese (zh-HK), fine-tuned on the Common Voice dataset and enhanced with a 5-gram language model.

Wav2vec2 model fine-tuned for Brazilian Portuguese using the CETUC dataset, trained with approximately 145 hours of Brazilian Portuguese speech data

This is an automatic speech recognition (ASR) model fine-tuned on Hindi speech datasets based on Facebook's wav2vec2-xls-r-300m model

An Arabic version of the ALBERT Xlarge pretrained language model, trained on approximately 4.4 billion words, supporting Modern Standard Arabic and some dialectal content.

An automatic speech recognition model fine-tuned on Marathi datasets based on facebook/wav2vec2-xls-r-300m

A Chinese (zh-CN) speech recognition model fine-tuned on the Common Voice dataset based on facebook/wav2vec2-large-xlsr-53

This is an automatic speech recognition model fine-tuned on Galician speech datasets based on facebook/wav2vec2-xls-r-300m.

Pre-trained Arabic BERT medium language model, trained on approximately 8.2 billion words of Arabic text resources

AraBERT is a pre-trained language model based on Google's BERT architecture, specifically designed for Arabic natural language understanding tasks.

A CTC model based on XLS-R with a 5-gram language model from Open Subtitles, primarily used for automatic speech recognition in Dutch and Flemish.

A BERT model for Arabic dialects, further trained on the HARD-Arabic-Dataset based on CAMeL-Lab's bert-base-camelbert-msa-eighth model

An Arabic named entity recognition model fine-tuned based on CAMeLBERT Mix, supporting entity recognition in Modern Standard Arabic, dialects, and Classical Arabic

A dialect identification model fine-tuned based on the CAMeLBERT Modern Standard Arabic model, supporting 21 Arabic dialect identifications.

A Nynorsk automatic speech recognition model based on the XLSR-300M architecture, trained on the NPSC dataset with low word error rate and character error rate.

A Transformer-based machine translation model for Swedish to Chinese, supporting multiple Chinese variants, developed by the Helsinki-NLP team

This is a machine translation model from Chinese to German based on the transformer-align architecture, supporting translation from various Chinese dialect variants such as Mandarin and Cantonese to German.

This is a Norwegian T5-based model trained on the Norwegian Colossal Corpus (NCC) using TPU v3-8.