FunASR/docs/papers.md

Papers

FunASR have implemented the following paper code

Speech Recognition Models

• Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition, INTERSPEECH 2022.
• Universal ASR: Unifying Streaming and Non-Streaming ASR Using a Single Encoder-Decoder Model, arXiv preprint arXiv:2010.14099, 2020.
• San-m: Memory equipped self-attention for end-to-end speech recognition, INTERSPEECH 2020
• Streaming Chunk-Aware Multihead Attention for Online End-to-End Speech Recognition, INTERSPEECH 2020
• Conformer: Convolution-augmented Transformer for Speech Recognition, INTERSPEECH 2020
• Sequence-to-sequence learning with Transducers, NIPS 2016

Multi-talker Speech Recognition Models

• MFCCA:Multi-Frame Cross-Channel attention for multi-speaker ASR in Multi-party meeting scenario, ICASSP 2022

Voice Activity Detection

• Deep-FSMN for Large Vocabulary Continuous Speech Recognition, ICASSP 2018

Punctuation Restoration

• CT-Transformer: Controllable time-delay transformer for real-time punctuation prediction and disfluency detection, ICASSP 2018

Language Models

• Attention Is All You Need, NEURIPS 2017

Speaker Verification

• X-VECTORS: ROBUST DNN EMBEDDINGS FOR SPEAKER RECOGNITION, ICASSP 2018

Speaker diarization

• Speaker Overlap-aware Neural Diarization for Multi-party Meeting Analysis, EMNLP 2022
• TOLD: A Novel Two-Stage Overlap-Aware Framework for Speaker Diarization, ICASSP 2023

Timestamp Prediction

• Achieving Timestamp Prediction While Recognizing with Non-Autoregressive End-to-End ASR Model, arXiv:2301.12343