Lattention: Lattice-attention in asr rescoring

[免费ChatGPT] [DeepSeek可用网址] [PDF] arxiv.org

Scaling laws for discriminative speech recognition rescoring models‏

Y Gu, PG Shivakumar, J Kolehmainen… - arxiv preprint arxiv …, 2023‏ - arxiv.org‏
Recent studies have found that model performance has a smooth power-law relationship, or
scaling laws, with training data and model size, for a wide range of problems. These scaling …‏
שמור צטט צוטט על ידי 9 מאמרים בנושא זה כל 6 הגרסאות פתיחה בתור HTML
[免费ChatGPT] [DeepSeek可用网址] [PDF] arxiv.org

Transformer-based Model for ASR N-Best Rescoring and Rewriting‏

IE Kang, C Van Gysel, MH Siu - arxiv preprint arxiv:2406.08207, 2024‏ - arxiv.org‏
Voice assistants increasingly use on-device Automatic Speech Recognition (ASR) to ensure
speed and privacy. However, due to resource constraints on the device, queries pertaining …‏
שמור צטט צוטט על ידי 2 מאמרים בנושא זה כל 5 הגרסאות פתיחה בתור HTML
[免费ChatGPT] [DeepSeek可用网址] [PDF] arxiv.org

EEL: Efficiently encoding lattices for reranking‏

P Singhal, J Xu, X Ye, G Durrett - arxiv preprint arxiv:2306.00947, 2023‏ - arxiv.org‏
Standard decoding approaches for conditional text generation tasks typically search for an
output hypothesis with high model probability, but this may not yield the best hypothesis …‏
שמור צטט צוטט על ידי 1 מאמרים בנושא זה כל 4 הגרסאות פתיחה בתור HTML
[免费ChatGPT] [DeepSeek可用网址] [PDF] arxiv.org

Personalization for bert-based discriminative speech recognition rescoring‏

J Kolehmainen, Y Gu, A Gourav… - arxiv preprint arxiv …, 2023‏ - arxiv.org‏
Recognition of personalized content remains a challenge in end-to-end speech recognition.
We explore three novel approaches that use personalized content in a neural rescoring step …‏
שמור צטט צוטט על ידי 3 מאמרים בנושא זה כל 5 הגרסאות פתיחה בתור HTML
[免费ChatGPT] [DeepSeek可用网址] [PDF] arxiv.org

Leveraging Cross-Utterance Context For ASR Decoding‏

R Flynn, A Ragni - arxiv preprint arxiv:2306.16903, 2023‏ - arxiv.org‏
While external language models (LMs) are often incorporated into the decoding stage of
automated speech recognition systems, these models usually operate with limited context …‏
שמור צטט צוטט על ידי 1 מאמרים בנושא זה כל 5 הגרסאות פתיחה בתור HTML
[免费ChatGPT] [DeepSeek可用网址] [HTML] amazon.science

[HTML][HTML] RNN-T lattice enhancement by grafting of pruned paths‏

M Novak, P Papadopoulos - 2022‏ - amazon.science‏
Abstract Recurrent Neural Network Transducers (RNN-T)—a streaming variant of end-to-
end models—became very popular in recent years. Since RNN-T networks condition the …‏
שמור צטט צוטט על ידי 3 מאמרים בנושא זה כל 3 הגרסאות במטמון