Summary of A Light-weight and Efficient Punctuation and Word Casing Prediction Model For On-device Streaming Asr, by Jian You et al.
A light-weight and efficient punctuation and word casing prediction model for on-device streaming ASR
by Jian You, Xiangfeng Li
First submitted to arxiv on: 18 Jul 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG); Sound (cs.SD); Audio and Speech Processing (eess.AS)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a lightweight and efficient model for predicting punctuation and word casing in real-time for on-device end-to-end streaming automatic speech recognition (ASR) systems. The model, based on Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM), jointly predicts punctuation and word casing, achieving 9% relative improvement over non-Transformer models on overall F1-score. Compared to a representative Transformer-based model, the proposed model is one-fortieth smaller and 2.5 times faster in inference time while obtaining comparable results. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper creates a new way to predict punctuation and word casing for automatic speech recognition (ASR) that works on devices like phones. Right now, there’s not much talk about this important topic. The team uses a special kind of AI model called a Transformer, but it’s too big for devices. So, they create a smaller, faster version using Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM). They test their model on some data and show that it does better than other models. |
Keywords
» Artificial intelligence » Cnn » F1 score » Inference » Neural network » Transformer
