Summary of Cmu’s Iwslt 2024 Simultaneous Speech Translation System, by Xi Xu and Siqi Ouyang and Brian Yan and Patrick Fernandes and William Chen and Lei Li and Graham Neubig and Shinji Watanabe
CMU’s IWSLT 2024 Simultaneous Speech Translation System
by Xi Xu, Siqi Ouyang, Brian Yan, Patrick Fernandes, William Chen, Lei Li, Graham Neubig, Shinji Watanabe
First submitted to arxiv on: 14 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 presents CMU’s entry to the IWSLT 2024 Simultaneous Speech Translation (SST) task, which aims to translate English speech into German text in real-time. The team developed an end-to-end speech-to-text system that combines a WavLM speech encoder, modality adapter, and Llama2-7B-Base decoder. They employed a two-stage training approach, aligning speech and text representations initially, followed by full fine-tuning on MuST-c v2 data using cross-entropy loss. For SST, they adapted their offline ST model with a simple fixed hold-n policy. The system achieved an offline BLEU score of 31.1 and 29.5 under 2 seconds latency on the MuST-C-v2 tst-COMMON dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us talk to machines in real-time. Imagine being able to have a conversation with someone who speaks German, but you only understand English. This is what this paper does! It’s like having a super smart translator that can translate spoken words into text as we speak. The team used special models and training methods to make their system work well for translating speech to text in real-time. They tested it on a big dataset and got really good results! |
Keywords
» Artificial intelligence » Bleu » Cross entropy » Decoder » Encoder » Fine tuning » Translation
