Summary of Don’t Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation, by Giorgos Vernikos and Andrei Popescu-belis
Don’t Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation
by Giorgos Vernikos, Andrei Popescu-Belis
First submitted to arxiv on: 12 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
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 This paper introduces QE-fusion, a novel method for neural machine translation systems to estimate probabilities of target sentences given source sentences. Unlike traditional methods like beam search, QE-fusion synthesizes translations using a quality estimation metric (QE) that correlates better with human judgments. The authors compare QE-fusion against recent reranking techniques and large language models (LLMs) used for translation, such as PolyLM, XGLM, Llama2, Mistral, ALMA, and Tower, demonstrating consistent improvements in translation quality measured by COMET and BLEURT scores. Notably, QE-fusion exhibits larger improvements for LLMs due to their ability to generate diverse outputs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine a machine that can translate languages really well. Right now, these machines don’t always choose the best translation option. This paper introduces a new way of translating called QE-fusion, which picks the best parts from multiple translations and combines them into one better translation. The authors tested this method against other ways of translating and found that it works much better. They also used special language models to test their method and saw even more improvement. This new approach can help us communicate with people who speak different languages more effectively. |
Keywords
* Artificial intelligence * Translation
