Abstract
This paper compares techniques to combine diverse parallel corpora for domain-specific phrase-based SMT system training. We address a common scenario where little in-domain data is available for the task, but where large background models exist for the same language pair. In particular, we focus on phrase table fill-up: a method that effectively exploits background knowledge to improve model coverage, while preserving the more reliable information coming from the in-domain corpus. We present experiments on an emerging transcribed speech translation task – the TED talks. While performing similarly in terms of BLEU and NIST scores to the popular log-linear and linear interpolation techniques, filled-up translation models are more compact and easy to tune by minimum error training.- Anthology ID:
- 2011.iwslt-evaluation.18
- Volume:
- Proceedings of the 8th International Workshop on Spoken Language Translation: Evaluation Campaign
- Month:
- December 8-9
- Year:
- 2011
- Address:
- San Francisco, California
- Editors:
- Marcello Federico, Mei-Yuh Hwang, Margit Rödder, Sebastian Stüker
- Venue:
- IWSLT
- SIG:
- SIGSLT
- Publisher:
- Note:
- Pages:
- 136–143
- Language:
- URL:
- https://aclanthology.org/2011.iwslt-evaluation.18
- DOI:
- Cite (ACL):
- Arianna Bisazza, Nick Ruiz, and Marcello Federico. 2011. Fill-up versus interpolation methods for phrase-based SMT adaptation. In Proceedings of the 8th International Workshop on Spoken Language Translation: Evaluation Campaign, pages 136–143, San Francisco, California.
- Cite (Informal):
- Fill-up versus interpolation methods for phrase-based SMT adaptation (Bisazza et al., IWSLT 2011)
- PDF:
- https://preview.aclanthology.org/ingest-2024-clasp/2011.iwslt-evaluation.18.pdf