On the feasibility of character n-grams pseudo-translation for Cross-Language Information Retrieval tasks
详细信息 查看全文
- 作者：Jesú ; s Vilaresa ; jesus.vilares@udc.es" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Manuel Vilaresb ; vilares@uvigo.es" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Miguel A. Alonsoa ; miguel.alonso@udc.es" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae ; Michael P. Oakesc ; michael.oakes@wlv.ac.uk" class="auth_mail" title="E-mail the corresponding authorAuthor Vitae
- 关键词：Cross-Language Information Retrieval ; Character n-grams ; Alignment algorithms for Machine Translation
- 刊名：Computer Speech & Language
- 出版年：2016
- 出版时间：March 2016
- 年：2016
- 卷：36
- 期：Complete
- 页码：136-164
- 全文大小：2242 K
文摘
- •
We analyze the use of character n-grams both as indexing and translation units for CLIR tasks.
- •
We study their effective application and consistency across languages.
- •
We use an algorithm of our own for parallel text alignment at the subword level.
- •
Tests were performed for seven languages, with English as the target language.
- •
Results confirm their feasibility and consistency, that their validity is not tied to a given implementation, and a remarkable robustness.