【24h】

Simple Transliteration for CLIR

机译:CLIR的简单音译

获取原文
获取原文并翻译 | 示例

摘要

This is an experiment in cross-lingual information retrieval for Indian languages, in a resource-poor situation. We use a simple grapheme-to-grapheme transliteration technique to transliterate parallel query-text between three morphologically similar Indian languages and compare the cross-lingual and mono-lingual performance. Where a state of the art system like the Google Translation tool performs roughly in the range of 60-90%, our transliteration technique achieves 20-60% of the mono-lingual performance. Though the figures are not impressive, we argue that in situations where linguistic resources are scarce, to the point of being non-existent, this can be a starting point of engineering retrieval effectiveness.
机译:这是在资源匮乏的情况下针对印度语言进行跨语言信息检索的实验。我们使用一种简单的字素到字素音译技术在三种形态相似的印度语言之间音译平行查询文本,并比较了跨语言和单语言的性能。最先进的系统(例如Google翻译工具)的性能大致在60-90%的范围内,而我们的音译技术可实现20-60%的单语种性能。尽管这些数字并不令人印象深刻,但我们认为,在语言资源稀缺的情况下(甚至不存在),这可能是工程检索效率的起点。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号