arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1404.4641cs.CL
收藏

组合分布式语义的多语言模型

Multilingual Models for Compositional Distributed Semantics

University of Oxford(牛津大学)

Karl Moritz Hermann, Phil Blunsom

摘要:

我们提出了一种新的技术,学习语义表示,扩展的分布假设,多语言数据和联合空间嵌入。我们的模型利用并行数据,并学习强烈对齐语义等价句子的嵌入,同时在不同句子之间保持足够的距离。该模型不依赖于单词对齐或任何句法信息,并成功地应用于许多不同的语言。我们也扩展了我们的方法来学习文档级别的语义表示。我们评估这些模型的两个跨语言的文档分类任务,优于现有的最先进的状态。通过定性分析和研究的旋转效应,我们证明,我们的表示是语义上合理的,可以捕捉语义关系,跨语言,而无需并行数据。

英文摘要:

We present a novel technique for learning semantic representations, which extends the distributional hypothesis to multilingual data and joint-space embeddings. Our models leverage parallel data and learn to strongly align the embeddings of semantically equivalent sentences, while maintaining sufficient distance between those of dissimilar sentences. The models do not rely on word alignments or any syntactic information and are successfully applied to a number of diverse languages. We extend our approach to learn semantic representations at the document level, too. We evaluate these models on two cross-lingual document classification tasks, outperforming the prior state of the art. Through qualitative analysis and the study of pivoting effects we demonstrate that our representations are semantically plausible and can capture semantic relationships across languages without parallel data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑