基于多译文的中文转述语料库建设及转述评价方案.PDFVIP

文章编号：1003-0077 （2011）00-0000-00 基于多译文的中文转述语料库建设及转述评价方案 1,2 1,2 2 2 1,2** 阮翀 ，施文娴 ，李岩昊 ，翁伊嘉 ，胡俊峰 （1.北京大学计算语言学教育部重点实验室，北京市 100871； 2. 北京大学信息科学技术学院，北京市 100871） 摘要：转述语料是转述现象研究的基础。针对目前学术界中文转述语料稀缺的现状，本文以《简爱》的多 1 个中文译本为基础，通过句对齐得到五万句级别的平行转述语料 。使用无监督的小句对齐和词对齐算法， 本文从语料中挖掘到九千多对词汇转述知识。同时，本文还复现和改进了机器翻译测评指标 Meteor ，使得 该指标更适合于中文转述句子的测评，并构造了一个中文句子转述测评数据集，以便对不同的转述知识和 评价指标进行比较。实验表明，本文算法挖掘到的词汇转述知识在封闭测试中不逊于《同义词词林》。 关键词：转述知识挖掘；转述评价指标；转述语料建设 中图分类号：TP391 文献标识码：A Multi-Translation Based Chinese Paraphrase Corpus Construction and Paraphrasing Evaluation Methods 1,2 1,2 2 2 1,2 RUAN Chong , SHI Wenxian , LI Yanhao , WENG Yijia , HU Junfeng (1. Key Laboratory of Computational Linguistics (Ministry of Education), Peking University, Beijing, 100871, China; 2. School of Electronics and Computer Science, Peking University, Beijing, 100871, China) Abstract: Paraphrase corpus is fundamental to research in paraphrase phenomenon, while Chinese paraphrase corpus is hardly available in academia. In this paper, we collected multiple Chinese translations of the novel Jane Eyre, and obtained around 50 thousand parallel paraphrasing sentences. Based on such corpus, we managed to extract more than 9,000 pairs of lexical paraphrase knowledge. We also reimplemented and improved a machine translation evaluation metric Meteor, making it more suitable for Chinese paraphrase evaluation. We constructed a Chinese paraphrase evaluation dataset to facilitate the comparison of different paraphrase knowledge sources and evaluation methods. The experiments show the quality of our mined knowledge is comparable to Tongyici C