arXiv:1305.1319 · 中英对照阅读
用于判别式图书摘要的新对齐方法:进行中的工作
New Alignment Methods for Discriminative Book Summarization Work in Progress
中文速览
把整本书和很短的情节摘要自动对应起来很难,因为摘要可能只用一句话概括书中几页甚至一整章,普通的逐词匹配方法很难胜任。作者提出了两种基于隐马尔可夫模型(HMM)的无监督方法:一种把书切成可变长度的段落,让摘要句子对应到合适的段落;另一种在逐词对齐的基础上改进转移学习、空位匹配和词语变体处理。实验显示,这些方法能提升基于原文句子抽取的图书摘要效果,但仍有很大改进空间。除了帮助摘要系统利用书籍配对数据学习,这种对齐也能揭示哪些情节和内容最容易被人类写进摘要。
摘要
我们研究将一本书的全文与人工撰写的摘要进行无监督对齐。这带来了其他文本对齐问题中未曾出现的挑战,包括长度差异,以及由此导致的违背预期:单个词语和短语本应彼此对齐,因为大段内容和章节可以浓缩为一个摘要短语。我们提出了两种专门针对这一问题、基于隐马尔可夫模型(HMM)的新方法,并证明这些方法在抽取式图书摘要任务中取得了提升。尽管仍有很大的改进空间,无监督对齐本身也具有内在价值,因为它能够帮助洞察一本书中哪些特征被认为值得纳入摘要。
We consider the unsupervised alignment of the full text of a book with a human-written summary. This presents challenges not seen in other text alignment problems, including a disparity in length and, consequent to this, a violation of the expectation that individual words and phrases should align, since large passages and chapters can be distilled into a single summary phrase. We present two new methods, based on hidden Markov models, specifically targeted to this problem, and demonstrate gains on an extractive book summarization task. While there is still much room for improvement, unsupervised alignment holds intrinsic value in offering insight into what features of a book are deemed worthy of summarization.
术语表
- unsupervised alignment
- 无监督对齐
- extractive summarization
- 抽取式摘要
- supervised discriminative summarization
- 有监督判别式摘要
- word-level alignment
- 词级对齐
- phrasal alignment
- 短语级对齐
- monolingual alignment
- 单语对齐
- comparable corpora
- 可比语料库
- Ziff-Davis corpus
- Ziff-Davis 语料库
- Project Gutenberg
- Project Gutenberg
- hidden Markov model (HMM)
- 隐马尔可夫模型(HMM)
- passage model
- 段落模型
- HMM alignment model
- HMM 对齐模型
- state
- 状态
- observation sequence
- 观测序列
- emission distribution
- 发射分布
- transition distribution
- 转移分布
- unigram language model
- 一元语言模型
- unigram distribution
- 一元分布
- relative frequency estimation
- 相对频率估计
- EM-like algorithm
- 类 EM 算法
- forward-backward algorithm
- 前向后向算法
- posterior distribution
- 后验分布
- likelihood
- 似然
- state boundary
- 状态边界
- source document
- 源文档
- target summary
- 目标摘要