基于加权AMR图的语义子图预测摘要算法
|
摘要
方法多数停留在挖掘词与词之间的浅层语义关系,没有很好地利用词句之间的完整语义信息,为此,提出一种改进的语义子图预测摘要的算法。将原始文本转化为相应的抽象语义表示(AMR)图,融合成一个AMR总图,基于WordNet语义词典对其进行冗余信息的过滤。在此基础上利用综合统计特征对不具有权值的AMR图节点赋予权值,通过筛选重要性程度高的部分构成语义摘要子图,并基于ROUGE指标和Smatch指标综合衡量生成摘要的质量。实验结果表明,与仅挖掘浅层语义关系的文本摘要基准算法相比,该算法ROUGE值和Smatch值明显提高。
Most of the existing text abstract methods stay in the shallow semantic relationship between words and w ords,and do not make good use of the complete semantic information betw een w ords. Therefore,an improved algorithm for semantic subgraph predictive summary is proposed. The algorithm transforms the original text into corresponding Abstract M eaning Representation( AM R) graphs,merges them into an AM R total graph,and filters the redundant information based on the WordNet semantic dictionary. On this basis,using the comprehensive statistical features assigns w eights to the AM R graph nodes that do not have w eights,and constructs the semantic summary subgraphs by filtering the parts w ith high importance,and comprehensively measures the quality of the abstracts based on the ROUGE index and the Smatch index. Experimental results show that compared w ith the text abstraction benchmark algorithm w hich only mines shallow semantic relations,the ROUGE value and Smatch value of the algorithm are significantly improved.
Most of the existing text abstract methods stay in the shallow semantic relationship between words and w ords,and do not make good use of the complete semantic information betw een w ords. Therefore,an improved algorithm for semantic subgraph predictive summary is proposed. The algorithm transforms the original text into corresponding Abstract M eaning Representation( AM R) graphs,merges them into an AM R total graph,and filters the redundant information based on the WordNet semantic dictionary. On this basis,using the comprehensive statistical features assigns w eights to the AM R graph nodes that do not have w eights,and constructs the semantic summary subgraphs by filtering the parts w ith high importance,and comprehensively measures the quality of the abstracts based on the ROUGE index and the Smatch index. Experimental results show that compared w ith the text abstraction benchmark algorithm w hich only mines shallow semantic relations,the ROUGE value and Smatch value of the algorithm are significantly improved.
引文
[1]NALLAPATI R,ZHOU B,SANTOS C N D,et al.Abstractive text summarization using sequence-tosequence RNNs and beyond[C]//Proceedings of CoNLL’16.Washington D.C.,USA:IEEE Press,2016:125-136.
[2]王萌,何婷婷,姬东鸿,等.基于HowNet概念获取的中文自动文摘系统[J].中文信息学报,2005,19(3):87-93.
[3]MILLER G A.WordNet:a lexical database for English[J].Communications of the ACM,1995,38(11):39-41.
[4]MIHALCEA R,TARAU P.TextRank:bringing order into texts[EB/OL].[2018-01-21].http://w w w.aclw eb.org/.
[5]李宝程.基于浅层语义分析的文本摘要方法研究与实现[D].成都:电子科技大学,2016.
[6]吴晓锋,宗成庆.一种基于LDA的CRF自动文摘方法[J].中文信息学报,2009,23(6):39-45.
[7]罗森林,白建敏,潘丽敏,等.融合句义特征的多文档自动摘要算法研究[J].北京理工大学学报,2016,36(10):1059-1064.
[8]BANARESCU L,BONIAL C,CAI S,et al.Abstract meaning representation for sembanking[C]//Proceedings of Linguistic Annotation Workshop on Interoperability w ith Discourse.Washington D.C.,USA:IEEE Press,2013:178-186.
[9]LIU F,FLANIGN J,THOMSO S,et al.Toward abstractive summarization using semantic representations[C]//Proceedings of Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.Washington D.C.,USA:IEEE Press,2015:1077-1086.
[10]HERMANN K M,KOCISKY T.Teaching machines to read and comprehend[C]//Proceedings of International Conference on Neural Information Processing Systems.[S.1.]:MIT Press,2015:1693-1701.
[11]曲维光,周俊生,吴晓东,等.自然语言句子抽象语义表示AMR研究综述[J].数据采集与处理,2017,32(1):26-36.
[12]李斌,闻媛,宋丽,等.融合概念对齐信息的中文AMR语料库的构建[J].中文信息学报,2017,31(6):93-102.
[13]SONG L,PENG X,ZHANG Y,et al.AMR-to-text generation with synchronous node replacement Grammar[EB/OL].[2018-01-21].http://www.aclweb.org.
[14]KONSTAS I,IYER S,YATSKAR M,et al.Neural AMR:sequence-to-sequence models for parsing and generation[EB/OL].[2018-01-21].http://www.ikonstas.net.
[15]杜秀英.基于聚类与语义相似分析的多文本自动摘要方法[J].情报杂志,2017,36(6):167-172.
[16]宁可,孙同晶,徐洁洁.面向海量数据的改进最近邻优先吸收聚类算法[J].计算机工程,2018,44(4):35-40.
[17]GOLDSTEIN J,MITTAL V,CARBONELL J,et al.Multidocument summarization by sentence extraction[C]//Proceedings of the 2000 NAACL-ANLP Workshop on Automatic Summarization Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2000:40-48.
[18]孟令阁,马建芬,张雪英.基于主题的SVM与MMR融合的会议摘要技术[J].计算机工程与设计,2016,37(10):2695-2699.
[19]刘寒磊,关毅,徐永东.多文档文摘中基于语义相似度的最大边缘相关技术研究[C]//全国计算语言学联合学术会议论文集.南京:[出版社不详],2005.
[20]TAN J,WAN X,XIAO J.Abstractive document summarization with a graph-based attentional neural model[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2017:1171-1181.
[21]SEE A,LIU P J,MANNING C D.Get to the point:summarization with pointer-generator networks[EB/OL].[2018-01-21].http://www.aclweb.org/.
[22]FLICK C.ROUGE:a package for automatic evaluation of summaries[C]//Proceedings of IEEE Workshop on Text Summarization Branches Out.Washington D.C.,USA:IEEE Press,2004:10.
[23]CAI S,KNIGHT K.Smatch:an evaluation metric for semantic feature structures[C]//Proceedings of IEEEMeeting of the Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2012:748-752.
[2]王萌,何婷婷,姬东鸿,等.基于HowNet概念获取的中文自动文摘系统[J].中文信息学报,2005,19(3):87-93.
[3]MILLER G A.WordNet:a lexical database for English[J].Communications of the ACM,1995,38(11):39-41.
[4]MIHALCEA R,TARAU P.TextRank:bringing order into texts[EB/OL].[2018-01-21].http://w w w.aclw eb.org/.
[5]李宝程.基于浅层语义分析的文本摘要方法研究与实现[D].成都:电子科技大学,2016.
[6]吴晓锋,宗成庆.一种基于LDA的CRF自动文摘方法[J].中文信息学报,2009,23(6):39-45.
[7]罗森林,白建敏,潘丽敏,等.融合句义特征的多文档自动摘要算法研究[J].北京理工大学学报,2016,36(10):1059-1064.
[8]BANARESCU L,BONIAL C,CAI S,et al.Abstract meaning representation for sembanking[C]//Proceedings of Linguistic Annotation Workshop on Interoperability w ith Discourse.Washington D.C.,USA:IEEE Press,2013:178-186.
[9]LIU F,FLANIGN J,THOMSO S,et al.Toward abstractive summarization using semantic representations[C]//Proceedings of Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.Washington D.C.,USA:IEEE Press,2015:1077-1086.
[10]HERMANN K M,KOCISKY T.Teaching machines to read and comprehend[C]//Proceedings of International Conference on Neural Information Processing Systems.[S.1.]:MIT Press,2015:1693-1701.
[11]曲维光,周俊生,吴晓东,等.自然语言句子抽象语义表示AMR研究综述[J].数据采集与处理,2017,32(1):26-36.
[12]李斌,闻媛,宋丽,等.融合概念对齐信息的中文AMR语料库的构建[J].中文信息学报,2017,31(6):93-102.
[13]SONG L,PENG X,ZHANG Y,et al.AMR-to-text generation with synchronous node replacement Grammar[EB/OL].[2018-01-21].http://www.aclweb.org.
[14]KONSTAS I,IYER S,YATSKAR M,et al.Neural AMR:sequence-to-sequence models for parsing and generation[EB/OL].[2018-01-21].http://www.ikonstas.net.
[15]杜秀英.基于聚类与语义相似分析的多文本自动摘要方法[J].情报杂志,2017,36(6):167-172.
[16]宁可,孙同晶,徐洁洁.面向海量数据的改进最近邻优先吸收聚类算法[J].计算机工程,2018,44(4):35-40.
[17]GOLDSTEIN J,MITTAL V,CARBONELL J,et al.Multidocument summarization by sentence extraction[C]//Proceedings of the 2000 NAACL-ANLP Workshop on Automatic Summarization Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2000:40-48.
[18]孟令阁,马建芬,张雪英.基于主题的SVM与MMR融合的会议摘要技术[J].计算机工程与设计,2016,37(10):2695-2699.
[19]刘寒磊,关毅,徐永东.多文档文摘中基于语义相似度的最大边缘相关技术研究[C]//全国计算语言学联合学术会议论文集.南京:[出版社不详],2005.
[20]TAN J,WAN X,XIAO J.Abstractive document summarization with a graph-based attentional neural model[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2017:1171-1181.
[21]SEE A,LIU P J,MANNING C D.Get to the point:summarization with pointer-generator networks[EB/OL].[2018-01-21].http://www.aclweb.org/.
[22]FLICK C.ROUGE:a package for automatic evaluation of summaries[C]//Proceedings of IEEE Workshop on Text Summarization Branches Out.Washington D.C.,USA:IEEE Press,2004:10.
[23]CAI S,KNIGHT K.Smatch:an evaluation metric for semantic feature structures[C]//Proceedings of IEEEMeeting of the Association for Computational Linguistics.Washington D.C.,USA:IEEE Press,2012:748-752.