基于分类的中文文本摘要方法

自动文本摘要是自然语言处理领域中一项重要的研究内容,根据实现方式的不同其分为摘录式和理解式,其中理解式文摘是基于不同的形式对原始文档的中心内容和概念的重新表示,生成的文摘中的词语无需与原始文档相同。提出了一种基于分类的理解式文摘模型。该模型将基于递归神经网络的编码-解码结构与分类结构相结合,并充分利用监督信息,从而获得更多的摘要特性;通过在编码-解码结构中使用注意力机制,模型能更精确地获取原文的中心内容。模型的两部分可以同时在大数据集下进行训练优化,训练过程简单且有效。所提模型表现出了优异的自动摘要性能。     

基于深度学习的文本自动摘要方案

针对自然语言处理（NLP）生成式自动摘要领域的语义理解不充分、摘要语句不通顺和摘要准确度不够高的问题，提出了一种新的生成式自动摘要解决方案，包括一种改进的词向量生成技术和一个生成式自动摘要模型。改进的词向量生成技术以Skip-Gram方法生成的词向量为基础，结合摘要的特点，引入词性、词频和逆文本频率三个词特征，有效地提高了词语的理解；而提出的Bi-MulRnn+生成式自动摘要模型以序列映射（seq2seq）与自编码器结构为基础，引入注意力机制、门控循环单元（GRU）结构、双向循环神经网络（BiRnn）、多层循环神经网络（MultiRnn）和集束搜索，提高了生成式摘要准确性与语句流畅度。基于大规模中文短文本摘要（LCSTS）数据集的实验结果表明，该方案能够有效地解决短文本生成式摘要问题，并在Rouge标准评价体系中表现良好，提高了摘要准确性与语句流畅度。     

基于文本结构和图卷积网络的生成式摘要

目前主流的生成式自动文摘采用基于编码器—解码器架构的机器学习模型,且通常使用基于循环神经网络的编码器。该编码器主要学习文本的序列化信息,对文本的结构化信息学习能力较差。从语言学的角度来讲,文本的结构化信息对文本重要内容的判断具有重要作用。为了使编码器能够获取文本的结构信息,该文提出了基于文本结构信息的编码器,其使用了图卷积神经网络对文本进行编码。该文还提出了一种规范融合层,旨在使模型在获取文本结构信息的同时,也能关注到文本的序列化信息。另外,该文还使用了多头注意力机制的解码器,以提高生成摘要的质量。实验结果表明,在加入该文所提出的文本结构信息编码器、规范融合层后,系统性能在ROUGE评价指标上有显著的提高。     

基于指针网络的抽取生成式摘要生成模型

作为自然语言处理中的热点问题,摘要生成具有重要的研究意义。基于Seq2Seq模型的生成式摘要模型取得了良好的效果,然而抽取式的方法具有挖掘有效特征并抽取文章重要句子的潜力,因此如何利用抽取式方法来改进生成式方法是一个较好的研究方向。鉴于此,提出了融合生成式和抽取式方法的模型。首先,使用TextRank算法并融合主题相似度来抽取文章中有重要意义的句子。然后,设计了融合抽取信息语义的基于Seq2Seq模型的生成式框架来实现摘要生成任务;同时,引入指针网络解决模型训练中的未登录词（OOV）问题。综合以上步骤得到最终摘要,并在CNN/Daily Mail数据集上进行验证。结果表明在ROUGE-1、ROUGE-2和ROUGE-L三个指标上所提模型比传统TextRank算法均有所提升,同时也验证了融合抽取式和生成式方法在摘要生成领域中的有效性。     

主题关键词信息融合的中文生成式自动摘要研究

随着大数据和人工智能技术的迅猛发展,传统自动文摘研究正朝着从抽取式摘要到生成式摘要的方向演化,从中达到生成更高质量的自然流畅的文摘的目的.近年来,深度学习技术逐渐被应用于生成式摘要研究中,其中基于注意力机制的序列到序列模型已成为应用最广泛的模型之一,尤其在句子级摘要生成任务（如新闻标题生成、句子压缩等）中取得了显著的效果.然而,现有基于神经网络的生成式摘要模型绝大多数将注意力均匀分配到文本的所有内容中,而对其中蕴含的重要主题信息并没有细致区分.鉴于此,本文提出了一种新的融入主题关键词信息的多注意力序列到序列模型,通过联合注意力机制将文本中主题下重要的一些关键词语的信息与文本语义信息综合起来实现对摘要的引导生成.在NLPCC 2017的中文单文档摘要评测数据集上的实验结果验证了所提方法的有效性和先进性.     

A multiagent,dynamic rank-driven multi-deme architecture for real-valued multiobjective optimization

As information is available in abundance for every topic on internet, condensing the important information in the form of summary would benefit a number of users. Hence, there is growing interest among the research community for developing new approaches to automatically summarize the text. Automatic text summarization system generates a summary, i.e. short length text that includes all the important information of the document. Since the advent of text summarization in 1950s, researchers have been trying to improve techniques for generating summaries so that machine generated summary matches with the human made summary. Summary can be generated through extractive as well as abstractive methods. Abstractive methods are highly complex as they need extensive natural language processing. Therefore, research community is focusing more on extractive summaries, trying to achieve more coherent and meaningful summaries. During a decade, several extractive approaches have been developed for automatic summary generation that implements a number of machine learning and optimization techniques. This paper presents a comprehensive survey of recent text summarization extractive approaches developed in the last decade. Their needs are identified and their advantages and disadvantages are listed in a comparative manner. A few abstractive and multilingual text summarization approaches are also covered. Summary evaluation is another challenging issue in this research field. Therefore, intrinsic as well as extrinsic both the methods of summary evaluation are described in detail along with text summarization evaluation conferences and workshops. Furthermore, evaluation results of extractive summarization approaches are presented on some shared DUC datasets. Finally this paper concludes with the discussion of useful future directions that can help researchers to identify areas where further research is needed.     

基于BERT-PGN模型的中文新闻文本自动摘要生成

针对文本自动摘要任务中生成式摘要模型对句子的上下文理解不够充分、生成内容重复的问题,基于BERT和指针生成网络（PGN）,提出了一种面向中文新闻文本的生成式摘要模型——BERT-指针生成网络（BERT-PGN）。首先,利用BERT预训练语言模型结合多维语义特征获取词向量,从而得到更细粒度的文本上下文表示;然后,通过PGN模型,从词表或原文中抽取单词组成摘要;最后,结合coverage机制来减少重复内容的生成并获取最终的摘要结果。在2017年CCF国际自然语言处理与中文计算会议（NLPCC2017）单文档中文新闻摘要评测数据集上的实验结果表明,与PGN、伴随注意力机制的长短时记忆神经网络（LSTM-attention）等模型相比,结合多维语义特征的BERT-PGN模型对摘要原文的理解更加充分,生成的摘要内容更加丰富,全面且有效地减少重复、冗余内容的生成,Rouge-2和Rouge-4指标分别提升了1.5%和1.2%。     

基于优势演员-评论家算法的强化自动摘要模型

针对长文本自动摘要任务中抽取式模型摘要较为冗余,而生成式摘要模型时常有关键信息丢失、摘要不准确和生成内容重复等问题,提出一种面向长文本的基于优势演员-评论家算法的强化自动摘要模型（A2C-RLAS）。首先,用基于卷积神经网络（CNN）和循环神经网络（RNN）的混合神经网络的抽取器（extractor）来提取原文关键句;然后,用基于拷贝机制和注意力机制的重写器（rewriter）来精炼关键句;最后,使用强化学习的优势演员-评论家（A2C）算法训练整个网络,把重写摘要和参考摘要的语义相似性（BERTScore值）作为奖励（reward）来指导抽取过程,从而提高抽取器提取句子的质量。在CNN/Daily Mail数据集上的实验结果表明,与基于强化学习的抽取式摘要（Refresh）模型、基于循环神经网络的抽取式摘要序列模型（SummaRuNNer）和分布语义奖励（DSR）模型等模型相比,A2C-RLAS的最终摘要内容更加准确、语言更加流畅,冗余的内容有效减少,且A2C-RLAS的ROUGE和BERTScore指标均有提升。相较于Refresh模型和SummaRuNNer模型,A2C-RLAS模型的ROUGE-L值分别提高了6.3%和10.2%;相较于DSR模型,A2C-RLAS模型的F1值提高了30.5%。     

融合BERT和卷积门控的生成式文本摘要方法

结合注意力机制的循环神经网络(RNN)模型是目前主流的生成式文本摘要方法,采用基于深度学习的序列到序列框架,但存在并行能力不足或效率低的缺陷,并且在生成摘要的过程中存在准确率低和重复率高的问题.为解决上述问题,提出一种融合BERT预训练模型和卷积门控单元的生成式摘要方法.该方法基于改进Transformer模型,在编码器阶段充分利用BERT预先训练的大规模语料,代替RNN提取文本的上下文表征,结合卷积门控单元对编码器输出进行信息筛选,筛选出源文本的关键内容;在解码器阶段,设计3种不同的Transformer,旨在探讨BERT预训练模型和卷积门控单元更为有效的融合方式,以此提升文本摘要生成性能.实验采用ROUGE值作为评价指标,在LCSTS中文数据集和CNN/Daily Mail英文数据集上与目前主流的生成式摘要方法进行对比的实验,结果表明所提出方法能够提高摘要的准确性和可读性.     

Assessing sentence scoring techniques for extractive text summarization

Text summarization is the process of automatically creating a shorter version of one or more text documents. It is an important way of finding relevant information in large text libraries or in the Internet. Essentially, text summarization techniques are classified as Extractive and Abstractive. Extractive techniques perform text summarization by selecting sentences of documents according to some criteria. Abstractive summaries attempt to improve the coherence among sentences by eliminating redundancies and clarifying the contest of sentences. In terms of extractive summarization, sentence scoring is the technique most used for extractive text summarization. This paper describes and performs a quantitative and qualitative assessment of 15 algorithms for sentence scoring available in the literature. Three different datasets (News, Blogs and Article contexts) were evaluated. In addition, directions to improve the sentence extraction results obtained are suggested.     

基于改进seq2seq模型的英汉翻译研究

目前机器翻译主要对印欧语系进行优化与评测,很少有对中文进行优化的,而且机器翻译领域效果最好的基于注意力机制的神经机器翻译模型-seq2seq模型也没有考虑到不同语言间语法的变换。提出一种优化的英汉翻译模型,使用不同的文本预处理和嵌入层参数初始化方法,并改进seq2seq模型结构,在编码器和解码器之间添加一层用于语法变化的转换层。通过预处理,能缩减翻译模型的参数规模和训练时间20%,且翻译性能提高0.4 BLEU。使用转换层的seq2seq模型在翻译性能上提升0.7～1.0 BLEU。实验表明,在规模大小不同的语料英汉翻译任务中,该模型与现有的基于注意力机制的seq2seq主流模型相比,训练时长一致,性能提高了1～2 BLEU。     

基于双编码器的短文本自动摘要方法

针对当前生成式文本摘要方法存在的语义信息利用不充分、摘要精度不够等问题,提出一种基于双编码器的文本摘要方法。首先,通过双编码器为序列映射（Seq2Seq）架构提供更丰富的语义信息,并对融入双通道语义的注意力机制和伴随经验分布的解码器进行了优化研究;然后,在词嵌入生成技术中融合位置嵌入和词嵌入,并新增词频-逆文档频率（TF-IDF）、词性（POS）、关键性得分（Soc）,优化词嵌入维度。所提方法对传统序列映射Seq2Seq和词特征表示进行优化,在增强模型对语义的理解的同时,提高了摘要的质量。实验结果表明,该方法在Rouge评价体系中的表现相比传统伴随自注意力机制的递归神经网络方法（RNN+atten）和多层双向伴随自注意力机制的递归神经网络方法（Bi-MulRNN+atten）提高10~13个百分点,其文本摘要语义理解更加准确、生成效果更好,拥有更好的应用前景。     

基于分层编码的深度增强学习对话生成

面向对话生成问题,提出一种构建对话生成模型的方法--基于分层编码的深度增强学习对话模型（EHRED）,用以解决当前标准序列到序列（seq2seq）结构采用最大似然函数作为目标函数所带来的易生成通用回答的问题。该方法结合了分层编码和增强学习技术,利用分层编码来对多轮对话进行建模,在标准seq2seq的基础上新增了中间层来加强对历史对话语句的记忆,而后采用了语言模型来构建奖励函数,进而用增强学习中的策略梯度方法代替原有的最大似然损失函数进行训练。实验结果表明EHRED能生成语义信息更丰富的回答,在标准的人工测评中,其效果优于当前广泛采用的标准seq2seq循环神经网络（RNN）模型5.7~11.1个百分点。     

主题感知的长文本自动摘要算法

长文本摘要生成一直是自动摘要领域的难题。现有方法在处理长文本的过程中,存在准确率低、冗余等问题。鉴于主题模型在多文档摘要中的突出表现,将其引入到长文本摘要任务中。另外,目前单一的抽取式或生成式方法都无法应对长文本的复杂情况。结合两种摘要方法,提出了一种针对长文本的基于主题感知的抽取式与生成式结合的混合摘要模型。并在TTNews和CNN/Daily Mail数据集上验证了模型的有效性,该模型生成摘要ROUGE分数与同类型模型相比提升了1~2个百分点,生成了可读性更高的摘要。     

Concept generalization and fusion for abstractive sentence generation

Text summarization is either extractive or abstractive. Extractive summarization is to select the most salient pieces of information (words, phrases, and/or sentences) from a source document without adding any external information. Abstractive summarization allows an internal representation of the source document so as to produce a faithful summary of the source. In this case, external text can be inserted into the generated summary. Because of the complexity of the abstractive approach, the vast majority of work in text summarization has adopted an extractive approach.In this work, we focus on concepts fusion and generalization, i.e. where different concepts appearing in a sentence can be replaced by one concept which covers the meanings of all of them. This is one operation that can be used as part of an abstractive text summarization system. The main goal of this contribution is to enrich the research efforts on abstractive text summarization with a novel approach that allows the generalization of sentences using semantic resources. This work should be useful in intelligent systems more generally since it introduces a means to shorten sentences by producing more general (hence abstractions of the) sentences. It could be used, for instance, to display shorter texts in applications for mobile devices. It should also improve the quality of the generated text summaries by mentioning key (general) concepts. One can think of using the approach in reasoning systems where different concepts appearing in the same context are related to one another with the aim of finding a more general representation of the concepts. This could be in the context of Goal Formulation, expert systems, scenario recognition, and cognitive reasoning more generally.We present our methodology for the generalization and fusion of concepts that appear in sentences. This is achieved through (1) the detection and extraction of what we define as generalizable sentences and (2) the generation and reduction of the space of generalization versions. We introduce two approaches we have designed to select the best sentences from the space of generalization versions. Using four NLTK¹ corpora, the first approach estimates the “acceptability” of a given generalization version. The second approach is Machine Learning-based and uses contextual and specific features. The recall, precision and F1-score measures resulting from the evaluation of the concept generalization and fusion approach are presented.     

面向连贯性强化的无真值依赖文本摘要模型

自动文本摘要技术旨在凝练给定文本,以篇幅较短的摘要有效反映出原文核心内容.现阶段,生成型文本摘要技术因能够以更加灵活丰富的词汇对原文进行转述,已成为文本摘要领域的研究热点.然而,现有生成型文本摘要模型在产生摘要语句时涉及对原有词汇的重组与新词的添加,易造成摘要语句不连贯、可读性低.此外,通过传统基于已标注数据的有监督训...     

Improving the readability and saliency of abstractive text summarization using combination of deep neural networks equipped with auxiliary attention mechanism

Rapid and exponential development of textual data in recent years has yielded to the need for automatic text summarization models which aim to automatically condense a piece of text into a shorter version. Although various unsupervised and machine learning-based approaches have been introduced for text summarization during the last decades, the emergence of deep learning has made remarkable progress in this field. However, deep learning-based text summarization models are still in their early steps of development and their potential has yet to be fully explored. Accordingly, a novel abstractive summarization model is proposed in this paper which utilized the combination of convolutional neural network and long short-term memory integrated with auxiliary attention in its encoder to increase the saliency and coherency of generated summaries. The proposed model was validated on CNN\Daily Mail and DUC-2004 datasets and empirical results indicated that not only the proposed model outperformed existing models in terms of ROUGE metric but also its generated summaries had higher saliency and readability compared to the baseline model according to human evaluation.

     

SFExt-PGAbs:两阶段长文档摘要模型

针对抽取式方法、生成式方法在长文档摘要上的流畅性、准确性缺陷以及在文档编码前截断原始文档造成的重要信息缺失问题,提出一种两阶段长文档摘要模型SFExt-PGAbs,由次模函数抽取式摘要SFExt与指针生成器生成式摘要PGAbs组成。SFExt-PGAbs模拟人类对长文档进行摘要的过程,首先使用SFExt在长文档中抽取出重要句子,过滤不重要且冗余的句子形成过渡文档,然后PGAbs接收过渡文档作为输入以生成流畅且准确的摘要。为获取与原始文档中心思想更为接近的过渡文档,在传统SFExt中拓展出位置重要性、准确性两个子方面,同时设计新的贪心算法。为研究不同特征提取器对生成摘要质量的影响,在PGAbs中应用两种循环神经网络。实验结果显示,在CNNDM测试集上,SFExt-PGAbs相较于基线模型生成了更为流畅、准确的摘要,ROUGE指标有较大提升。同时,子方面拓展后的SFExt也能抽取得到更准确的摘要。     

Effective approaches to combining lexical and syntactical information for code summarization

Natural language summaries of source codes are important during software development and maintenance. Recently, deep learning based models have achieved good performance on the task of automatic code summarization, which encode token sequence or abstract syntax tree (AST) of code with neural networks. However, there has been little work on the efficient combination of lexical and syntactical information of code for better summarization quality. In this paper, we propose two general and effective approaches to leveraging both types of information: a convolutional neural network that aims to better extract vector representation of AST node for downstream models; and a Switch Network that learns an adaptive weight vector to combine different code representations for summary generation. We integrate these approaches into a comprehensive code summarization model, which includes a sequential encoder for token sequence of code and a tree based encoder for its AST. We evaluate our model on a large Java dataset. The experimental results show that our model outperforms several state-of-the-art models on various metrics, and the proposed approaches contribute a lot to the improvements.     

融合复制机制和input-feeding方法的中文自动摘要模型

针对中文自动摘要准确率不高的问题,在含有注意力机制的序列到序列（sequence-to-sequence,seq2seq）基础模型的解码器中融合了复制机制和input-feeding方法,提出了准确率更高的中文自动摘要模型。首先,该模型使用指针网络将出现在源序列中的OOV（out-of-vocabulary）词扩展到固定词典,以实现从源序列复制OOV词到生成序列中;其次,input-feeding方法用于跟踪已生成序列的注意力决定信息以提升模型输出准确率。在NLPCC2018数据集上的实验结果表明,与基础模型相比,所提出模型获得了更高的ROUGE得分,验证了该方法的可行性。