Cooperating rewrite processes for natural-language analysis

General methods for understanding a natural language based on the intensive use of rewrite rules and on the existence of several cooperating processes are put forward. The choice of Horn-clause logic as the underlying formalism for semantic representations, together with the employment of unification as pattern-matching procedure and depth-first search with backtracking were derived from logic-programming ideas, in particular from the use of PROLOG. Several examples are presented to illustrate how these methods work.     

Generate domain-specific sentiment lexicon for review sentiment analysis

Multimedia Tools and Applications - Lexicon-based approaches for review sentiment analysis have attracted significant attention in recent years. Lots of sentiment lexicon generation methods have...     

Discourse strategies for generating natural-language text

If a generation system is to produce text in response to a given communicative goal, it must be able to determine what to include in its text and how to organize this information so that it can be easily understood. In this paper, a computational model of discourse strategies is presented that can be used to guide the generation process in its decisions about what to say next. The model is based on an analysis of naturally occurring texts and represents strategies that can be used for three communicative goals: define, compare, and describe. We show how this model has been implemented in text, a system which generates paragraph-length responses to questions about database structure.     

Availability analysis for the design of distributed processing networks

This paper presents an analytical model for evaluating the availability of distributed processing systems. The model, which is based on a well-known probability formula, has been implemented in FORTRAN and used for the design of a real-time distributed processing system that requires high availability. It is applied to a sample distributed processing system to illustrate its usefulness and to provide good insight into the design of highly available distributed processing systems. Effects of homogeneous/heterogeneous clustering and standby spare are illustrated. The concept of availability threshold where the distributed system may gain or lose availability over the individual processor availability is introduced. It has proven to be a valuable tool for the tradeoffs in the allocation of system requirements. The concluding remarks summarize the lessons learned and suggest cost-effective ways to achieve higher availability for the design of distributed processing systems.     

SentiTurkNet: a Turkish polarity lexicon for sentiment analysis

Sentiment analysis aims to extract the sentiment polarity of given segment of text. Polarity resources that indicate the sentiment polarity of words are commonly used in different approaches. While English is the richest language in regard to having such resources, the majority of other languages, including Turkish, lack polarity resources. In this work we present the first comprehensive Turkish polarity resource, SentiTurkNet, where three polarity scores are assigned to each synset in the Turkish WordNet, indicating its positivity, negativity, and objectivity (neutrality) levels. Our method is general and applicable to other languages. Evaluation results for Turkish show that the polarity scores obtained through this method are more accurate compared to those obtained through direct translation (mapping) from SentiWordNet.     

Annotated news corpora and a lexicon for sentiment analysis in Slovene

In this study, we introduce Slovene web-crawled news corpora with sentiment annotation on three levels of granularity: sentence, paragraph and document levels. We describe the methodology and tools that were required for their construction. The corpora contain more than 250,000 documents with political, business, economic and financial content from five Slovene media resources on the web. More than 10,000 of them were manually annotated as negative, neutral or positive. All corpora are publicly available under a Creative Commons copyright license. We used the annotated documents to construct a Slovene sentiment lexicon, which is the first of its kind for Slovene, and to assess the sentiment classification approaches used. The constructed corpora were also utilised to monitor within-the-document sentiment dynamics, its changes over time and relations with news topics. We show that sentiment is, on average, more explicit at the beginning of documents, and it loses sharpness towards the end of documents.     

Morphological image sequence processing

We present a morphological multi-scale method for image sequence processing, which results in a truly coupled spatio-temporal anisotropic diffusion. The aim of the method is not to smooth the level-sets of single frames but to denoise the whole sequence while retaining geometric features such as spatial edges and highly accelerated motions. This is obtained by an anisotropic spatio-temporal level-set evolution, where the additional artificial time variable serves as the multi-scale parameter. The diffusion tensor of the evolution depends on the morphology of the sequence, given by spatial curvatures of the level-sets and the curvature of trajectories (=acceleration) in sequence-time. We discuss different regularization techniques and describe an operator splitting technique for solving the problem. Finally we compare the new method with existing multi-scale image sequence processing methodologies.     

A lexicon for knowledge-based MT

In knowledge-based machine translation (KBMT), the lexicon can be specified and acquired only in close connection with the specification and acquisition of the world model (ontology) and the specification of the text meaning representation (interlingua) language. The former supplies the atoms for the specification of text meaning and provides world knowledge to support the inference processes necessary for a variety of disambiguation and meaning assignment operations. The latter is necessary for the formulation of the semantic zone of the lexicon entries, which can be viewed as containing the static building blocks of the text meaning representation. This is the view taken in the Mikrokosmos KBMT project.     

Template design for file processing applications

Even though the data processing domain is relatively well-defined, application-specific details vary too widely to conceive of totally automating programming and specification support. Conventional tools lack the semantic knowledge to constructively support software development. On the other hand, it is possible to design a prompting and guiding system that can elicit the application details. The limits of automated programming support possible for file and data processing applications were identified by means of a case study of COBOL and PL/I programs. The data gathered from this study facilitated the definition of seventeen widely applicable templates. The resulting template set has immediate value as a manual specification and documentation tool. The templates encourage programmers to make well-considered design decisions and enforce standardized documentation.     

融合情感词典与上下文语言模型的文本情感分析

词嵌入技术在文本情感分析中发挥着重要的作用,但是传统的Word2Vec、GloVe等词嵌入技术会产生语义单一的问题。针对上述问题提出了一种融合情感词典与上下文语言模型ELMo的文本情感分析模型SLP-ELMo。首先,利用情感词典对句子中的单词进行筛选;其次,将筛选出的单词输入字符卷积神经网络（char-CNN）,从而产生每个单词的字符向量;然后,将字符向量输入ELMo模型进行训练;此外,在ELMo向量的最后一层加入了注意力机制,以便更好地训练词向量;最后,将词向量与ELMo向量并行融合并输入分类器进行文本情感分类。与现有的多个模型对比,所提模型在IMDB和SST-2这两个数据集上均得到了更高的准确率,验证了模型的有效性。     

基于情感词典和堆叠残差的双向长短期记忆网络的情感分析

情感分析作为自然语言处理（NLP）的细分研究方向经历了使用情感词典、机器学习和深度学习分析的发展过程。针对使用一般化的深度学习模型作为文本分类器对于特定领域的网络评论类型的文本的分析的精准度较低,训练时发生过拟合现象以及情感词典覆盖率低、编纂工作量大的问题,提出了基于情感词典和堆叠残差的双向长短期记忆（Bi-LSTM）网络的情感分析模型。首先,借助情感词典中情感词的设计覆盖“教育机器人”研究领域内的专业词汇,从而弥补Bi-LSTM模型在分析此类文本时精准度的不足;然后,使用Bi-LSTM和SnowNLP来降低情感词典的编纂体量。长短期记忆（LSTM）网络的“记忆门”“遗忘门”结构可以在保证充分考虑评论文本中的前后词语的关联性的同时,适时选择遗忘一些已分析词语,从而避免反向传播时的梯度爆炸问题。而在将堆叠残差的Bi-LSTM引入后,不仅使得模型的层数加深至8层,而且还使残差网络避免了叠加LSTM时会导致的“退化”问题;最后,通过适当设置和调整两部分的得分权重,并将总分使用Sigmoid激活函数标准化到［0,1］的区间上,按照［0,0.5］,（0.5,1］的区间划分分别表示负面和正面情绪,完成情感分类。实验结果表明,在“教育机器人”评论数据集中,所提模型对于情感分类准确率相较于标准的LSTM模型提升了约4.5个百分点,相较于BERT提升了约2.0个百分点。综上,所提模型将基于情感词典和深度学习模型的情感分类方法一般化;而通过修改情感词典中的情感词汇并适当调整深度学习模型的结构和层数,所提模型可以应用于电子商务平台中各类商品的购物评价的精确情感分析,从而帮助企业洞悉消费者的购物心理和市场需求,同时也可以为消费者提供商品质量的一种参考标准。     

多成分正则化约束的断层图像重建算法

压缩感知理论为低采样率下稀疏性信号的高质量重建提供理论基础和广泛的应用前景,然而现有大部分研究通常假设原始图像是分段光滑的,不适用于纹理丰富的图像。基于多成分分析理论将原始图像分为卡通和纹理两部分,并且利用各成分在不同变换下稀疏的性质,提出一种多成分正则化的断层图像重建模型。采用分裂Bregman方法将多正则项解耦合,分解成最小二乘问题和去噪问题求解,提出该模型的交替迭代算法。对MRI和CT图像进行仿真实验,并同新近的多种算法进行比较。实验结果表明,该模型能够较好地保持图像纹理等中小尺度信息,收敛速度快。     

Morphological analysis for investigating artistic images

This paper describes an approach for automatically analyzing the alterations of an original artwork during its reproduction. The overall deformation of the artwork is modelled by a piecewise linear model, where regions of the artwork that feature similar alterations are automatically inferred and assigned to the different model components. Model complexity, that is, the required number of affine components required for registration, is automatically estimated using a statistical stability analysis. The main challenge is to simultaneously solve three tasks: (i) inferring the correspondences between both shapes, (ii) identifying the groups in the image that share the same transformation, and (iii) estimating the transformation of these groups. Our approach is tested on controlled scenarios as well as on real historical images.