自然语言处理：方法与技术

无论是一项科学前沿研究，还是一门应用学科，自然语言处理已引起人们越来越多的关注与重视，本文讨论了它的理论基础，主要计算模式，系统的实现技术，并扼要介绍了这一领域新发展动向。     

自然语言处理中的语句语义表示格式研究

语句语义表示格式(语句格式)体现了语句各组成成分之间的组合关系,可给计算机提供指导性的知识。该文在HNC理论框架指导下,对目前普遍采用的语句格式提出了一种全面改进的方案,给出了新的语句格式编码,通过实例阐述了带来的变化。测试数据表明,新语句格式具有很强的通用性和非常好的表示能力,更有利于计算机的处理。     

Scheduling aspects in keyword extraction problem

The amount of big data collected during human–computer interactions requires natural language processing (NLP) applications to be executed efficiently, especially in parallel computing environments. Scalability and performance are critical in many NLP applications such as search engines or web indexers. However, there is a lack of mathematical models helping users to design and apply scheduling theory for NLP approaches. Moreover, many researchers and software architects reported various difficulties related to common NLP benchmarks. Therefore, this paper aims to introduce and demonstrate how to apply a scheduling model for a class of keyword extraction approaches. Additionally, we propose methods for the overall performance evaluation of different algorithms, which are based on processing time and correctness (quality) of answers. Finally, we present a set of experiments performed in different computing environments together with obtained results that can be used as reference benchmarks for further research in the field.     

文字可视化技术在虚拟环境中的应用(英文)

文字和图像是人类日常交流常用的两种方式,同时它们之间有着很深的联系。我们常常将图像转化为文字(如通过文字来描述一副绘画作品),反之亦然(如通过一幅画来再现一段故事),但这个过程的转换对计算机系统而言则是一个挑战。随着自然语言处理技术和计算机图形技术的发展,人们逐渐意识到结合这两种技术,并通过整合人与人之间的交流方式来进一步提高人机界面的可能性,使之成为信息可视化领域一个新的研究方向之一。本文通过回顾基于文字到可视化技术的应用发展进程,并在对当前的相关研究进行了探讨的基础上提出了我们对故事可视化系统开发的解决方案。     

Natural language processing: a prolog perspective

Natural language processing (NLP) is a vibrant field of interdisciplinary Computer Science research. Ultimately, NLP seeks to build intelligence into software so that software will be able to process a natural language as skillfully and artfully as humans. Prolog, a general purpose logic programming language, has been used extensively to develop NLP applications or components thereof. This report is concerned with introducing the interested reader to the broad field of NLP with respect to NLP applications that are built in Prolog or from Prolog components.     

计算机数控系统光滑时间最优轨迹规划

基于控制向量参数化(CVP)方法, 研究了计算机数控(CNC)系统光滑时间最优轨迹规划方法. 通过在规划问题中引入加加速度约束, 实现轨迹的光滑给进. 引入时间归一化因子, 将加加速度约束的时间最优轨迹规划问题转化为固定时间的一般性最优控制问题. 以路径参数对时间的三阶导数(伪加加速度)和终端时刻为优化变量, 并采用分段常数近似伪加加速度, 将最优控制问题转化为一般的非线性规划(NLP)问题进行求解. 针对加加速度、加速度等过程不等式约束, 引入约束凝聚函数, 将过程约束转化为终端时刻约束, 从而显著减少约束计算. 构造目标和约束函数的Hamiltonian函数, 利用伴随方法获得求解NLP问题所需的梯度.     

自然语言处理在信息检索中的应用综述

在信息检索^①发展的过程中,研究者们不断尝试着将自然语言处理应用到检索里,希望能够为检索效果提高带来帮助。然而这些尝试的结果大多和研究者们最初的设想相反,自然语言处理在大多数情况下没有改进信息检索效果,甚至反而起了负面作用。即便有一些帮助,也往往是微小的,远远不如自然语言处理所需要的计算消耗那么大。研究者们对这些现象进行了分析,认为: 自然语言处理更适合于应用在需要精确结果的任务中,例如问答系统、信息抽取等;自然语言处理需要针对信息检索进行优化才可能发挥积极作用。最新的一些进展(例如在语言模型中加入自然语言处理)在一定程度上印证了这一结论。     

一个通用的混合非线性规划问题的演化算法

提出了一种新的求解非线性规划问题的演化算法，它是在郭涛算法的基础上提出的，新算法的主要特点是引入了变维子空间，加入了子空间搜索过程和规范化约束条件以及增加了处理带等式约束的实数规划，整数规划，0－1规划和混合整数规划问题的功能，使之成为一种求解非线性规划（NLP）问题的通用算法，数值实验表明，新算法不仅是一种通用的算法，而且与已有算法的计算结果相比，其解的精确度也最好。     

Knowledge Mining: A Cross-disciplinary Survey

Machine Intelligence Research - Knowledge mining is a widely active research area across disciplines such as natural language processing (NLP), data mining (DM), and machine learning (ML). The...     

A Robust Algorithm for Solving Nonlinear Programming Problems

In this paper, we introduce a new algorithm for solving nonlinear programming (NLP) problems. It is an extension of Guo's algorithm [1] which possesses enhanced capabilities for solving NLP problems. These capabilities include: a) extending the variable subspace, b) adding a search process over subspaces and normalized constraints, c) using an adaptive penalty function, and d) adding the ability to deal with integer NLP problems, 0-1 NLP problems, and mixed-integer NLP problems which have equality constraints. These four enhancements increase the capabilities of the algorithm to solve nonlinear programming problems in a more robust and universal way. This paper will present results of numerical experiments which show that the new algorithm is not only more robust and universal than its competitors, but also its performance level is higher than any others in the literature.     

NLP: Managing a Creative Revolution

Neuro-Linguistic Programming (NLP) is a technology for personal excellence based on research investigation. The process involves sensitivity to effective behaviours, and the development of those behaviours. Some presuppositions of NLP are examined, together with practical components and evidence of its effectiveness. The author explains NLP, drawing on organisational experiences.     

A study of one class of NLP problems arising in parametric semi-infinite programming

The paper deals with a nonlinear programming (NLP) problem that depends on a finite number of integers (parameters). This problem has a special form, and arises as an auxiliary problem in study of solutions' properties of parametric semi-infinite programming (SIP) problems with finitely representable compact index sets. Therefore, it is important to provide a deep study of this NLP problem and its properties w.r.t. the values of the parameters. We are especially interested in the case when optimal solutions of the NLP problem satisfy certain properties due to some specific requirements arising in parametric SIP. We establish the values of the parameters for which optimal solutions of the corresponding NLP problem fulfil the needed properties, and suggest an algorithm that determines the right values of the parameters. An example is proposed to illustrate the application of the algorithm.     

Smooth time-optimal tool trajectory generation for CNC manufacturing systems

An optimization approach is proposed in this paper for generating smooth and time-optimal path constrained tool trajectory for Cartesian computer numerical control (CNC) manufacturing systems. The desired smooth time-optimal trajectory generation (STOTG) problem is formulated as a general optimal control problem. And axis jerk (derivative of acceleration with respect to time) constraints are introduced into this problem to remove discontinuities of the acceleration profiles. The desired smoothness of the trajectory can be accomplished by adjusting the values of jerk constraints. A control vector parameterization (CVP) method is applied to convert the optimal control problem into a nonlinear programming (NLP) problem which can be solved conveniently and effectively. The third derivative of the path parameter with respect to time (pseudo-jerk) and jerk act as optimization variables. The pseudo-jerk is approximated as piecewise constant, thus for at least second-order continuous parametric path, the resulted optimized trajectory with respect to time is also at least second-order continuous. Sequential quadratic programming (SQP) method is used to solve the NLP problem, through which numerical solution is obtained. Non-smooth (i.e. without considering jerk constraints) time-optimal trajectory generation (non-STOTG) problem is also considered in this paper for the purpose of comparison. Solutions of time-optimal trajectory generation (TOTG) problems for two test paths are performed to verify the effectiveness of the proposed approach.     

Real-time nonlinear programming by amplitude modulation

The nonlinear programming (NLP) is an optimization technique that minimizes arbitrary nonlinear cost functions. To apply the NLP to real-time applications, however, the estimation of the gradient of the cost function is remained as a challenge. The extremum-seeking control (ESC) optimizes the cost functions in real-time, but it involves the complicated design of filters for simultaneous estimation of the gradient. In this paper, a complementary method that optimizes an arbitrary multivariable cost function in real-time is proposed. Taking the advantages of both NLP and ESC, the variables are updated by the steepest descent method of NLP, while the gradient of the cost function is continuously estimated by the amplitude modulation as in ESC. Unlike the ESC, the proposed method does not require the design of complicated filters. The optimization performance is verified by simulations on time-varying and noisy cost functions, as well as automatic controller tuning applications.     

A survey of discourse parsing

Discourse parsing is an important research area in natural language processing (NLP), which aims to parse the discourse structure of coherent sentences. In this survey, we introduce several different kinds of discourse parsing tasks, mainly including RST-style discourse parsing, PDTB-style discourse parsing, and discourse parsing for multiparty dialogue. For these tasks, we introduce the classical and recent existing methods, especially neural network approaches. After that, we describe the applications of discourse parsing for other NLP tasks, such as machine reading comprehension and sentiment analysis. Finally, we discuss the future trends of the task.     

Improving Open Information Extraction with Distant Supervision Learning

Neural Processing Letters - Open information extraction (Open IE), as one of the essential applications in the area of Natural Language Processing (NLP), has gained great attention in recent years....     

Development of Web Service for Japanese Text Triplification

Linked Open Data (LOD) is recently attracting attention as a vast amount of distributed knowledge base on the Web. Thus, semi-structured data such as tables and hierarchical data in several domains have been triplified to the LOD. Moreover, triplification of unstructured data such as text and sensor data is actively studied in the research area. We thus developed a Web API for extracting triples from natural sentences, which is useful for the promotion of the LOD. In this paper, we first describe the service specification, its technical background, and the evaluation of the current extraction accuracy, and then introduce several use cases of the service. Although this service adopts a novel combination of a method using ontology-based rules and a machine learning method using Conditional Random Field based on a probability distribution, the main contributions are in the practical aspect, that is, mashup of several natural language processing (NLP) techniques for the text triplification, and its deployment as a public Web API, so that non-experts of NLP can easily use it.     

Reduction constraints for the global optimization of NLPs

Convergence of branch‐and‐bound algorithms for the solution of NLPs is obtained by finding ever‐nearer lower and upper bounds to the objective function. The lower bound is calculated by constructing a convex relaxation of the NLP. Reduction constraints are new linear problem constraints which are (a) linearly independent from the existing constraints; (b) redundant with reference to the original NLP formulation; (c) not redundant with reference to its convex relaxation. Thus, they can be successfully employed to reduce the feasible region of the convex relaxation without cutting the feasible region of the original NLP.     

生成对抗网络在各领域应用研究进展

随着深度学习的快速发展, 生成式模型领域也取得了显著进展. 生成对抗网络(Generative adversarial network, GAN)是一种无监督的学习方法, 它是根据博弈论中的二人零和博弈理论提出的. GAN具有一个生成器网络和一个判别器网络, 并通过对抗学习进行训练. 近年来, GAN成为一个炙手可热的研究方向. GAN不仅在图像领域取得了不错的成绩, 还在自然语言处理(Natural language processing, NLP)以及其他领域崭露头角. 本文对GAN的基本原理、训练过程和传统GAN存在的问题进行了阐述, 进一步详细介绍了通过损失函数的修改、网络结构的变化以及两者结合的手段提出的GAN变种模型的原理结构, 其中包括: 条件生成对抗网络(Conditional GAN, CGAN)、基于Wasserstein 距离的生成对抗网络(Wasserstein-GAN, WGAN)及其基于梯度策略的WGAN (WGAN-gradient penalty, WGAN-GP)、基于互信息理论的生成对抗网络(Informational-GAN, InfoGAN)、序列生成对抗网络(Sequence GAN, SeqGAN)、Pix2Pix、循环一致生成对抗网络(Cycle-consistent GAN, Cycle GAN)及其增强Cycle-GAN (Augmented CycleGAN). 概述了在计算机视觉、语音与NLP领域中基于GAN和相应GAN变种模型的基本原理结构, 其中包括: 基于CGAN的脸部老化应用(Face aging CGAN, Age-cGAN)、双路径生成对抗网络(Two-pathway GAN, TP-GAN)、表示解析学习生成对抗网络(Disentangled representation learning GAN, DR-GAN)、对偶学习生成对抗网络(DualGAN)、GeneGAN、语音增强生成对抗网络(Speech enhancement GAN, SEGAN)等. 介绍了GAN在医学、数据增强等领域的应用情况, 其中包括: 数据增强生成对抗网络(Data augmentation GAN, DAGAN)、医学生成对抗网络(Medical GAN, MedGAN)、无监督像素级域自适应方法(Unsupervised pixel-level domain adaptation method, PixelDA). 最后对GAN未来发展趋势及方向进行了展望.