NEW-Synative

“旧词”在计算机语言中的“新意”语言词汇是承载文化信息、反映人类社会生活的最明显的标志，特定时期的新词语是这一时期人们思维方式和生活方式的集中反映。

随着网络技术的发展，产生了大量的新词语。

计算机英语中也有一些词语在已有词汇的基础上增添了新的含义，即“旧词”添“新意”。

比如“table”和“form”，通用的英文含义为“表格，表”，但学习过ACCESS数据库的人都知道，在数据库中，二者表示的具体含义并不相同，前者为“数据库中的子表格”，后者指“用于简化数据输入和查询的窗体”。

下面将一些在计算机英语中出现“新意”的常见“旧词”分类列举如下：一、人名、地名Delphi（特尔斐）：古希腊城市名，被古希腊人当成世界的中心，因有阿波罗神殿而出名。

而在计算机英语中Delphi则指美国宝兰（Borland）公司的一种可视化、面向对象、事件驱动的电脑编程语言。

Java：初见此词有些人会想到印度尼西亚的爪哇岛，喜欢喝咖啡的人会想到“爪哇咖啡”，因为Java岛以盛产咖啡豆闻名于世。

而Java在计算机英语中则指由美国太阳电脑公司研制开发的一种程序语言。

此程序语言主要用于撰写、制作多媒体文档，尤其是动画与动态图像。

以Java命名此程序语言还有一段有趣的故事——据说，Java刚刚被设计出来时，设计人员在公司附近一家名为Java的咖啡店商讨起名问题，长时间都没有达成一致，最后有人提议索性就用这家咖啡店的店名来命名。

Mosaic：此词原是形容词，意思是“摩西的”。

摩西是旧约圣经中的先知。

在计算机英语中Mosaic被网络公司的安德森用来命名其所开发的浏览器。

Troglodyte：源于希腊语，意思是“（史前的）穴居人；隐居者”。

现在被用来指那些终日在电脑房中搞计算机的人，这些人像洞穴中的蝙蝠一样，日夜颠倒，废寝忘食。

二、人、动物Bug：通用的英文含义为“（英国）臭虫；（美口）虫，昆虫”。

在计算机英语中指的是由于设计疏忽而遗留在软件中的漏洞或隐藏在软件中而使软件或操作系统不能正常运行的错误程序。

Java中冷门的synthetic关键字原理解读看JAVA的反射时，看到有个synthetic ，还有⼀个⽅法isSynthetic() 很好奇，就了解了⼀下：1.定义Any constructs introduced by a Java compiler that do not have a corresponding construct in the source code must be marked as synthetic, except for default constructors, the class initialization method, and the values and valueOf methods of the Enum class.⼤意为：由java编译器⽣成的（除了像默认构造函数这⼀类的）⽅法，或者类2.实例既然知道synthetic⽅法和synthetic类是由编译器⽣成的，那到底编译器会怎么⽣成这些东西，⼜在什么情况下会⽣成这些东西呢？先看⼀段代码：import static ng.System.out;public final class DemonstrateSyntheticMethods{public static void main(final String[] arguments){DemonstrateSyntheticMethods.NestedClass nested =new DemonstrateSyntheticMethods.NestedClass();out.println("String: " + nested.highlyConfidential);}private static final class NestedClass{private String highlyConfidential = "Don't tell anyone about me";private int highlyConfidentialInt = 42;private Calendar highlyConfidentialCalendar = Calendar.getInstance();private boolean highlyConfidentialBoolean = true;}}编译之后，可以看到三个⽂件：其中，最下⾯的这个类⽂件很好解释，就是我们的主class，中间的⽂件，是我们的内部类，上⾯的⽂件，后⾯再讲，我们先看⼀下中间这个内部类2.1 内部类的反编译结果⽤javap 反编译DemonstrateSyntheticMethods$NestedClass.class，得到如下结果：javap DemonstrateSyntheticMethods\$NestedClass.classCompiled from "DemonstrateSyntheticMethods.java"final class DemonstrateSyntheticMethods$NestedClass {DemonstrateSyntheticMethods$NestedClass(DemonstrateSyntheticMethods$1);static ng.String access$100(DemonstrateSyntheticMethods$NestedClass);}先把构造函数放⼀边，我们来看这个标⿊的⽅法access$100 这个是怎么回事呢？我们的源⽂件⾥找不到这个access⽅法啊？2.2 synthetic⽅法这个⽅法就是编译器⽣成的synthetic⽅法，读者不信的话，可以⽤method.isSynthetic() 去验证⼀下。

ChatGPT技术对话生成的主题一致性的保证方法ChatGPT是一种基于人工智能的技术，它可以生成自然语言的对话。

然而，在使用ChatGPT进行对话生成时，一个重要的问题是如何保证生成的对话主题一致性。

本文将探讨一些方法，帮助确保ChatGPT生成的对话具有一致的主题。

首先，我们可以通过限制ChatGPT的输入话题来保证对话的一致性。

通过引入一个话题模块，我们可以预先将对话限定在特定的主题范围内。

这样做可以防止ChatGPT在生成回复时跳跃到不相关的话题上。

例如，如果对话的主题是关于电影的，我们可以将对话模型的输入设置为电影相关的问题或提示，确保生成的回复始终与电影有关。

其次，我们可以使用上下文信息来维持对话的一致性。

ChatGPT可以通过上下文理解对话的前后文，并生成具有连贯性的回复。

在对话的每一轮中，ChatGPT可以通过对上一轮对话内容的理解来生成回复，从而确保回复与之前的对话主题保持一致。

这可以通过对模型进行预训练和微调来实现，让ChatGPT能够更好地理解上下文并生成连贯的回复。

此外，我们还可以使用评估指标来衡量对话的一致性，并对ChatGPT的回复进行评估和调整。

一种常用的评估指标是BLEU得分，它可以衡量生成的回复与参考回复之间的相似度。

通过设置合适的阈值，我们可以过滤掉那些与对话主题不一致的回复，并优先选择与主题一致的回复作为最终的生成结果。

另外，引入外部知识库或常识推理可以进一步提升ChatGPT的对话一致性。

通过将外部知识集成到ChatGPT中，模型可以更好地理解和生成与主题相关的回复。

例如，ChatGPT可以通过检索相关的知识或根据常识进行推理，以确保生成的回复与对话主题保持一致。

然而，需要注意的是，保证对话的一致性并不意味着完全限制ChatGPT的创造力。

有时候，一些略微偏离主题的回复也可以增加对话的多样性和趣味性。

因此，在确保对话一致性的同时，我们也需要平衡对话的多样性和创造性，以提供更好的对话体验。

intriguing properties of neural networks 精读Intriguing Properties of Neural NetworksIntroduction:Neural networks are a type of machine learning model inspired by the human brain's functioning. They are composed of interconnected nodes known as neurons that work together to process and analyze complex data. Neural networks have gained immense popularity due to their ability to learn, adapt, and make accurate predictions. In this article, we will delve into some of the intriguing properties of neural networks and explore how they contribute to their success in various fields.1. Non-linearity:One of the key properties of neural networks is their ability to model nonlinear relationships in data. Traditional linear models assume a linear relationship between input variables and the output. However, neural networks introduce non-linear activation functions that allow them to capture complex patterns and correlations. This property enables neural networks to excel in tasks such as image recognition, natural language processing, and voice recognition.2. Parallel Processing:Neural networks possess the remarkable ability to perform parallel processing. Unlike traditional algorithms that follow a sequential execution path, neural networks operate by simultaneously processing multiple inputs in parallel. This parallel architecture allows for faster and efficientcomputations, making neural networks suitable for handling large-scale datasets and real-time applications.3. Distributed Representation:Neural networks utilize distributed representation to process and store information. In traditional computing systems, data is stored in a centralized manner. However, neural networks distribute information across interconnected neurons, enabling efficient storage, retrieval, and association of knowledge. This distributed representation enhances their ability to learn complex patterns and generalize from limited training examples.4. Adaptability:Neural networks exhibit a high degree of adaptability, enabling them to adjust their internal parameters and optimize their performance based on changing input. Through a process called backpropagation, neural networks continuously learn from the errors they make during training. This iterative learning process allows them to adapt to new data and improve their accuracy over time. The adaptability of neural networks makes them robust to noise, varying input patterns, and changing environments.5. Feature Extraction:Neural networks are adept at automatically extracting relevant features from raw data. In traditional machine learning approaches, feature engineering is often a time-consuming and manual process. However, neural networks can learn to identify important features directly from the input data. This property eliminates the need for human intervention and enables neuralnetworks to handle complex, high-dimensional data without prior knowledge or domain expertise.6. Capacity for Representation:Neural networks possess an impressive capacity for representation, making them capable of modeling intricate relationships in data. Deep neural networks, in particular, with multiple layers, can learn hierarchies of features, capturing both low-level and high-level representations. This property allows neural networks to excel in tasks such as image recognition, where they can learn to detect complex shapes, textures, and objects.Conclusion:The intriguing properties of neural networks, such as non-linearity, parallel processing, distributed representation, adaptability, feature extraction, and capacity for representation, contribute to their exceptional performance in various domains. These properties enable neural networks to tackle complex problems, make accurate predictions, and learn from diverse datasets. As researchers continue to explore and enhance the capabilities of neural networks, we can expect these models to revolutionize fields such as healthcare, finance, and autonomous systems.。

人工智能是一门新兴的具有挑战力的学科。

自人工智能诞生以来,发展迅速,产生了许多分支。

诸如强化学习、模拟环境、智能硬件、机器学习等。

但是,在当前人工智能技术迅猛发展,为人们的生活带来许多便利。

下面是搜索整理的人工智能英文参考文献的分享，供大家借鉴参考。

人工智能英文参考文献一：[1]Lars Egevad,Peter Str?m,Kimmo Kartasalo,Henrik Olsson,Hemamali Samaratunga,Brett Delahunt,Martin Eklund. The utility of artificial intelligence in the assessment of prostate pathology[J]. Histopathology,2020,76(6).[2]Rudy van Belkom. The Impact of Artificial Intelligence on the Activities ofa Futurist[J]. World Futures Review,2020,12(2).[3]Reza Hafezi. How Artificial Intelligence Can Improve Understanding in Challenging Chaotic Environments[J]. World Futures Review,2020,12(2).[4]Alejandro Díaz-Domínguez. How Futures Studies and Foresight Could Address Ethical Dilemmas of Machine Learning and Artificial Intelligence[J]. World Futures Review,2020,12(2).[5]Russell T. Warne,Jared Z. Burton. Beliefs About Human Intelligence in a Sample of Teachers and Nonteachers[J]. Journal for the Education of the Gifted,2020,43(2).[6]Russell Belk,Mariam Humayun,Ahir Gopaldas. Artificial Life[J]. Journal of Macromarketing,2020,40(2).[7]Walter Kehl,Mike Jackson,Alessandro Fergnani. Natural Language Processing and Futures Studies[J]. World Futures Review,2020,12(2).[8]Anne Boysen. Mine the Gap: Augmenting Foresight Methodologies with Data Analytics[J]. World Futures Review,2020,12(2).[9]Marco Bevolo,Filiberto Amati. The Potential Role of AI in Anticipating Futures from a Design Process Perspective: From the Reflexive Description of “Design” to a Discussion of Influences by the Inclusion of AI in the Futures Research Process[J]. World Futures Review,2020,12(2).[10]Lan Xu,Paul Tu,Qian Tang,Dan Seli?teanu. Contract Design for Cloud Logistics (CL) Based on Blockchain Technology (BT)[J]. Complexity,2020,2020.[11]L. Grant,X. Xue,Z. Vajihi,A. Azuelos,S. Rosenthal,D. Hopkins,R. Aroutiunian,B. Unger,A. Guttman,M. Afilalo. LO32: Artificial intelligence to predict disposition to improve flow in the emergency department[J]. CJEM,2020,22(S1).[12]A. Kirubarajan,A. Taher,S. Khan,S. Masood. P071: Artificial intelligence in emergency medicine: A scoping review[J]. CJEM,2020,22(S1).[13]L. Grant,P. Joo,B. Eng,A. Carrington,M. Nemnom,V. Thiruganasambandamoorthy. LO22: Risk-stratification of emergency department syncope by artificial intelligence using machine learning: human, statistics or machine[J]. CJEM,2020,22(S1).[14]Riva Giuseppe,Riva Eleonora. OS for Ind Robots: Manufacturing Robots Get Smarter Thanks to Artificial Intelligence.[J]. Cyberpsychology, behavior and social networking,2020,23(5).[15]Markus M. Obmann,Aurelio Cosentino,Joshy Cyriac,Verena Hofmann,Bram Stieltjes,Daniel T. Boll,Benjamin M. Yeh,Matthias R. Benz. Quantitative enhancement thresholds and machine learning algorithms for the evaluation of renal lesions using single-phase split-filter dual-energy CT[J]. Abdominal Radiology,2020,45(1).[16]Haytham H. Elmousalami,Mahmoud Elaskary. Drilling stuck pipe classification and mitigation in the Gulf of Suez oil fields using artificial intelligence[J]. Journal of Petroleum Exploration and Production Technology,2020,10(10).[17]Rüdiger Schulz-Wendtland,Karin Bock. Bildgebung in der Mammadiagnostik –Ein Ausblick <trans-title xml:lang="en">Imaging in breast diagnostics—an outlook [J]. Der Gyn?kologe,2020,53(6).</trans-title>[18]Nowakowski Piotr,Szwarc Krzysztof,Boryczka Urszula. Combining an artificial intelligence algorithm and a novel vehicle for sustainable e-waste collection[J]. Science of the Total Environment,2020,730.[19]Wang Huaizhi,Liu Yangyang,Zhou Bin,Li Canbing,Cao Guangzhong,Voropai Nikolai,Barakhtenko Evgeny. Taxonomy research of artificial intelligence for deterministic solar power forecasting[J]. Energy Conversion and Management,2020,214.[20]Kagemoto Hiroshi. Forecasting a water-surface wave train with artificial intelligence- A case study[J]. Ocean Engineering,2020,207.[21]Tomonori Aoki,Atsuo Yamada,Kazuharu Aoyama,Hiroaki Saito,Gota Fujisawa,Nariaki Odawara,Ryo Kondo,Akiyoshi Tsuboi,Rei Ishibashi,Ayako Nakada,Ryota Niikura,Mitsuhiro Fujishiro,Shiro Oka,Soichiro Ishihara,Tomoki Matsuda,Masato Nakahori,Shinji Tanaka,Kazuhiko Koike,Tomohiro Tada. Clinical usefulness of a deep learning‐based system as the first screening on small‐bowel capsule endoscopy reading[J]. Digestive Endoscopy,2020,32(4).[22]Masashi Fujii,Hajime Isomoto. Next generation of endoscopy: Harmony with artificial intelligence and robotic‐assisted devices[J]. Digestive Endoscopy,2020,32(4).[23]Roberto Verganti,Luca Vendraminelli,Marco Iansiti. Innovation and Design in the Age of Artificial Intelligence[J]. Journal of Product Innovation Management,2020,37(3).[24]Yuval Elbaz,David Furman,Maytal Caspary Toroker. Modeling Diffusion in Functional Materials: From Density Functional Theory to Artificial Intelligence[J]. Advanced Functional Materials,2020,30(18).[25]Dinesh Visva Gunasekeran,Tien Yin Wong. Artificial Intelligence in Ophthalmology in 2020: A Technology on the Cusp for Translation and Implementation[J]. Asia-Pacific Journal of Ophthalmology,2020,9(2).[26]Fu-Neng Jiang,Li-Jun Dai,Yong-Ding Wu,Sheng-Bang Yang,Yu-Xiang Liang,Xin Zhang,Cui-Yun Zou,Ren-Qiang He,Xiao-Ming Xu,Wei-De Zhong. The study of multiple diagnosis models of human prostate cancer based on Taylor database by artificial neural networks[J]. Journal of the Chinese Medical Association,2020,83(5).[27]Matheus Calil Faleiros,Marcello Henrique Nogueira-Barbosa,Vitor Faeda Dalto,JoséRaniery Ferreira Júnior,Ariane Priscilla Magalh?es Tenório,Rodrigo Luppino-Assad,Paulo Louzada-Junior,Rangaraj Mandayam Rangayyan,Paulo Mazzoncini de Azevedo-Marques. Machine learning techniques for computer-aided classification of active inflammatory sacroiliitis in magnetic resonance imaging[J]. Advances in Rheumatology,2020,60(1078).[28]Balamurugan Balakreshnan,Grant Richards,Gaurav Nanda,Huachao Mao,Ragu Athinarayanan,Joseph Zaccaria. PPE Compliance Detection using Artificial Intelligence in Learning Factories[J]. Procedia Manufacturing,2020,45.[29]M. Stévenin,V. Avisse,N. Ducarme,A. de Broca. Qui est responsable si un robot autonome vient à entra?ner un dommage ?[J]. Ethique et Santé,2020.[30]Fatemeh Barzegari Banadkooki,Mohammad Ehteram,Fatemeh Panahi,Saad Sh. Sammen,Faridah Binti Othman,Ahmed EL-Shafie. Estimation of Total Dissolved Solids (TDS) using New Hybrid Machine Learning Models[J]. Journal of Hydrology,2020.[31]Adam J. Schwartz,Henry D. Clarke,Mark J. Spangehl,Joshua S. Bingham,DavidA. Etzioni,Matthew R. Neville. Can a Convolutional Neural Network Classify Knee Osteoarthritis on Plain Radiographs as Accurately as Fellowship-Trained Knee Arthroplasty Surgeons?[J]. The Journal of Arthroplasty,2020.[32]Ivana Nizetic Kosovic,Toni Mastelic,Damir Ivankovic. Using Artificial Intelligence on environmental data from Internet of Things for estimating solar radiation: Comprehensive analysis[J]. Journal of Cleaner Production,2020.[33]Lauren Fried,Andrea Tan,Shirin Bajaj,Tracey N. Liebman,David Polsky,Jennifer A. Stein. Technological advances for the detection of melanoma: Part I. Advances in diagnostic techniques[J]. Journal of the American Academy of Dermatology,2020.[34]Mohammed Amoon,Torki Altameem,Ayman Altameem. Internet of things Sensor Assisted Security and Quality Analysis for Health Care Data Sets Using Artificial Intelligent Based Heuristic Health Management System[J]. Measurement,2020.[35]E. Lotan,C. Tschider,D.K. Sodickson,A. Caplan,M. Bruno,B. Zhang,Yvonne W. Lui. Medical Imaging and Privacy in the Era of Artificial Intelligence: Myth, Fallacy, and the Future[J]. Journal of the American College of Radiology,2020.[36]Fabien Lareyre,Cédric Adam,Marion Carrier,Juliette Raffort. Artificial Intelligence in Vascular Surgery: moving from Big Data to Smart Data[J]. Annals of Vascular Surgery,2020.[37]Ilesanmi Daniyan,Khumbulani Mpofu,Moses Oyesola,Boitumelo Ramatsetse,Adefemi Adeodu. Artificial intelligence for predictive maintenance in the railcar learning factories[J]. Procedia Manufacturing,2020,45.[38]Janet L. McCauley,Anthony E. Swartz. Reframing Telehealth[J]. Obstetrics and Gynecology Clinics of North America,2020.[39]Jean-Emmanuel Bibault,Lei Xing. Screening for chronic obstructive pulmonary disease with artificial intelligence[J]. The Lancet Digital Health,2020,2(5).[40]Andrea Laghi. Cautions about radiologic diagnosis of COVID-19 infection driven by artificial intelligence[J]. The Lancet Digital Health,2020,2(5).人工智能英文参考文献二：[41]K. Orhan,I. S. Bayrakdar,M. Ezhov,A. Kravtsov,T. ?zyürek. Evaluation of artificial intelligence for detecting periapical pathosis on cone‐beam computed tomography scans[J]. International Endodontic Journal,2020,53(5).[42]Avila A M,Mezi? I. Data-driven analysis and forecasting of highway traffic dynamics.[J]. Nature communications,2020,11(1).[43]Neri Emanuele,Miele Vittorio,Coppola Francesca,Grassi Roberto. Use of CT andartificial intelligence in suspected or COVID-19 positive patients: statement of the Italian Society of Medical and Interventional Radiology.[J]. La Radiologia medica,2020.[44]Tau Noam,Stundzia Audrius,Yasufuku Kazuhiro,Hussey Douglas,Metser Ur. Convolutional Neural Networks in Predicting Nodal and Distant Metastatic Potential of Newly Diagnosed Non-Small Cell Lung Cancer on FDG PET Images.[J]. AJR. American journal of roentgenology,2020.[45]Coppola Francesca,Faggioni Lorenzo,Regge Daniele,Giovagnoni Andrea,Golfieri Rita,Bibbolino Corrado,Miele Vittorio,Neri Emanuele,Grassi Roberto. Artificial intelligence: radiologists' expectations and opinions gleaned from a nationwide online survey.[J]. La Radiologia medica,2020.[46]?. ? ? ? ? [J]. ,2020,25(4).[47]Savage Rock H,van Assen Marly,Martin Simon S,Sahbaee Pooyan,Griffith Lewis P,Giovagnoli Dante,Sperl Jonathan I,Hopfgartner Christian,K?rgel Rainer,Schoepf U Joseph. Utilizing Artificial Intelligence to Determine Bone Mineral Density Via Chest Computed Tomography.[J]. Journal of thoracic imaging,2020,35 Suppl 1.[48]Brzezicki Maksymilian A,Bridger Nicholas E,Kobeti? Matthew D,Ostrowski Maciej,Grabowski Waldemar,Gill Simran S,Neumann Sandra. Artificial intelligence outperforms human students in conducting neurosurgical audits.[J]. Clinical neurology and neurosurgery,2020,192.[49]Lockhart Mark E,Smith Andrew D. Fatty Liver Disease: Artificial Intelligence Takes on the Challenge.[J]. Radiology,2020,295(2).[50]Wood Edward H,Korot Edward,Storey Philip P,Muscat Stephanie,Williams George A,Drenser Kimberly A. The retina revolution: signaling pathway therapies, genetic therapies, mitochondrial therapies, artificial intelligence.[J]. Current opinion in ophthalmology,2020,31(3).[51]Ho Dean,Quake Stephen R,McCabe Edward R B,Chng Wee Joo,Chow Edward K,Ding Xianting,Gelb Bruce D,Ginsburg Geoffrey S,Hassenstab Jason,Ho Chih-Ming,Mobley William C,Nolan Garry P,Rosen Steven T,Tan Patrick,Yen Yun,Zarrinpar Ali. Enabling Technologies for Personalized and Precision Medicine.[J]. Trends in biotechnology,2020,38(5).[52]Fischer Andreas M,Varga-Szemes Akos,van Assen Marly,Griffith L Parkwood,Sahbaee Pooyan,Sperl Jonathan I,Nance John W,Schoepf U Joseph. Comparison of Artificial Intelligence-Based Fully Automatic Chest CT Emphysema Quantification to Pulmonary Function Testing.[J]. AJR. American journal ofroentgenology,2020,214(5).[53]Moore William,Ko Jane,Gozansky Elliott. Artificial Intelligence Pertaining to Cardiothoracic Imaging and Patient Care: Beyond Image Interpretation.[J]. Journal of thoracic imaging,2020,35(3).[54]Hwang Eui Jin,Park Chang Min. Clinical Implementation of Deep Learning in Thoracic Radiology: Potential Applications and Challenges.[J]. Korean journal of radiology,2020,21(5).[55]Mateen Bilal A,David Anna L,Denaxas Spiros. Electronic Health Records to Predict Gestational Diabetes Risk.[J]. Trends in pharmacological sciences,2020,41(5).[56]Yao Xiang,Mao Ling,Lv Shunli,Ren Zhenghong,Li Wentao,Ren Ke. CT radiomics features as a diagnostic tool for classifying basal ganglia infarction onset time.[J]. Journal of the neurological sciences,2020,412.[57]van Assen Marly,Banerjee Imon,De Cecco Carlo N. Beyond the Artificial Intelligence Hype: What Lies Behind the Algorithms and What We Can Achieve.[J]. Journal of thoracic imaging,2020,35 Suppl 1.[58]Guzik Tomasz J,Fuster Valentin. Leaders in Cardiovascular Research: Valentin Fuster.[J]. Cardiovascular research,2020,116(6).[59]Fischer Andreas M,Eid Marwen,De Cecco Carlo N,Gulsun Mehmet A,van Assen Marly,Nance John W,Sahbaee Pooyan,De Santis Domenico,Bauer Maximilian J,Jacobs Brian E,Varga-Szemes Akos,Kabakus Ismail M,Sharma Puneet,Jackson Logan J,Schoepf U Joseph. Accuracy of an Artificial Intelligence Deep Learning Algorithm Implementing a Recurrent Neural Network With Long Short-term Memory for the Automated Detection of Calcified Plaques From Coronary Computed Tomography Angiography.[J]. Journal of thoracic imaging,2020,35 Suppl 1.[60]Ghosh Adarsh,Kandasamy Devasenathipathy. Interpretable Artificial Intelligence: Why and When.[J]. AJR. American journal of roentgenology,2020,214(5).[61]M.Rosario González-Rodríguez,M.Carmen Díaz-Fernández,Carmen Pacheco Gómez. Facial-expression recognition: An emergent approach to the measurement of tourist satisfaction through emotions[J]. Telematics and Informatics,2020,51.[62]Ru-Xi Ding,Iván Palomares,Xueqing Wang,Guo-Rui Yang,Bingsheng Liu,Yucheng Dong,Enrique Herrera-Viedma,Francisco Herrera. Large-Scale decision-making: Characterization, taxonomy, challenges and future directions from an Artificial Intelligence and applications perspective[J]. Information Fusion,2020,59.[63]Abdulrhman H. Al-Jebrni,Brendan Chwyl,Xiao Yu Wang,Alexander Wong,Bechara J. Saab. AI-enabled remote and objective quantification of stress at scale[J]. Biomedical Signal Processing and Control,2020,59.[64]Gillian Thomas,Elizabeth Eisenhauer,Robert G. Bristow,Cai Grau,Coen Hurkmans,Piet Ost,Matthias Guckenberger,Eric Deutsch,Denis Lacombe,Damien C. Weber. The European Organisation for Research and Treatment of Cancer, State of Science in radiation oncology and priorities for clinical trials meeting report[J]. European Journal of Cancer,2020,131.[65]Muhammad Asif. Are QM models aligned with Industry 4.0? A perspective on current practices[J]. Journal of Cleaner Production,2020,258.[66]Siva Teja Kakileti,Himanshu J. Madhu,Geetha Manjunath,Leonard Wee,Andre Dekker,Sudhakar Sampangi. Personalized risk prediction for breast cancer pre-screening using artificial intelligence and thermal radiomics[J]. Artificial Intelligence In Medicine,2020,105.[67]. Evaluation of Payer Budget Impact Associated with the Use of Artificial Intelligence in Vitro Diagnostic, Kidneyintelx, to Modify DKD Progression:[J]. American Journal of Kidney Diseases,2020,75(5).[68]Rohit Nishant,Mike Kennedy,Jacqueline Corbett. Artificial intelligence for sustainability: Challenges, opportunities, and a research agenda[J]. International Journal of Information Management,2020,53.[69]Hoang Nguyen,Xuan-Nam Bui. Soft computing models for predicting blast-induced air over-pressure: A novel artificial intelligence approach[J]. Applied Soft Computing Journal,2020,92.[70]Benjamin S. Hopkins,Aditya Mazmudar,Conor Driscoll,Mark Svet,Jack Goergen,Max Kelsten,Nathan A. Shlobin,Kartik Kesavabhotla,Zachary A Smith,Nader S Dahdaleh. Using artificial intelligence (AI) to predict postoperative surgical site infection: A retrospective cohort of 4046 posterior spinal fusions[J]. Clinical Neurology and Neurosurgery,2020,192.[71]Mei Yang,Runze Zhou,Xiangjun Qiu,Xiangfei Feng,Jian Sun,Qunshan Wang,Qiufen Lu,Pengpai Zhang,Bo Liu,Wei Li,Mu Chen,Yan Zhao,Binfeng Mo,Xin Zhou,Xi Zhang,Yingxue Hua,Jin Guo,Fangfang Bi,Yajun Cao,Feng Ling,Shengming Shi,Yi-Gang Li. Artificial intelligence-assisted analysis on the association between exposure to ambient fine particulate matter and incidence of arrhythmias in outpatients of Shanghai community hospitals[J]. Environment International,2020,139.[72]Fatemehalsadat Madaeni,Rachid Lhissou,Karem Chokmani,Sebastien Raymond,Yves Gauthier. Ice jam formation, breakup and prediction methods based on hydroclimatic data using artificial intelligence: A review[J]. Cold Regions Science and Technology,2020,174.[73]Steve Chukwuebuka Arum,David Grace,Paul Daniel Mitchell. A review of wireless communication using high-altitude platforms for extended coverage and capacity[J]. Computer Communications,2020,157.[74]Yong-Hong Kuo,Nicholas B. Chan,Janny M.Y. Leung,Helen Meng,Anthony Man-Cho So,Kelvin K.F. Tsoi,Colin A. Graham. An Integrated Approach of Machine Learning and Systems Thinking for Waiting Time Prediction in an Emergency Department[J]. International Journal of Medical Informatics,2020,139.[75]Matteo Terzi,Gian Antonio Susto,Pratik Chaudhari. Directional adversarial training for cost sensitive deep learning classification applications[J]. Engineering Applications of Artificial Intelligence,2020,91.[76]Arman Kilic. Artificial Intelligence and Machine Learning in Cardiovascular Health Care[J]. The Annals of Thoracic Surgery,2020,109(5).[77]Hossein Azarmdel,Ahmad Jahanbakhshi,Seyed Saeid Mohtasebi,Alfredo Rosado Mu?oz. Evaluation of image processing technique as an expert system in mulberry fruit grading based on ripeness level using artificial neural networks (ANNs) and support vector machine (SVM)[J]. Postharvest Biology and Technology,2020,166.[78]Wafaa Wardah,Abdollah Dehzangi,Ghazaleh Taherzadeh,Mahmood A. Rashid,M.G.M. Khan,Tatsuhiko Tsunoda,Alok Sharma. Predicting protein-peptide binding sites with a deep convolutional neural network[J]. Journal of Theoretical Biology,2020,496.[79]Francisco F.X. Vasconcelos,Róger M. Sarmento,Pedro P. Rebou?as Filho,Victor Hugo C. de Albuquerque. Artificial intelligence techniques empowered edge-cloud architecture for brain CT image analysis[J]. Engineering Applications of Artificial Intelligence,2020,91.[80]Masaaki Konishi. Bioethanol production estimated from volatile compositions in hydrolysates of lignocellulosic biomass by deep learning[J]. Journal of Bioscience and Bioengineering,2020,129(6).人工智能英文参考文献三：[81]J. Kwon,K. Kim. Artificial Intelligence for Early Prediction of Pulmonary Hypertension Using Electrocardiography[J]. Journal of Heart and Lung Transplantation,2020,39(4).[82]C. Maathuis,W. Pieters,J. van den Berg. Decision support model for effects estimation and proportionality assessment for targeting in cyber operations[J]. Defence Technology,2020.[83]Samer Ellahham. Artificial Intelligence in Diabetes Care[J]. The American Journal of Medicine,2020.[84]Yi-Ting Hsieh,Lee-Ming Chuang,Yi-Der Jiang,Tien-Jyun Chang,Chung-May Yang,Chang-Hao Yang,Li-Wei Chan,Tzu-Yun Kao,Ta-Ching Chen,Hsuan-Chieh Lin,Chin-Han Tsai,Mingke Chen. Application of deep learning image assessment software VeriSee? for diabetic retinopathy screening[J]. Journal of the Formosan Medical Association,2020.[85]Emre ARTUN,Burak KULGA. Selection of candidate wells for re-fracturing in tight gas sand reservoirs using fuzzy inference[J]. Petroleum Exploration and Development Online,2020,47(2).[86]Alberto Arenal,Cristina Armu?a,Claudio Feijoo,Sergio Ramos,Zimu Xu,Ana Moreno. Innovation ecosystems theory revisited: The case of artificial intelligence in China[J]. Telecommunications Policy,2020.[87]T. Som,M. Dwivedi,C. Dubey,A. Sharma. Parametric Studies on Artificial Intelligence Techniques for Battery SOC Management and Optimization of Renewable Power[J]. Procedia Computer Science,2020,167.[88]Bushra Kidwai,Nadesh RK. Design and Development of Diagnostic Chabot for supporting Primary Health Care Systems[J]. Procedia Computer Science,2020,167.[89]Asl? Bozda?,Ye?im Dokuz,?znur Begüm G?k?ek. Spatial prediction of PM 10 concentration using machine learning algorithms in Ankara, Turkey[J]. Environmental Pollution,2020.[90]K.P. Smith,J.E. Kirby. Image analysis and artificial intelligence in infectious disease diagnostics[J]. Clinical Microbiology and Infection,2020.[91]Alklih Mohamad YOUSEF,Ghahfarokhi Payam KAVOUSI,Marwan ALNUAIMI,Yara ALATRACH. Predictive data analytics application for enhanced oil recovery in a mature field in the Middle East[J]. Petroleum Exploration and Development Online,2020,47(2).[92]Omer F. Ahmad,Danail Stoyanov,Laurence B. Lovat. Barriers and pitfalls for artificial intelligence in gastroenterology: Ethical and regulatory issues[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[93]Sanne A. Hoogenboom,Ulas Bagci,Michael B. Wallace. Artificial intelligence in gastroenterology. The current state of play and the potential. How will it affect our practice and when?[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[94]Douglas K. Rex. Can we do resect and discard with artificial intelligence-assisted colon polyp “optical biopsy?”[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[95]Neal Shahidi,Michael J. Bourke. Can artificial intelligence accurately diagnose endoscopically curable gastrointestinal cancers?[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[96]Michael Byrne. Artificial intelligence in gastroenterology[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[97]Piet C. de Groen. Using artificial intelligence to improve adequacy of inspection in gastrointestinal endoscopy[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[98]Robin Zachariah,Andrew Ninh,William Karnes. Artificial intelligence for colon polyp detection: Why should we embrace this?[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[99]Alexandra T. Greenhill,Bethany R. Edmunds. A primer of artificial intelligence in medicine[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[100]Tomohiro Tada,Toshiaki Hirasawa,Toshiyuki Yoshio. The role for artificial intelligence in evaluation of upper GI cancer[J]. Techniques and Innovations in Gastrointestinal Endoscopy,2020,22(2).[101]Yahui Jiang,Meng Yang,Shuhao Wang,Xiangchun Li,Yan Sun. Emerging role of deep learning‐based artificial intelligence in tumor pathology[J]. Cancer Communications,2020,40(4).[102]Kristopher D. Knott,Andreas Seraphim,Joao B. Augusto,Hui Xue,Liza Chacko,Nay Aung,Steffen E. Petersen,Jackie A. Cooper,Charlotte Manisty,Anish N. Bhuva,Tushar Kotecha,Christos V. Bourantas,Rhodri H. Davies,Louise A.E. Brown,Sven Plein,Marianna Fontana,Peter Kellman,James C. Moon. The Prognostic Significance of Quantitative Myocardial Perfusion: An Artificial Intelligence–Based Approach Using Perfusion Mapping[J]. Circulation,2020,141(16).[103]Muhammad Asad,Ahmed Moustafa,Takayuki Ito. FedOpt: Towards Communication Efficiency and Privacy Preservation in Federated Learning[J]. Applied Sciences,2020,10(8).[104]Wu Wenzhi,Zhang Yan,Wang Pu,Zhang Li,Wang Guixiang,Lei Guanghui,Xiao Qiang,Cao Xiaochen,Bian Yueran,Xie Simiao,Huang Fei,Luo Na,Zhang Jingyuan,Luo Mingyan. Psychological stress of medical staffs during outbreak of COVID-19 and adjustment strategy.[J]. Journal of medical virology,2020.[105]. Eyenuk Fulfills Contract for Artificial Intelligence Grading of Retinal Images[J]. Telecomworldwire,2020.[106]Kim Tae Woo,Duhachek Adam. Artificial Intelligence and Persuasion: A Construal-Level Account.[J]. Psychological science,2020,31(4).[107]McCall Becky. COVID-19 and artificial intelligence: protecting health-care workers and curbing the spread.[J]. The Lancet. Digital health,2020,2(4).[108]Alca?iz Mariano,Chicchi Giglioli Irene A,Sirera Marian,Minissi Eleonora,Abad Luis. [Autism spectrum disorder biomarkers based on biosignals, virtual reality and artificial intelligence].[J]. Medicina,2020,80 Suppl 2.[109]Cong Lei,Feng Wanbing,Yao Zhigang,Zhou Xiaoming,Xiao Wei. Deep Learning Model as a New Trend in Computer-aided Diagnosis of Tumor Pathology for Lung Cancer.[J]. Journal of Cancer,2020,11(12).[110]Wang Fengdan,Gu Xiao,Chen Shi,Liu Yongliang,Shen Qing,Pan Hui,Shi Lei,Jin Zhengyu. Artificial intelligence system can achieve comparable results to experts for bone age assessment of Chinese children with abnormal growth and development.[J]. PeerJ,2020,8.[111]Hu Wenmo,Yang Huayu,Xu Haifeng,Mao Yilei. Radiomics based on artificial intelligence in liver diseases: where we are?[J]. Gastroenterology report,2020,8(2).[112]Batayneh Wafa,Abdulhay Enas,Alothman Mohammad. Prediction of the performance of artificial neural networks in mapping sEMG to finger joint angles via signal pre-investigation techniques.[J]. Heliyon,2020,6(4).[113]Aydin Emrah,Türkmen ?nan Utku,Namli G?zde,?ztürk ?i?dem,Esen Ay?e B,Eray Y Nur,Ero?lu Egemen,Akova Fatih. A novel and simple machine learning algorithm for preoperative diagnosis of acute appendicitis in children.[J]. Pediatric surgery international,2020.[114]Ellahham Samer. Artificial Intelligence in Diabetes Care.[J]. The Americanjournal of medicine,2020.[115]David J. Winkel,Thomas J. Weikert,Hanns-Christian Breit,Guillaume Chabin,Eli Gibson,Tobias J. Heye,Dorin Comaniciu,Daniel T. Boll. Validation of a fully automated liver segmentation algorithm using multi-scale deep reinforcement learning and comparison versus manual segmentation[J]. European Journal of Radiology,2020,126.[116]Binjie Fu,Guoshu Wang,Mingyue Wu,Wangjia Li,Yineng Zheng,Zhigang Chu,Fajin Lv. Influence of CT effective dose and convolution kernel on the detection of pulmonary nodules in different artificial intelligence software systems: A phantom study[J]. European Journal of Radiology,2020,126.[117]Georgios N. Kouziokas. A new W-SVM kernel combining PSO-neural network transformed vector and Bayesian optimized SVM in GDP forecasting[J]. Engineering Applications of Artificial Intelligence,2020,92.[118]Qingsong Ruan,Zilin Wang,Yaping Zhou,Dayong Lv. A new investor sentiment indicator ( ISI ) based on artificial intelligence: A powerful return predictor in China[J]. Economic Modelling,2020,88.[119]Mohamed Abdel-Basset,Weiping Ding,Laila Abdel-Fatah. The fusion of Internet of Intelligent Things (IoIT) in remote diagnosis of obstructive Sleep Apnea: A survey and a new model[J]. Information Fusion,2020,61.[120]Federico Caobelli. Artificial intelligence in medical imaging: Game over for radiologists?[J]. European Journal of Radiology,2020,126.以上就是关于人工智能参考文献的分享，希望对你有所帮助。

Learning New Compositions from Given OnesJi DonghongDepartment of Computer ScienceTsinghua UniversityBeijing, 100084, P. R. ChinaEmail: jdh@He JunDepartment of Computer ScienceHerbin Institute of TechnologyEmail: hj@Huang ChangningDepartment of Computer ScienceTsinghua UniversityBeijing, 100084, P. R. ChinaEmail: hcn@Abstract:In this paper, we study the problem of learning new compositions of words from given ones with a specific syntactic structure, e.g., A-N or V-N structures. We first cluster words according to the given compositions, then construct a cluster-based compositional frame for each word cluster, which contains both new and given compositions relevant with the words in the cluster. In contrast to other methods, we don’t pre-define the number of clusters, and formalize the problem of clustering words as a non-linear optimization one, in which we specify the environments of words based on word clusters to be determined, rather than their neighboring words. To solve the problem, we make use of a kind of cooperative evolution strategy to design an evolutionary algorithm.Key words: word compositions, evolutionary learning, clustering, natural language processing.1. IntroductionWord compositions have long been a concern in lexicography(Benson et al. 1986; Miller et al. 1995), and now as a specific kind of lexical knowledge, it has been shown that they have an important role in many areas in natural language processing, e.g., parsing, generation, lexicon building, word sense disambiguation, and information retrieving, etc.(e.g., Abney 1989, 1990; Benson et al. 1986; Yarowsky 1995; Church and Hanks 1989; Church, Gale, Hans, and Hindle 1989). But due to the huge number of words, it is impossible to list all compositions between words by hand in dictionaries. So an urgent problem occurs: how to automatically acquire word compositions?In general, word compositions fall into two categories: free compositions and bound compositions, i.e., collocations. Free compositions refer to those in which words can be replaced by other similar ones, while in bound compositions, words cannot be replaced freely(Benson 1990). Free compositions are predictable, i.e., their reasonableness can be determined according to the syntactic and semantic properties of the words in them. While bound compositions are not predictable, i.e., their reasonableness cannot be derived from the syntactic and semantic propertiesof the words in them(Smadja 1993).N ow with the availability of large-scale corpus, automatic acquisition of word compositions, especially word collocations from them have been extensively studied(e.g., Choueka et al. 1988; Church and Hanks 1989; Smadja 1993). The key of their methods is to make use of some statistical means, e.g., frequencies or mutual information, to quantify the compositional strength between words. These methods are more appropriate for retrieving bound compositions, while less appropriate for retrieving free ones. This is because in free compositions, words are related with each other in a more loose way, which may result in the invalidity of mutual information and other statistical means in distinguishing reasonable compositions from unreasonable ones.In this paper, we start from a different point to explore the problem of automatic acquisition of free compositions. Although we cannot list all free compositions, we can select some typical ones as those specified in some dictionaries(e.g., Benson 1986; Zhang et al. 1994). According to the properties held by free compositions, we can reasonably suppose that selected compositions can provide strong clues for others. Furthermore we suppose that words can be classified into clusters, with the members in each cluster similar in their compositional ability, which can be characterized as the set of the words able to combined with them to form meaningful phrases. Thus any given composition, although specifying the relation between two words literally, suggests the relation between two clusters. So for each word(or cluster), there exist some word clusters, the word (or the words in the cluster) can and only can combine with the words in the clusters to form meaningful phrases. We call the set of these clusters compositional frame of the word (or the cluster).A seemingly plausible method to determine compositional frames is to make use of pre-defined semantic classes in some thesauri(e.g., Miller et al. 1993; Mei et al. 1996). The rationale behind the method is to take such an assumption that if one word can be combined with another one to form a meaningful phrase, the words similar to them in meaning can also be combined with each other. But it has been shown that the similarity between words in meaning doesn’t correspond to the similarity in compositional ability(Zhu 1982). So adopting semantic classes to construct compositional frames will result in considerable redundancy.An alternative to semantic class is word cluster based on distributional environment (Brown et al., 1992), which in general refers to the surrounding words distributed around certain word (e.g., Hatzivassiloglou et al., 1993; Pereira et al., 1993), or the classes of them(Bensch et al., 1995), or more complex statistical means (Dagan et al., 1993). According to the properties of the clusters in compositional frames, the clusters should be based on the environment, which, however, is narrowed in the given compositions. Because the given compositions are listed by hand, it is impossible to make use of statistical means to form the environment, the remaining choices are surrounding words or classes of them.Pereira et al.(1993) put forward a method to cluster nouns in V-N compositions, taking the verbs which can combine with a noun as its environment. Although its goal is to deal with the problem of data sparseness, it suffers from the problem itself. A strategy to alleviate the effects of the problem is to cluster nouns and verbs simultaneously. But as a result, the problem of word clustering becomes a bootstrapping one, or a non-linear one: the environment is also to be determined. Bensch et al. (1995) proposed a definite method to deal with the generalized version of the non-linear problem, but it suffers from the problem of local optimization.In this paper, we focus on A-N compositions in Chinese, and explore the problem of learningnew compositions from given ones. In order to copy with the problem of sparseness, we take adjective clusters as nouns’ environment, and take noun clusters as adjectives’ environment. In order to avoid local optimal solutions, we propose a cooperative evolutionary strategy. The method uses no specific knowledge of A-N structure, and can be applied to other structures.The remainder of the paper is organized as follows: in section 2, we give a formal description of the problem. In section 3, we discuss a kind of cooperative evolution strategy to deal with the problem. In section 4, we explore the problem of parameter estimation. In section 5, we present our experiments and the results as well as their evaluation. In section 6, we give some conclusions and discuss future work.2. Problem SettingGiven an adjective set and a noun set, suppose for each noun, some adjectives are listed as its compositional instances1. Our goal is to learn new reasonable compositions from the instances. To do so, we cluster nouns and adjectives simultaneously and build a compositional frame for each noun.Suppose A is the set of adjectives, N is the set of nouns, for any a symbol 206 \f "Symbol" \s 10.5∈}A, let f(a)symbol 205 \f "Symbol" \s 10.5⊆}N be the instance set of a, i.e., the set of nouns in N which can be combined with a, and for any n symbol 206 \f "Symbol" \s 10.5∈}N, let g(n)symbol 205 \f "Symbol" \s 10.5⊆}A be the instance set of n, i.e., the set of adjectives in A which can be combined with n. We first give some formal definitions in the following:Definition 1 partitionSuppose U is a non-empty finite set, we call <U1, U2, ..., U k> a partition of U, if:i) for any U i, and U j, i symbol 185 \f "Symbol" \s 10.5≠}j, U i symbol 199 \f "Symbol" \s10.5∩}U j•symbol 102 \f "Symbol" \s 10.5φ,ii) U=U ii k1≤≤UWe call U i a cluster of U.Suppose U•< A1, A2, ..., A p > is a partition of A, V•<N1, N2, ..., N q> is a partition of N, f and g are defined as above, for any N i, let g(N i)•{A j:_symbol 36 \f "Symbol" \s 10.5∃n symbol 206 \f "Symbol" \s 10.5∈}N i, A j symbol 199 \f "Symbol" \s 10.5∩}g(n)symbol 185 \f "Symbol" \s 10.5≠}symbol 102 \f "Symbol" \s 10.5φ}, and for anyn, let δ<>U V n,()•symbol 124 \f "Symbol" \s 10.5|{a:_symbol 36 \f "Symbol" \s 10.5∃A j, A j symbol 206 \f "Symbol" \s 10.5∈}g(N k), a symbol 206 \f "Symbol" \s 10.5∈}A j}symbol 45 \f "Symbol" \s 10.5−g(n)symbol 124 \f "Symbol" \s 10.5|, where n symbol 206 \f "Symbol" \s 10.5∈}N k. Intuitively,δ<>U V n,()is the number of the new instances relevant with n. We define the general learning amount as the following:Definition 2 learning amountδ<>U V,1 The compositional instances of the adjectives can be inferred from those of the nouns.δ<>U V ,•δ<>∈∑n N n ,()Based on the partitions of both nouns and adjectives, we can define the distance between nouns and that between adjectives.Definition 3 distance between wordsfor any a symbol 206 \f "Symbol" \s 10.5∈}A , let f a ()={N i :_1symbol 163 \f "Symbol" \s 10.5≤}i symbol 163 \f "Symbol" \s 10.5≤}q , N i symbol 199 \f "Symbol" \s 10.5∩}f(a)symbol 185 \f "Symbol" \s 10.5≠}symbol 102 \f "Symbol" \s 10.5φ}, for any n symbol 206 \f "Symbol" \s 10.5∈}N , let g n U ()={A i :_1symbol 163 \f "Symbol" \s 10.5≤}i symbol 163 \f "Symbol" \s 10.5≤}p , A i symbol 199 \f "Symbol" \s 10.5∩}g(n)symbol 185 \f "Symbol" \s 10.5≠}symbol 102 \f "Symbol" \s 10.5φ}, for any two nouns n 1 and n 2, any two adjectives a 1 and a 2, we define the distances between them respectively as the following: i) dis n n U (,)12=1-g n g n g n g n U U U U ()()()()1212∩∪ii) dis a a V (,)12=1-f a f a f a f a V V V V ()()()()1212∩∪According to the distances between words, we can define the distances between word sets .Definition 4 distance between word setsGiven any two adjective sets X 1, X 2symbol 205 \f "Symbol" \s 10.5⊆}A , any two noun sets Y 1, Y 2symbol 205 \f "Symbol" \s 10.5⊆}N , their distances are:i) dis X X V (,)12=a X a X V dis a a 112212∈∈,max{(,)}ii) dis Y Y U (,)12 =n Y n Y dis n n 112212∈∈,max {(,)}Intuitively, the distance between word sets refer to the biggest distance between words respectively in the two sets.We formalize the problem of clustering nouns and adjectives simultaneously as an optimization problem with some constraints. (1)To determine a partition U •<A 1, A 2, ..., A p > of A , and a partition V •<N 1, N 2, ..., N q > of N , where p , q symbol 62 \f "Symbol" \s 10.5>0, which satisfies i) and ii), and minimize δ<>U V ,.i) for any a 1, a 2symbol 206 \f "Symbol" \s 10.5∈}A i , 1symbol 163 \f "Symbol" \s 10.5≤}i symbol 163 \f "Symbol" \s 10.5≤}p , dis a a (,)12<t 1; for A i and A j , 1symbol 163 \f "Symbol" \s 10.5≤}i symbol 185 \f "Symbol" \s 10.5≠}j symbol 163 \f "Symbol" \s 10.5≤}p , dis A A V i j (,)symbol 179 \f "Symbol" \s 10.5≥} t 1;ii) for any n 1, n 2symbol 206 \f "Symbol" \s 10.5∈}N i , 0symbol 163 \f"Symbol" \s 10.5≤}i symbol 163 \f "Symbol" \s 10.5≤}q , dis n n U (,)12<t 1; for N i and N j , 1symbol 163 \f "Symbol" \s 10.5≤}i symbol 185 \f "Symbol" \s 10.5≠}j symbol 163 \f "Symbol" \s 10.5≤}q , dis N N i j (,)symbol 179 \f "Symbol" \s 10.5≥} t 2;where 0symbol 163 \f "Symbol" \s 10.5≤}t 1, t 2symbol 163 \f "Symbol" \s 10.5≤}1.Intuitively, the conditions i) and ii) make the distances between words within clusters smaller, and those between different clusters bigger, and to minimize δ<>U V ,means to minimize the distances between the words within clusters. In fact, (U , V ) can be seen as an abstraction model over given compositions, and t 1, t 2 can be seen as its abstraction degree . Consider the two special case: one is t 1=t 2 =0, i.e., the abstract degree is the lowest, when the result is that one noun forms a cluster and on adjective forms a cluster, which means that no new compositions are learned. The other is t 1=t 2 =1, the abstract degree is the highest, when a possible result is that all nouns form a cluster and all adjectives form a cluster, which means that all possible compositions, reasonable or unreasonable, are learned. So we need estimate appropriate values for the two parameters, in order to make an appropriate abstraction over given compositions, i.e., make the compositional frames contain as many reasonable compositions as possible, and as few unreasonable ones as possible.3. Cooperative EvolutionSince the beginning of evolutionary algorithms, they have been applied in many areas in AI(Davis et al., 1991; Holland 1994). Recently, as a new and powerful learning strategy,cooperative evolution has gained much attention in solving complex non-linear problem. In this section, we discuss how to deal with the problem (1) based on the strategy.According to the interaction between adjective clusters and noun clusters, we adopt such a cooperative strategy: after establishing the preliminary solutions, for any preliminary solution, we optimize N ’s partition based on A ’s partition, then we optimize A ’s partition based on N ’s partition, and so on, until the given conditions are satisfied.3.1 Preliminary SolutionsWhen determining the preliminary population, we also cluster nouns and adjectives respectively.However, we see the environment of a noun as the set of all adjectives which occur with it in given compositions, and that of an adjective as the set of all the nouns which occur with it in given compositions. Compared with (1), the problem is a linear clustering one.Suppose a 1, a 2symbol 206 \f "Symbol" \s 10.5∈}A , f is defined as above, we define the linear distance between them as (2):(2) dis (a 1, a 2)•1-f a f a f a f a ()()()()1212∩∪Similarly, we can define the linear distance between nouns dis (n 1, n 2) based on g . In contrast, we call the distances in definition 3 non-linear distances .According to the linear distances between adjectives, we can determine a preliminary partition of N : randomly select an adjective and put it into an empty set X , then scan the otheradjectives in A , for any adjective in A -X, if its distances from the adjectives in X are all smaller than t 1, then put it into X , finally X forms a preliminary cluster. Similarly, we can build another preliminary cluster in (A -X ). So on, we can get a set of preliminary clusters, which is just a partition of A . According to the different order in which we scan the adjectives, we can get different preliminary partitions of A .Similarly, we can determine the preliminary partitions of N based on the linear distances between nouns. A partition of A and a partition of N forms a preliminary solution of (1), and all possible preliminary solutions forms the population of preliminary solutions, which we also call the population of 0th generation solutions.3.2 Evolution OperationIn general, evolution operation consists of recombination, mutation and selection. Recombination makes two solutions in a generation combine with each other to form a solution belonging to nextgeneration. Suppose <U i 1(); V i 1()> and <U i 2(); V i 2()> are two ith generation solutions, whereU i 1() and U i 2() are two partitions of A , V i 1() and V i 2() are two partitions of N , then <U i 1()•V i 2()>and <U i 2(); V i 1()> forms two possible (i +1)th generation solutions.Mutation makes a solution in a generation improve its fitness , and evolve into a new onebelonging to next generation. Suppose <U i (); V i ()> is a ith generation solution, whereU i ()•<A 1, A 2, ..., A p >•V i ()•<N 1, N 2, ..., N q > are partitions of A and N respectively, the mutation is aimed at optimizing Vi () into V i ()£«1 based on U i (), and makes V i ()£«1 satisfy the condition ii)in (1), or optimizing U i () into Ui ()£«1 based on V i (), and makes U i ()£«1 satisfy the condition i)in (1), then moving words across clusters to minimize δ<>U V ,.We design three steps for mutation operation: splitting , merging and moving , the former two are intended for the partitions to satisfy the conditions in (1), and the third intended to minimizeδ<>U V ,. In the following, we take the evolution of V i ()£«1 as an example to demonstrate the three steps.symbol 183 \f "Symbol" \s 10.5•} Splitting Procedure . For any N k , 1symbol 163 \f "Symbol"\s 10.5≤}k symbol 163 \f "Symbol" \s 10.5≤}q , if there exist n 1, n 2symbol 206 \f "Symbol" \s 10.5∈}N k , such that dis n n U i ()(,)12symbol 179 \f "Symbol" \s 10.5≥} t 2, then splitting N k into two subsets X and Y . The procedure is given as the following:i) Put n 1 into X , n 2 into Y ,ii) Select the noun in (N k -(X symbol 200 \f "Symbol" \s 10.5∪}Y )) whose distance from n 1is the smallest, and put it into X ,iii) Select the noun in (N k -(X symbol 200 \f "Symbol" \s 10.5∪}Y )) whose distance from n 2is the smallest, and put it into Y,iv) Repeat ii) and iii), until X symbol 200 \f "Symbol" \s 10.5∪}Y = N k .For X (or Y ), if there exist n 1, n 2symbol 206 \f "Symbol" \s 10.5∈}X (or Y ), dis n n U i ()(,)12symbol 179 \f "Symbol" \s 10.5≥} t 2, then we can make use of the above procedure to split it into moresmaller sets. Obviously, we can split any N k in V i () into several subsets which satisfy thecondition ii) in (1) by repeating the procedure.symbol 183 \f "Symbol" \s 10.5•} Merging procedure . If there exist N j and N k , where1symbol 163 \f "Symbol" \s 10.5≤}j,k symbol 163 \f "Symbol" \s 10.5≤}q , such that dis N N U j k i ()(,)symbol 60 \f "Symbol" \s 10.5<t 2, then merging them into a new cluster.It is easy to prove that U i () and V i () will meet the condition i) and ii) in (1) respectively, after splitting and merging procedure.symbol 183 \f "Symbol" \s 10.5•} Moving procedure . We call moving n from N j to N k a word move , where 1symbol 163 \f "Symbol" \s 10.5≤}j symbol 185 \f "Symbol" \s 10.5≠}k symbol 163 \f "Symbol" \s 10.5≤}q , denoted as (n , N j , N k ), if the condition (ii) remains satisfied. The procedure is as the following:i) Select a word move (n , N j , N k ) which minimizes δ<>U V ,,ii) Move n from N j to N k ,iii) Repeat i) and ii)•until there are no word moves which reduce δ<>U V ,.After the three steps, U i ()and V i () evolve into U i ()£«1and V i ()£«1 respectively.Selection operation selects the solutions among those in the population of certain generation according to their fitness. We define the fitness of a solution as its learning amount.We use J i to denote the set of ith generation solutions, H (i , i+1), as in (3), specifies the similarity between ith generation solutions and (i +1)th generation solutions.(3) H (i , i+1)=Min U V J Min U V J U V i i i U V i i i i i i i {:_(,)}{:_(,)}()()()(),()(),()()δδ<>+++<>++∈∈11111Let t 3 be a threshold for H (i , i+1), the following is the general evolutionary algorithm:Procedure Clustering (A , N , f , g );begini) Build preliminary solution population I 0,ii) Determine 0th generation solution set J 0 according to their fitness,iii) Determine I i+1 based on J i :a) Recombination: if (U i 1(),V i 1()), (U i 2(),V i 2())symbol 206 \f "Symbol" \s10.5∈}J i , then (U i 1(),V i 2()), (U i 2(),V i 1())symbol 206 \f "Symbol" \s10.5∈}I i+1,b) Mutation: if (U i (),V i ())symbol 206 \f "Symbol" \s 10.5∈}J i , then(U i (),V i ()£«1), (U i ()£«1,V i ())symbol 206 \f "Symbol" \s 10.5∈}I i+1,iv) Determine J i+1 from I i+1 according to their fitness,v) If H (i ,i+1)>t 3, then exit, otherwise goto iii),end •After determining the clusters of adjectives and nouns, we can construct the compositional frame for each noun cluster or each noun. In fact, for each noun cluster N i , g (N i )={A j :_symbol 36 \f "Symbol" \s 10.5∃n symbol 206 \f "Symbol" \s 10.5∈}N i , A j symbol 199 \f "Symbol" \s 10.5∩}g (n )symbol 185 \f "Symbol" \s 10.5≠}symbol 102 \f"Symbol" \s 10.5φ} is just its compositional frame, and for any noun in N i , g (N i ) is also its compositional frame. Similarly, for each adjective (or adjective cluster), we can also determine its compositional frame.4. Parameter EstimationThe parameters t 1 and t 2 in (1) are the thresholds for the distances between the clusters of A and N respectively. If they are too big, the established frame will contain more unreasonable compositions, on the other hand, if they are too small, many reasonable compositions may not be included in the frame. Thus, we should determine appropriate values for t 1 and t 2, which makes the fame contain as many reasonable compositions as possible, meanwhile as few unreasonable ones as possible.Suppose F i is the compositional frame of N i , let F =<F 1, F 2, ..., F q >, for any F i , let A F i •{a :_symbol 36 \f "Symbol" \s 10.5∃X symbol 206 \f "Symbol" \s 10.5∈}F i , a symbol 206 \f "Symbol" \s 10.5∈}X }. Intuitively, A F i is the set of the adjectives learned as the compositionalinstances of the noun in N i . For any n symbol 206 \f "Symbol" \s 10.5∈}N i , we use A n to denote the set of all the adjectives which in fact can modify n to form a meaningful phrase, we now define deficiency rate and redundancy rate of F . For convenience, we use symbol 100 \f "Symbol" \s 10.5δF to represent symbol 100 \f "Symbol" \s 10.5δ(, ).Definition 5 Deficiency rate symbol 97 \f "Symbol" \s 10.5αFsymbol 97 \f "Symbol" \s 10.5αF = n N n F i q nn N i iA A A∈≤≤∈∑∑∑−1Intuitively, symbol 97 \f "Symbol" \s 10.5αF refers to the ratio between the reasonable compositions which are not learned and all the reasonable ones.Definition 6 Redundancy rate symbol 98 \f "Symbol" \s 10.5βFsymbol 98 \f "Symbol" \s 10.5βF •n N F ni q Fi i A A ∈≤≤∑∑−1δIntuitively, symbol 98 \f "Symbol" \s 10.5βF refers to the ratio between unreasonable compositions which are learned and all the learned ones.So the problem of estimating t 1 and t 2 can be formalized as (5):(5) to find t 1 and t 2, which makes symbol 97 \f "Symbol" \s 10.5αF =0, and symbol 98 \f "Symbol" \s 10.5βF =0.But, (5) may exists no solutions, because its constraints are two strong, on one hand, the sparseness of instances may cause symbol 97 \f "Symbol" \s 10.5αF not to get 0 value, even if t 1and t 2 close to 1, on the other hand, the difference between words may cause symbol 98 \f "Symbol" \s 10.5βF not to get 0 value, even if t 1 and t 2 close to 0. So we need to weaken (5).In fact, both symbol 97 \f "Symbol" \s 10.5αF and symbol 98 \f "Symbol" \s 10.5βF can be seen as the functions of t 1 and t 2, denoted as symbol 97 \f "Symbol" \s 10.5αF (t 1, t 2) and symbol 98 \f "Symbol" \s 10.5βF (t 1,t 2) respectively. Given some values for t 1 and t 2, we can compute symbol 97\f "Symbol" \s 10.5αF and symbol 98 \f "Symbol" \s 10.5βF . Although there may exist no values (t 1*, t 2*) for (t 1, t 2), such that symbol 97 \f "Symbol" \s 10.5αF (t 1*, t 2*)=symbol 98 \f "Symbol" \s 10.5βF (t 1*, t 2*)=0, but with t 1 and t 2 increasing, symbol 97 \f "Symbol" \s 10.5αF tends to decrease, while symbol 98 \f "Symbol" \s 10.5βF tends to increase. So we can weaken (5) as (6).(6) to find t 1 and t 2, which maximizes (7). (7) ααF t t T t t F t t T t t t t T t t t t T t t (,)(,)(,)(,)(,)(,)**(,)(,)******12112122121211212212∈∈∑∑−whereT 1(t 1*, t 2*)={(t 1, t 2):_0symbol 163 \f "Symbol" \s 10.5≤}t 1symbol 163\f "Symbol" \s 10.5≤}t 1*, 0symbol 163 \f "Symbol" \s 10.5≤}t 2symbol 163 \f "Symbol" \s 10.5≤}t 2*}, T 2(t 1*, t 2*)={(t 1, t 2):_t 1*<t 1symbol 163 \f "Symbol" \s 10.5≤}1, t 2*<t 2symbol 163 \f "Symbol" \s 10.5≤}1}Intuitively, if we see the area ([0, 1]; [0, 1]) as a sample space for t 1 and t 2, T 1(t 1*, t 2*) and T 2(t 1*, t 2*) are its sub-areas. So the former part of (7) is the mean deficiency rate of the points in T 1(t 1*, t 2*), and the latter part of (7) is the mean deficiency rate of the points in T 2(t 1*, t 2*). To maximize (7) means to maximize its former part, while to minimize its latter part. So ourweakening (5) into (6) lies in finding a point (t 1*, t 2*), such that the mean deficiency rate of thesample points in T 2(t 1*, t 2*) tends to be very low, rather than finding a point(t 1*, t 2*), such that its deficiency rate is 0.5 Experiment Results and EvaluationWe randomly select 30 nouns and 43 adjectives, and retrieve 164 compositions(see Appendix I)between them from Xiandai Hanyu Cihai (Zhang et al. 1994), a word composition dictionary of Chinese. After checking by hand, we get 342 reasonable compositions(see Appendix I), among which 177 ones are neglected in the dictionary. So the sufficiency rate 2(denoted as symbol 103 \f "Symbol" \s 10.5γ) of these given compositions is 47.9%.We select 0.95 as the value of t 3, and let t 1=0.0, 0.1, 0.2, …, 1.0, t 2=0.0, 0.1, 0.2, …, 1.0respectively, we get 121 groups of values for symbol 97 \f "Symbol" \s 10.5αF and symbol 98 \f "Symbol" \s 10.5βF . Fig.1 and Fig.2 demonstrate the distribution of symbol 97 \f "Symbol" \s 10.5αF and symbol 98 \f "Symbol" \s 10.5βF respectively.2 Sufficiency rate refers to the ratio between given reasonable compositions and all reasonable ones.19deficiencyratet2t1 Fig. 1 The distribution of symbol 97 \f "Symbol" \s 10.5αF.19redundancerate(%)t2(1/10)t1(1/10)Fig. 2 The distribution of symbol 98 \f "Symbol" \s 10.5βF.For any given t1, and t2,we found (7) get its biggest value when t1=0.4 and t2=0.4, so we select 0.4 as the appropriate value for both t1 and t2. The result is listed in Appendix II. From Fig.1 and Fig.2, we can see that when t1=0.4 and t2=0.4, both symbol 97 \f "Symbol" \s 10.5αF and symbol 98 \f "Symbol" \s 10.5βF get smaller values. With the two parameters increasing, symbol 97 \f "Symbol" \s 10.5αF decreases slowly, while symbol 98 \f "Symbol" \s 10.5βF increases severely, which demonstrates the fact that the learning of new compositions from the given ones has reached the limit at the point: the other reasonable compositions will be learned at a cost of severely raising the redundancy rate.From Fig.1, we can see that symbol 97 \f "Symbol" \s 10.5αF generally increases as t1 and t2 increase, this is because that to increase the thresholds of the distances between clusters means to raise the abstract degree of the model, then more reasonable compositions will be learned. On the other hand, we can see from Fig.2 that when t1symbol 179 \f "Symbol" \s 10.5≥}0.4, t2symbol 179 \f "Symbol" \s 10.5≥}0.4, symbol 98 \f "Symbol" \s 10.5βF roughly increases as t1 and t2 increase3,3 On some points, it may be not the case.。

ChatGPT技术对话生成中的实时响应性解析近年来，人工智能技术的快速发展为我们的生活带来了许多便利和创新。

在人工智能领域中，对话生成技术是一个备受关注的研究方向。

ChatGPT作为一种先进的对话生成模型，具有强大的语言理解和生成能力，但其实时响应性一直是一个需要解决的难题。

在对话生成中，实时响应性是指模型能够在与用户进行对话时快速生成回应，使对话更加流畅和自然。

然而，由于ChatGPT模型的复杂性和计算资源的限制，实时响应性一直是一个挑战。

为了解决这个问题，研究人员提出了一系列方法和技术。

首先，对话生成中的实时响应性解析需要考虑模型的响应时间。

传统的ChatGPT模型通常需要较长的推理时间，导致对话的延迟和不连贯。

为了提高实时性，研究人员提出了一种基于缓存的方法。

该方法通过将用户的输入和模型的中间状态保存在缓存中，以便在下一次对话中重用，从而减少计算时间和延迟。

这种方法可以显著提高模型的实时响应性，使对话更加流畅。

其次，实时响应性解析还需要考虑对话的上下文理解和生成。

在对话中，理解和生成的速度对于实时响应性至关重要。

为了提高对话的实时性，研究人员提出了一种基于预测的方法。

该方法通过预测用户的下一个输入，提前生成可能的回应，并在用户输入后进行调整。

这种方法可以减少生成的时间和计算量，提高对话的实时性。

此外，实时响应性解析还需要考虑对话的语境和语言表达。

在对话中，语境和语言表达的准确性对于实时响应性至关重要。

为了提高对话的实时性，研究人员提出了一种基于语境感知的方法。

该方法通过对对话的语境进行建模，理解用户的意图和需求，并生成与之相符的回应。

这种方法可以提高对话的实时响应性，使对话更加连贯和自然。

综上所述，ChatGPT技术对话生成中的实时响应性解析是一个具有挑战性的问题。

为了提高实时性，研究人员提出了基于缓存、基于预测和基于语境感知的方法。

这些方法可以显著提高模型的实时响应性，使对话更加流畅和自然。

new reference用法New Reference的用法作为一名资深的创作者，你可能已经听说过”new reference”这个词，但你知道它的真正用法吗？在这篇文章中，我将详细介绍”new reference”的用法，并为你提供一些实用的例子。

1. 创建新的引用在许多编程语言中，我们可以使用”new reference”来创建新的引用。

它通常用于动态分配内存或创建新的对象。

下面是一个示例：new reference exampleRef = new Object();在这个例子中，我们使用”new reference”关键字来创建一个名为exampleRef的引用，该引用指向一个新创建的Object对象。

2. 引用参数“new reference”还可以用于函数的参数。

通过使用引用参数，我们可以直接传递引用，而不是复制整个对象。

这对于操作大型对象或需要更改对象状态的函数非常有用。

下面是一个示例：void modifyObject(new reference objRef) {= newValue;}在这个例子中，我们定义了一个名为modifyObject的函数，并使用”new reference”关键字来声明一个引用参数objRef。

在函数内部，我们可以通过该引用来修改对象的属性。

3. 引用的复制与传递使用”new reference”关键字，我们可以将一个引用的值复制到另一个引用中。

这样，两个引用将指向相同的对象。

这在需要多个引用操作同一个对象时非常有用。

下面是一个示例：new reference ref1 = someObject;new reference ref2 = ref1;在这个例子中，我们将ref1的值复制到了ref2中，所以它们都指向相同的someObject对象。

这意味着，通过任何一个引用修改someObject，都会影响到其他引用。

4. 引用的销毁使用”new reference”关键字创建的引用，可以通过销毁来释放对象所占用的内存。

Python中__new__方法详解及使用在Python中，每个类都会有一个特殊的方法叫做`__new__`。

这个方法会在对象实例化之前被调用，并返回一个类的实例对象。

在本文中，我将详细介绍`__new__`方法以及如何使用它。

### `__new__`方法的作用`__new__`方法在对象实例化之前被调用，它的作用是创建并返回一个类的实例对象。

通常情况下，我们不需要重写`__new__`方法，因为Python会自动生成一个默认的`__new__`方法来完成这个任务。

但是，有些情况下，我们可能需要重写`__new__`方法来控制对象的创建过程。

### `__new__`方法的使用`__new__`方法常用于以下几种情况：1.自定义不可变类型的对象创建过程2.控制对象的创建过程3. 与`__init__`方法配合使用下面我们将逐一介绍这些情况。

####1.自定义不可变类型的对象创建过程在Python中，不可变类型的对象（如字符串、元组等）是无法改变的。

当我们尝试改变一个不可变对象时，实际上是创建了一个新的对象。

这就涉及到了`__new__`方法。

通过重写`__new__`方法，我们可以自定义不可变类型对象的创建过程。

例如，我们可以创建一个不可变的Point对象，它包含x和y两个坐标，并且无法改变。

```pythonclass Point:def __new__(cls, x, y):instance = super(.__new__(cls)instance.x = xinstance.y = yreturn instancep = Point(1, 2)print(p.x, p.y) # 输出：1 2p.x = 3 # 报错：AttributeError: can't set attribute```在上面的例子中，我们重写了Point类的`__new__`方法。

这个方法先调用了`super(.__new__(cls)`来创建一个Point对象的实例，然后再为这个实例对象添加x和y属性，并最后返回实例对象。

同态增晰耦合改进的区域生长的键盘字符目标识别算法摘要：键盘作为诸多智能设备指令输入端口，其表面字符完整程度直接影响到设备使用。

为了提高键盘在制造过程中的生产质量，需要对其生产流水线完成视觉检测。

而在当前键盘表面字符目标检测过程中存在成像模糊、分割不准确以及识别率低等问题。

对此，设计基于同态增晰与区域生长的键盘按键识别检测算法。

首先，引入同态增晰算法，对模糊图像清晰化处理。

然后嵌入全局特征，改进了区域生长算法，准确分割并提取出图像中数字按键目标区域。

最后基于最近邻算法对数字按键图像样本库进行机器学习，完成按键数字识别，从而建立起键盘数字按键质量检查系统。

实验数据显示，与当前数字识别算法相比，面对成像模糊的按键数字图像时，该数字按键检查算法具备更高的准确性与稳定性。

关键词：键盘识别；同态增晰；区域生长；最近邻算法；图像样本库；质量检查中图分类号：TN911?34；TP391 文献标识码： A 文章编号：1004?373X（2016）12?0143?06Abstract：The keyboard is the instructions input port of various intelligent devices，and its surface characters complete degree can directly affect on device use. In order to improvethe production quality of the keyboard in manufacturing process，it is necessary to complete the visual inspection of the production line. There are problems of blurred imaging，inaccurate segmentation and low recognition rate in the current target detection of keyboard surface characters. To solve the above problems，the keyboard’s key identification and detection algorithm based on homomorphic clarification and region growing was designed. Firstly the homomorphic clarification algorithm is introduced to do clear processing of fuzzy image. And then the global feature is embedded to improve the region growing algorithm，and accurately segment and extract the target region of number keys in the image. Finally，the machine learning for the image sample library of number key is conducted based on nearest neighbor algorithm （NNA）to identify the key number and establish the quality inspection system for the number keys on keyboard. The experimental data shows that，in comparison with the current number recognition algorithm，the proposed number key inspection algorithm has higher accuracy and stability while identifying the key’s number image with blurred imaging.Keywords：keyboard identification；homomorphic clarification；region growing；nearest neighbor algorithm；image sample library；quality inspection0 引言随着电子行业与制造业的快速发展，智能设备也在不断进驻生产车间，在降低企业人力成本与提高制造质量的同时，也为推进物联网、大数据建设做好了基础准备[1?3]。

Challenges in Building an Arabic-English GHMT Systemwith SMT ComponentsNizar Habash†,Bonnie Dorr‡,Christof Monz§†Center for Computational Learning Systems,Columbia Universityhabash@‡Department of Computer Science,University of Marylandbonnie@§Department of Computer Science,Queen Mary,University of Londonchristof@AbstractThe research context of this paper is de-veloping hybrid machine translation(MT)systems that exploit the advantages oflinguistic rule-based and statistical MTsystems.Arabic,as a morphologicallyrich language,is especially challengingeven without addressing the hybridiza-tion question.In this paper,we describethe challenges in building an Arabic-English generation-heavy machine trans-lation(GHMT)system and boosting itwith statistical machine translation(SMT)components.We present an extensiveevaluation of multiple system variants andreport positive results on the advantages ofhybridization.1IntroductionThe research context of this work is developing hy-brid machine translation(MT)systems that exploit the advantages of linguistic rule-based and statisti-cal MT systems.Arabic,as an example of a mor-phologically rich language,is especially challeng-ing even without addressing the hybridization ques-tion.In this paper,we describe the challenges in building an Arabic-English generation-heavy ma-chine translation(GHMT)system(Habash,2003a) and extending it with statistical machine translation (SMT)components.A major challenge for working with Arabic is the proliferation of inconsistent morphological repre-sentations in different resources and tools for Arabic natural language processing(NLP)(Habash,2006). This inconsistency is only heightened when trying to interface linguistically-aware MT approaches with surface-based statistical MT approaches,where the level of representation of the phrase(beyond the word)is not consistent.We describe how we ad-dress this issue in our system and present an exten-sive evaluation addressing its various strengths and weaknesses.We show positive improvements when extending our basic GHMT system with SMT com-ponents.The remainder of this paper is organized as fol-lows:the next section(Section2)discusses previ-ous work on hybridization in MT.It is followed by a discussion of Arabic-specific challenges for MT implementations in Section3.Section4describes the Arabic components of our basic GHMT system. Section5describes the extensions we made to in-tegrate SMT components into the GHMT system. Section6presents three evaluations of multiple MT system variants.2Previous WorkWe discuss research related to our approach in the areas of generation-heavy MT and MT hybridiza-tion.2.1Generation-Heavy MTGHMT is an asymmetrical hybrid approach that addresses the issue of MT resource poverty in source-poor/target-rich language pairs by exploiting symbolic and statistical target-language resources (Habash and Dorr,2002;Habash,2003a;Habash, 2003b).Expected source-language resources in-clude a syntactic parser and a simple one-to-many translation dictionary.No transfer rules or complex interlingual representations are used.Rich target-language symbolic resources such as word lexical semantics,categorial variations and subcategoriza-tion frames are used to overgenerate multiple struc-tural variations from a target-language-glossed syn-tactic dependency representation of source-language sentences.This symbolic overgeneration accounts for possible translation divergences,cases where the underlying concept or“gist”of a sentence is dis-tributed differently in two languages such as to put butter and to butter(Dorr,1993).The overgen-eration is constrained by multiple statistical target-language models including surface n-grams and structural n-grams.The source-target asymmetry of systems developed in this approach makes them more easily retargetable to new source languages (provided a source-language parser and translation dictionary).In this paper,we describe these two spe-cific extensions for Arabic in detail(Section4). 2.2MT HybridizationResearch into MT hybrids has increased over the last few years as research in the two main competing paradigms,rule-based MT and corpus-based(statis-tical)MT,is approaching a plateau in performance. In the case of statistical approaches this has recently led to approaches that not just rely on surface forms but also incorporate symbolic knowledge such as morphological information and syntactic structure. In the next two subsections,we review this body of research.Our own research however,differs in that we are approaching the hybridization question from the opposite direction,i.e.,how to incorporate SMT components into rule-based systems(Senel-lart,2006).Nonetheless,the research on SMT-based hybrids has influenced many of our decisions and di-rections.2.2.1Morphology-Based ApproachesThe anecdotal intuition in thefield is that reduc-tion of morphological sparsity often improves trans-lation quality.This reduction can be achieved by in-creasing training data or via morphologically-driven preprocessing(Goldwater and McClosky,2005). Recent investigations of the effect of morphology on SMT quality focused on morphologically rich lan-guages such as German(S.Nießen,2004);Span-ish,Catalan,and Serbian(Popovi´c and Ney,2004); and Czech(Goldwater and McClosky,2005).These studies examined the effects of various kinds of to-kenization,lemmatization and part-of-speech(POS) tagging and showed a positive effect on SMT qual-ity.Lee(2004)investigated the use of automatic alignment of POS tagged English and affix-stem segmented Arabic to determine appropriate tok-enizations of Arabic.Her results showed that mor-phological preprocessing helps,but only for smaller corpora.Habash and Sadat(2006)reached simi-lar conclusions on a much larger set of experiments including various preprocessing schemes and tech-niques.They showed that genre-variation interacts with preprocessing decisions.Within our approach,working with Arabic mor-phology is especially challenging.We discuss this issue in more detail in Section3.2.2.2Syntax-Based ApproachesMore recently a number of statistical MT ap-proaches included syntactic information as part of the preprocessing phase,the decoding phase or the n-best rescoring phase.Collins et al.(2005)incorporate syntactic infor-mation as part of preprocessing the parallel corpus.A series of transformations on the source parse trees are applied to make the order of the source language side closer to that of the target language.The same reordering is done for a new source sentence before decoding.They show a modest statistically signifi-cant improvement over basic phrase-based MT. Quirk et al.(2005)use sub-graphs of dependency trees to deal with word order differences between the source and the target language.During training, dependency graphs on the source side are projected onto the target side by using the alignment links be-tween words in the two languages.The use of syn-tactic information is the main difference between their approach and phrase-based statistical MT ap-proaches.During decoding,the different sub-graphs are combined in order to generate the most likely dependency tree.This approach has been shown to provide significant improvements over a phrase-based SMT system.Och et al.(2004)experimented with a wide range of syntactic features to rescore the n-best lists gener-ated by their statistical MT system.Although some features—e.g.,POS tags and parse-tree to string mappings—lead to slight improvements over the baseline,larger improvements are obtained by using simpler,non-syntactic features,such as IBM Model 1alignments.Similar to Collins et al.(2005)and Quirk et al. (2005),our approach uses source-language syntac-tic(specifically dependency)representations to cap-ture generalizations about the source-language text. Unlike both of them,we do not use or learn specific mappings between the syntactic structure of source and target languages.Instead,our approach maps the source language to a syntactically language-independent representation which forms the basis for target-language generation.3Arabic ChallengesArabic is a morphologically complex language with a large set of morphological features.These fea-tures are realized using both concatenative(affixes and stems)and templatic morphology(root and pat-terns)with a variety of morphological and phono-logical adjustments that appear in word orthography and interact with orthographic variations.As a re-sult,there are many different possible representa-tions of Arabic morphological tokens that have been used in different resources for Arabic NLP(Habash, 2006).For statistical MT,in principle,it does not matter what level of morphological representation is used so long as the input is on the same level as that of the training data.However,in practice,there are certain concerns with issues such as sparsity,ambiguity,and training data size.Symbolic MT approaches tend to capture more abstract generalities about the lan-guages they translate between compared to statisti-cal MT.This comes at a cost of being more com-plex than statistical MT,involving more human ef-fort,and depending on already existing resources for morphological analysis and parsing.This dependence on existing resources highlights the problem of variation in morphological represen-tations for Arabic.In a typical situation,the in-put/output text of an MT system is in simple white-space tokenization.But,a statistical parser(such as (Collins,1997)or(Bikel,2002))trained out-of-the-box on the Penn Arabic Treebank(Maamouri et al., 2004)assumes the same kind of tokenization it uses (4-way normalized segments into conjunction,parti-cle,word and pronominal clitic).This means,a sep-arate tokenizer is needed to convert input text to this representation(Habash and Rambow,2005;Diab et al.,2004).An additional issue with a treebank-trained sta-tistical parser is that its input/output is in normal-ized segmentation that does not contain morpholog-ical information such as features or lexemes that are important for translation:Arabic-English dictionar-ies use lexemes and proper translation of features, such as number and tense,which requires access to these features in both source and target languages. As a result,additional conversion is needed to relate the normalized segmentation to the lexeme and fea-ture levels.Of course,in principle,the treebank and parser could be modified to be at the desired level of representation(i.e.lexeme and features).But this may be a labor-intensive task for researchers inter-ested in MT.4Extending GHMT to ArabicAs described earlier,our English-targeted GHMT system can be used with a new source language given that a dependency parse and a word-based translation lexicon are provided.In the following sub-sections,we describe these two components in our Arabic-English GHMT system.The reusable English generation component is called EXERGE (Expansive Rich Generation for English),which is discussed in detail in(Habash,2003a).4.1Analysis IssuesThis sub-section describes the necessary steps for processing an Arabic input sentence.4.1.1Tokenization and POS TaggingFor tokenization,we use the Penn Arabic Tree-bank(PATB)tokenization scheme,which is most compatible with statistical parsers trained on the PATB(Maamouri et al.,2004).For the POS tagset, we use the collapsed tagset for PATB(24tags).We use the Morphological Analysis and Disambiguation (MADA)tool for Arabic preprocessing(Habash andRambow,2005)together with TOKAN,a general to-kenizer for Arabic(Habash,2006).MADA uses the A LMORGEANA(Arabic Lexeme-based Morpholog-ical Analysis and Generation)which is an alternative engine to Buckwalter’s AraMorph that uses the same lexicalfiles.4.1.2ChunkingWe employ a rule-based segment chunker to ad-dress two issues.First,the Arabic sentence length, which averages over35words with PATB tokeniza-tion(in the news genre),slows down the parser and increases its chances of producing null parses.Sec-ond,the use of punctuation and numbers in by-lines in news requires template handling in analysis and generation,which needs to be updated depending on the genre.Instead,we choose to preserve source-language order for such cases by chunking them out and treating them as special chunk separators that are translated independently.The rules currently im-plemented use the following chunk separators.POS information is used in this process.•Arabic conjunction proclitic w/CC1and •Numbers(CD)and punctuation(PUNC)•The subordinating conjunction An/IN that On average,sentences had3.3chunk separators.4.1.3ParsingFor parsing,we used the Bikel parser(Bikel, 2002)trained on the PATB(Part1).The default out-put of the parser is on unlabeled constituency repre-sentation.The tokens in the parser are surface words in the PATB tokenization scheme.4.1.4PostparsingThe specifications of EXERGE require an in-put dependency tree labeled with minimal syntac-tic relations(subj,obj,obj2,and mod).More-over,the nodes must have lexemes and features from a pre-specified set of feature names and val-ues(Habash,2003a).The output of the parsing step undergoes operations such as relation labeling and 1All Arabic transliterations in this chapter are provided in the Buckwalter transliteration scheme(Buckwalter,2002).node-structure modification.Some of these opera-tions are similar to the Spanish post-parsing process-ing for Matador(Spanish-English GHMT)(Habash, 2003b).Constituency to Dependency We convert con-stituencies to dependencies using modified head-percolation rules from Bikel parser applied with the Const2Dep tool2(Habash and Rambow,2004). Lexeme Selection MADA is only a morpho-logical disambiguation tool that makes no sense-disambiguation choices.Therefore,multiple lex-emes are still available as ambiguous options at the tree nodes.In some cases,the parser overrides the POS tag that was chosen initially by MADA.As a result,we need to re-visit discarded morphologi-cal analyses again.We re-apply the A LMORGEANA system on the tokenized words and thenfilter analy-ses using the following criteria.In case no analysis matches,all analyses are passed on to the nextfilter.•Analyses with PATB tokenizable clitics are ig-nored because the word is already tokenized.•Analyses that match the word’s POS are se-lected.Others are ignored.The POS match-ing is fuzzy since the tagset used by A L-MORGEANA(15tags)is more coarse than the PATB tagset(24tags).Also,since there arecommon cases of mismatch in Arabic,certain seemingly mismatched cases are allowed,e.g., noun,adjective and proper noun.•We use a statistical unigram lexeme and fea-ture model.The model was trained on PATB (part1and part2)and1million words from Arabic Gigaword(Graff,2003)disambiguated using MADA.The lexemes are chosen based on their unigram counts.Ties are broken with feature unigrams.Dependency Tree Restructuring The following operations are applied to the dependency tree:•Idafa Handling:The Idafa construction is a syntactic construction indicating the relation-ship of possession between two nouns,i.e., Noun1of Noun2.Nouns in this construction 2The Const2Dep tool was provided by Rebbecca Hwa.are modified to include an intervening node that has no surface value but is glossed to of/’s/*empty*.•The untokenized prefix Al+the is turned intoa separate node that is a dependent on the wordit is attached to.•Feature mapping:We map Arabic-specific fea-tures to language-independent features used in EXERGE.For example,the untokenized prefix s+will is mapped to the feature TENSE:FUT and the Arabic perfective aspect verb feature is turned into TENSE:PAST.Relation Labeling An Arabic subject may be:(a) pro-dropped(verb conjugated),(b)pre-verbal(full conjugation with verb),or(c)post-verbal(3per-son and gender agreement only).Third mascu-line/feminine singular verbs are often ambiguous as to whether they are case(a),where the adjacent noun is an object,or(c),where the adjacent noun is a subject.A verb can have no,one or two objects. Pronominal objects are always cliticized to the verb, which means they can appear between the verb and the nominal subject.For passive verbs,the sub-ject position is reserved for a*PRO*and the fea-ture is passed along.In principle,Arabic’s rich case system can account for the different configurations and also allow many variations in order,but since most cases are diacritical(and thus optionally writ-ten),that information is not always available.Arabic prose(non-poetry)writers generally avoid such syn-tactic acrobatics.We use heuristics based on Arabic syntax to deter-mine the relation of the verb to its nominal(common and proper),pronominal and relativizing children.4.1.5Subtree Phrase ConstructionEach node in the dependency tree is annotated with the full projection of the subtree it heads.This subtree phrase is later used to interface with the sta-tistical MT component.4.2Lexical Translation IssuesOne of the main challenges in resource usage was the transformation of the lexicon of the Buckwal-ter Arabic morphological analyzer(BAMA)(Buck-walter,2002)into a form that was readily usable by our GHMT system.The original Buckwalter lexi-con contained English glosses to Arabic stem entries used in morphological analysis.Since the glosses are attached to stems,they are sometime inflected for number or voice,although generally they are in lexeme form.Our initial extraction of a translation dictionary from BAMA produced a resource in the following form:[Arabic Word][Tab][POS][Tab][English Word][Tab](comment) For example:$ariyk_1N associates(female)We implemented a lexical reformatting procedure to address several issues with this lexicon.Thefirst issue is the inclusion of plural forms as in the example above—where the singular form ap-pears elsewhere in the lexicon(independently)—or related forms where the entry contains a synonym: $ariyk_1N associate$ariyk_1N partnerNote that,in addition to the redundancy inherent in these related entries,the use of parentheticals is inconsistent,e.g.,the comment“female”appears in only one of the entries above.The lexical reformatting procedure transforms these three entries into the following single line: $ariyk_1N associate/partnerwhere the plural form is assumed to be handled by GHMT during generation of the English surface form.In addition to the redundancy issues above,the material in the parentheses was often combined with a slash(/)in ways that were not uniform throughout the original lexicon.Consider the following exam-ple:<imArAtiy˜_1AJ Emirati(of/from_the_UAE)The material in the parentheses above is shorthand for“of the UAE”and“from the UAE”.Our lexical reformatting procedure transforms this entry into the following single line:<imArAtiy˜_1AJ Emirati(from_the_UAE/ of_the_UAE)Often,this same inconsistency with the slash(/) appeared in the English translation itself,as in the following entry:>a$ad˜_1N more/most_intensewhich was converted by our reformatting procedure into the following:>a$ad˜_1N more_intense/most_intense Beyond depluralizing and making slashes consis-tent,we also addressed the issue of passive conver-sion,where we transform a passive(but not cop-ula/adjective)English translation of an Arabic verb into an active form.Consider the following exam-ples from the original lexicon:>a$Ad_1V be_built>a$Ad_1V be_commended>a$Ad_1V be_praised$Abah_1V be_similarThese entries may be combined with other active forms that occur in the original dictionary:>a$Ad_1V build>a$Ad_1V commend>a$Ad_1V praise$Abah_1V resembleto yield the following two single lines:>a$Ad_1V build/commend/praise$Abah_1V be_similar/resemble5Integration of SMT Components into GHMTThe main challenge for integrating SMT compo-nents into GHMT is that the conception of the phrase (anything beyond a single word)is radically differ-ent.Phrase-based SMT systems take a phrase to be a sequence of words with no hidden underlying structure(Koehn,2004).On the other hand,for sys-tems that use parsers,like GHMT,a phrase has a linguistic structure that defines it and its behavior in a bigger context.Both kinds come with problems. Statistical phrases are created from alignments, which may not be clean.This results in jagged edges to many phrases.For example,the phrase.on the other hand,the(con-taining seven words starting with a period and end-ing with“the”)overlaps multiple linguistic phrase boundaries.Another related phenomenon is that of statistical hallucination,e.g.,the translation of AlswdAn w(literally,Sudan and)into enterprises and banks.Linguistic phrases come with a different set of problems.Since parsing technology for Arabic is still behind English,3many linguistic phrases are 3The parser we used in this paper is among the best avail-able,yet its performance for Arabic is in the lower70s percent mis-parsed creating symbolic hallucinations that af-fect the rest of the system.A common example is in-correctly attaching a prepositional phrase that mod-ifies a complete sentence to one of its noun phrases. We investigate two variants of a basic approach to using statistical phrases in the GHMT system.As phrase-based SMT system we use,Pharaoh(Koehn, 2004).We limit the statistical translation-table phrases used to those that correspond to completely projectable subtrees in the linguistic dependency representation of the input sentence.More complex solutions that use statistical phrases covering parts of a linguistic phrase are left for future work.In thefirst variant,(G HMT+Phrase Table,hence-forth G HMT P H T),we use the phrase table produced by Pharaoh as a multi-word surface dictionary.In the generation process,when a subtree is matched to an entry in this dictionary,an additional path in the generation lattice is created using the phrase-table entry in addition to the basic GHMT generation.In the second variant,(G HMT+Pharoah,hence-forth G HMT P HAROH),we use a phrase-based SMT system(Koehn,2004)to translate the subtree pro-jections for all the subtrees in the input sentence. These translations are added as alternatives to the basic G HMT system.Results comparing these two variants and a few others are described in Section6. The basic idea here is to exploit GHMT’s focus on phrase structure generation(global level)together with a phrase-based SMT system’s robustness(lo-cal phrases).One particular case in Arabic that we investigate later is the position of the subject relative to the verb.When we have a correct parse,moving the subject,which follows the verb in Arabic over 35%of the time,to a preverbal position is easy for GHMT(given a correct parse)but can be hard for a phrase-based SMT system,especially with sub-ject noun phrases exceeding the system’s distortion limit.6EvaluationWe use the standard NIST MTEval datasets for the years2003,2004and2005(henceforth MT03, MT04and MT05,respectively).4The2002MTEval test set was used for Minimum Error Training(Och, (labeled constituency PARSEV AL F-1score).4/speech/tests/mt/Table1:True-cased results of various systems on NIST MTEval test setsTest Set Metric G IST G HMT G HMT P H T G HMT P HAROH P HARAOH B W P HAROAH MT03BLEU0.08110.14790.23620.33790.41280.4162 NIST 5.1846 6.05287.32138.25699.92059.9300 MT04BLEU0.06510.14020.21100.27770.35460.3522 NIST 4.3904 6.09357.09817.58349.20389.1291 MT05BLEU0.06070.1450.23130.32390.39350.3960 NIST 4.7259 6.26367.48368.36879.69809.6615 Table2:Genre-specific true-cased results of various systems on NIST MT04test set Genre Metric G IST G HMT G HMT P H T G HMT P HAROH P HARAOH B W P HAROAH News BLEU0.08170.16170.25820.34340.42660.4244 NIST 4.8989 6.3587.61438.31329.72069.6796 Speech BLEU0.04290.12760.18210.24470.30880.3043 NIST 3.2993 5.3923 6.2022 6.63547.87967.7164 Editorial BLEU0.05750.11440.15420.19140.27040.2703 NIST 3.7633 4.9751 5.4724 5.46087.23447.18122003).All of the training data used here are available from the Linguistic Data Consortium(LDC).We use an Arabic-English parallel corpus of about5mil-lion words to train the translation model.5For Arabic preprocessing the Arabic Treebank scheme was used,see(Habash and Sadat,2006).All sys-tems use the same surface trigram language model, trained on approximately340million words of En-glish newswire text from the English Gigaword cor-pus.6English preprocessing simply included down-casing,separating punctuation from words and split-ting off“’s”.Trigram language models are imple-mented using the SRILM toolkit(Stolcke,2002). Both BLEU(Papineni et al.,2002;Callison-Burch et al.,2006)and NIST(Doddington,2002)metric scores are reported.All scores are computed against four references with n-grams of maximum length four.As a post-processing step,the translations of all systems are true-cased,and all results reported below refer to the case-sensitive BLEU and NIST scores.We conducted three sets of evaluations that ex-plore different aspects of the data sets and the system variants:a full system evaluation,a genre-specific 5The parallel text includes Arabic News,eTIRR,English translation of Arabic Treebank,and Ummah.6Distributed by the Linguistic Data Consortium: .evaluation,and a qualitative evaluation of specific linguistic phenomena.6.1Full EvaluationSix system variants are compared:•G IST is a simple gisting system that produces a sausage lattice from the English glosses in the output of the Buckwalter Arabic morphological analyzer(BAMA).Arabic word order is pre-served and English realization is limited to the variants provided in BAMA.•G HMT is the system described in Section4.The lexical translation is limited to the Buck-walter lexicon.•G HMT P H T is a variant of G HMT that uses a statistical phrase table as support multi-word surface dictionary(see Section5).•G HMT P HAROH is the second variant discussed in Section5.It uses Pharaoh to generate sub-tree phrases.•P HARAOH B W is the phrase-based SMT system Pharaoh trained on the basic training set in ad-dition to the entries in the Buckwalter lexicon.•P HAROAH is the phrase-based SMT system Pharaoh trained only on the basic training set.The results of the full systems are presented in Table1.The lowest performing system is G IST,as expected.G HMT,using only the Buckwalter lexicon and no other training data,more than doubles the G IST score.This indicates that the system is actually making more correct lexical choices and word order realization beyond simple gisting.G HMT P H T and G HMT P HAROH provide substan-tial improvements over G HMT.In G HMT P H T,only 54.6%of subtreesfind a match in the phrase table;as opposed to G HMT P HAROH which guarantees a sta-tistical translation for all subtrees.This accounts for the large difference between the two scores.This is a positive result for improving a non-statistical MT system with SMT components.However,the scores are still lower than the fully statistical system.We discuss the differences further in Section6.3.The primarily statistical systems P HAROAH and P HARAOH B W outperform all other systems. P HAROAH does better than P HARAOH B W for MT03 and MT05but not for MT04.In all three cases,the differences are not statistically significant.As the amount of dependence on training data in-creases,we see a bigger divide between the differ-ent data sets.MT03and MT05behave similarly but MT04lags behind.One of the reason behind this behavior is that MT04is a mixed genre data set.In the next section,we examine the differences in the genres in more detail.6.2Genre EvaluationThe MTEval2004data set is special in that it has a mix of genre(200documents:100news,50 speeches and50editorials).The training data we used is all Arabic news.We wanted to investigate the difference in behavior among variants with different types of symbolic and statistical resources.Table2 presents the scores for genre-specific subsets of the MT04test set.The difference in scores across the different sys-tems is consistent with the full evaluation in Table1. The difference across the genre is very clear,with the news subset performing at a similar score level to that of the MT03and MT05test sets in Table1. Upon examination of the documents in MT04,we see several variations across the genres that explain the differences.Particularly,speeches and editori-als have a much higher rate offirst and second per-son pronouns and verbs,include interrogative sen-tences,and use moreflowery andfiery language than news.Out-of-vocabulary(OOV)rates in the the dif-ferent subsets as measured against the basic train-ing set data is as follows:news(2.02%),speeches (2.01%)and editorials(2.34%).The differences are very small.This confirms that it is style/use differ-ence that is the biggest contributor to the difference in scores.The fact that we see similar differences in G IST and G HMT as in P HAROAH contradicts our hypothe-sis that G HMT is more genre-independent than SMT approaches.We believe this is a result of the Ara-bic linguistic resources we use being biased towards news-genre.For example,the Arabic treebank used for training the parser is only in the news genre.The Buckwalter lexicon potentially also has some inter-nal bias toward news genre because it was developed in tandem with the Arabic treebank.6.3Qualitative EvaluationAutomatic evaluation systems are often criticized for not capturing linguistic subtleties.This is clearly apparent in thefield’s moving back toward using hu-man evaluation metrics such as HTER(Snover et al., 2006).We conducted a small evaluation of verb and subject realization in eight random documents from MT04.The documents contained47sentences and reflect the distribution of genre in the MT04test set. We compare three systems G HMT,G HMT P HAROH and P HAROAH.The evaluation was conducted using one bilingual Arabic-English speaker(native Ara-bic,almost native English).The task is to deter-mine for every verb that appears in the Arabic input whether it is realized or not in the English transla-tion.If realized,we then determine whether its sub-ject is mapped to it correctly.Since translation diver-gences can cause an Arabic verb to appear as a noun in English,a nominalized translation is accepted as a valid realization.The subject of a non-verbal trans-lation is considered correctly assigned if the mean-ing of the relationship of the original subject-verb pair is preserved.Correct realization of the verb ob-ject was not considered here,and neither was non-verbal Arabic translations to verb forms in English. The results are presented in Table3for each genre and also collectively.For each of the three sys-tems studied,two columns are presented.Thefirst。

2021⁃01⁃10计算机应用,Journal of Computer Applications 2021,41(1):43-47ISSN 1001⁃9081CODEN JYIIDU http ：//基于邻居信息聚合的子图同构匹配算法徐周波，李珍，刘华东*，李萍（广西可信软件重点实验室（桂林电子科技大学），广西桂林541004）（∗通信作者电子邮箱ldd@ ）摘要：图匹配在现实中被广泛运用，而子图同构匹配是其中的研究热点，具有重要的科学意义与实践价值。

现有子图同构匹配算法大多基于邻居关系来构建约束条件，而忽略了节点的局部邻域信息。

对此，提出了一种基于邻居信息聚合的子图同构匹配算法。

首先，将图的属性和结构导入到改进的图卷积神经网络中进行特征向量的表示学习，从而得到聚合后的节点局部邻域信息；然后，根据图的标签、度等特征对匹配顺序进行优化，以提高算法的效率；最后，将得到的特征向量和优化的匹配顺序与搜索算法相结合，建立子图同构的约束满足问题（CSP ）模型，并结合CSP 回溯算法对模型进行求解。

实验结果表明，与经典的树搜索算法和约束求解算法相比，该算法可以有效地提高子图同构的求解效率。

关键词：子图同构；约束满足问题；图卷积神经网络；信息聚合；图匹配中图分类号：TP391文献标志码：ASubgraph isomorphism matching algorithm based on neighbor informationaggregationXU Zhoubo ，LI Zhen ，LIU Huadong *，LI Ping（Guangxi Key Laboratory of Trusted Software （Guilin University of Electronic Technology ），Guilin Guangxi 541004，China ）Abstract:Graph matching is widely used in reality ，of which subgraph isomorphic matching is a research hotspot and has important scientific significance and practical value.Most existing subgraph isomorphism algorithms build constraints based on neighbor relationships ，ignoring the local neighborhood information of nodes.In order to solve the problem ，a subgraph isomorphism matching algorithm based on neighbor information aggregation was proposed.Firstly ，the aggregated local neighborhood information of the nodes was obtained by importing the graph attributes and structure into the improved graph convolutional neural network to perform the representation learning of feature vector.Then ，the efficiency of the algorithm was improved by optimizing the matching order according to the characteristics such as the label and degree of the graph.Finally ，the Constraint Satisfaction Problem （CSP ）model of subgraph isomorphism was established by combining the obtained feature vector and the optimized matching order with the search algorithm ，and the model was solved by using the CSP backtracking algorithm.Experimental results show that the proposed algorithm significantly improves the solving efficiency of subgraph isomorphism compared with the traditional tree search algorithm and constraint solving algorithm.Key words:subgraph isomorphism;Constraint Satisfaction Problem (CSP);graph convolutional neural network;information aggregation;graph matching0引言图匹配技术被广泛地应用于社交网络、网络安全、计算生物学和化学等领域［1］中。

Syntactic FunctionAbstract: Syntactic function plays a role in the sentence. Also it is important for English learning. One sentence has many components. Different components have different functions. And we can analysis these components by syntactic function so that we can understand the sentence well. In other hand it is help us to improve our writing skill.Key words: syntactic, function, subject, predicate, object, direct object, indirect object, the relation between classes and functions1. IntroductionForm word to text, words, word groups and phrases, and clauses can not occur at random, they have to follow certain rules of ordering. Syntax is the study of the rules governing the ways different constituents are combined to form sentences in a language, or the study of the interrelationships between elements in sentence structures. It included syntactic relations, grammatical construction and its constituents, syntactic function and so on. I want to introduce syntactic function.2. Syntactic FunctionThe syntactic function shows the relationship between a linguistic form and other parts of the linguistic pattern in which it is used. Names of functions are expressed in term of subjects, objects, predicator and others. I will explain them one by one.2.1 SubjectFirst, the definition of subject, in some languages, it refers to one of the nouns in the nominative case. We can find the typical example in Latin, such as pater and filius. In our text book, there are two examples. (a) Pater filium amat. In English, it means the father loves the son. We can know that the subject is the father. (b) Paterm filius amat. It refers to the son loves the father. It is no doubt that the subject is the son.Of course, I reached some information. And I found there was another definition. In English, the subject of a sentence is often said to be the doer of the action, while the object is the person or thing acted upon by the doer. Let’s see the example. The first sentence, Mary slapped John. According to the definition, Mary is the subject,because she is the doer of the action.2.2 PredicatePredicate refers to a major constituent of sentence structure in a binary analysis in which all obligatory constituents other than the subject were considered together. It usually expresses actions, processes, and states that relate to the subject.Predicate has four functions. It specifies time reference other than reference to the time of the speech event, such as past tense, present tense or future tense. It specifies various additional aspects and phrases like seeming, trying, hoping and so on. Then it also specifies the voice: active or passive. As last, it specifies the progress, just like action, event, mental process, behavior, relation or existence. For instance, Peter broke the glass. The predicate is what says something about the subject. The subject is Peter, so the rest of this sentence is the object “broke the glass”. Broke shows the past tense.2.3 ObjectThere is no exact definition of object. Traditionally, subject can be defined as the doer of an action, object may refer to the “receiver” or “goal” of an action. I’d like to give an example. Mother bought a doll. We can see the object is a doll.Another example is that mother gave my sister a doll. So the object is my sister a doll. And it further classified into direct object and indirect object. There is one example. Mother gave a doll to my sister.There are two objects, a doll and my sister. A doll is the indirect object. And my sister is the direct object. We usually put the indirect object in front of the direct object. Sometimes, the object can become the subject in a passive transformation. The origin sentence is John broke the glass. The object is the glass. If we change this sentence into passive. The glass was broke by John. We can notice that the object is the glass.2.4 The Relation between Classes and Functions.Classes and functions each other, but not in any one-to-one relation. In other words, one class has many functions. I am willing to give some examples. The boys are playing football.The boys and football are all nouns. But they have different functions. In this sentence, the boys is the subject, while the football is the object. Hecame here last month. The object is last month. In this sentence it is an adverbial. The Summer Palace.Summer is a noun. And in this phrases, it is also a modifier.At the same time, a function can be fulfilled by several classes. For, instance, the subject of a sentence can be realized by a noun, pronoun, numeral, infinitive and so on. There are some examples. The dogs are barking. The dogs is the subject. Also it is noun. We will stay here. We is both the subject and pronoun.3. ConclusionAs far as I am concerned, syntactic function is very important for English learners. First of all, it is good for our study of words. In English, there are many different kinds of words, such as subject, object, predicate and so on. For me, it is difficult to distinguish them clearly. In my beginning of studying English, I was so confused with these. To be honest, I also do not know the syntactic of Chinese. In syntactic function, some definitions express clearly about them. And also there are some examples which are simple sentences so that we can understand well. Different kinds of words have different characteristic, and they can be not changed. Secondly, it is benefit for our study of grammar. When we study the syntactic function, it is also a process to learn grammar. At last, it will help us to understand English sentence. In the simple sentence, there is one subject, one object and one predicate. It is easy for us to understand. However, in the complex sentence, it is quite different. Maybe the object is one sentence. If we can learn syntactic function well, the complex sentence is easy to understand. In a word, syntactic function is important for us.Reference:Linguistics A Course Book Fourth Edition 北京大学出版社2012/p-381366690.html/view/6b9f14ed172ded630b1cb65a.html。

[推荐]计算机专业英语1500词1. file n. 文件；v. 保存文件2. command n. 命令，指令3. use v. 使用，用途4. program n. 程序5. line n. (数据，程序)行，线路6. if conj. 如果7. display vt. 显示，显示器8. set v. 设置，n. 集合9. key n. 键，关键字，关键码10. list n. 列表，显示，v. 打印11. by prep. 凭，*，沿12. press v. 按，压13. with prep. 用，与，随着14. format n. 格式15. change v. 更换，改变，变动16. cursor n. 光标17. directory n. 目录，索引簿18. from prep. 从，来自，以来19. menu n. 菜单，目录20. option n. 任选，选择，可选项21. character n. 字符，符号，特性22. current n. 电流23. type n. 型，类型；v. 打印24. screen n. 屏幕，屏；v. 屏蔽25. specify v. 指定，规定，确定26. move v. 移动27. disk n. 盘，磁盘28. text n. 正文，文本29. drive v. 驱动；n. 驱动器30. see v. 看，看出，查看31. name n. 名，名称；vt. 命名32. record n. 记录33. box n. 箱，匣，(逻辑)框34. database n. 数据库35. help v. & n. 帮助36. memory n. 记忆存储，存储器37. which pron. 哪个，a. 那一个38. all a. 全，全部；ad. 完全39. on ad. 接通，导电，开40. copy n. 复制，v. 拷贝41. shell n. 壳，外壳42. delete vt. 删除，删去，作废43. enter v. 键入，送入44. margin n. 余量，边缘，边际45. mark n. 标记；vt. 加标记46. also ad. & conj. 也，亦，还47. do v. 做，干；n. 循环48. information n. 信息，情报49. choose v. 挑选，选择，选定50. select vt. 选择51. group n. 组，群52. first a. & ad. & n. 第一，首先53. field n. 字段，域，栏，场54. procedure n. 过程，程序，工序55. print v. 打印，印刷56. return v. 返回，回送57. number n. 数字，号码；vt. 编号58. selected a. 精选的59. want v. 需要，应该，缺少60. window n. 窗口61. message n. 信息，消息，电文62. dialog n. & vt. 对话63. example n. 例子，实例64. create vt. 创立，建立65. insert vt. 插入66. related a. 相关的67. item n. 项，项目，条款68. edit vt. 编辑，编排，编篡69. marked a. 有记号的70. area n. (区)域，面积，方面71. parameter n. 参数，参变量72. then ad. & conj. 那时，则73. variable a. 可变的；n. 变量74. tab n. 制表键75. up ad. 上，向上，a. 高的76. string n. 行，字符串77. each a. & ad. 各(自)，每个78. active a. 激活的，活动的79. topic n. 题目，论题80. start v. 起动，开始，启动81. mode n. 态，方式，模82. selection n. 选择83. function n. 函数，功能，操作84. word n. 字(词)，单词85. make vt. 制造，形成，接通86. right a. 右边的，正确的87. value n. 值88. button n. 按钮89. index n. 索引，变址，指数90. without prep. 没有，在…以外91. appear vi. 出现，显现，好像92. left a. & n. 左边(的)93. save v. 保存94. next n. 下一次，a. 其次95. off ad. (设备)关着，脱离96. following a. 下列的，以下的97. control v. 控制，支配，管理98. only a. 唯一的，ad. 仅仅99. user n. 用户100. end n. 结束，终点，端点101. system n. 系统102. contain vt. 包含，包括103. time n. 时间；vt. 计时104. letter n. 字母，信105. data n. 数据106. setting n. 设置，调整107. desire v. & n. 期望108. position n. 位置；vt. 定位109. down ad. 落下，降低，减少110. task n. 任务；v. 派给…任务111. view n. & v. 视图，景象112. switch n. & v. 开关，转换，切换113. include vt. 包括，包含114. get v. 得到，获得，取115. default v. 缺省，预置，约定116. structure n. 结构，构造，构件117. into prep. 向内，进入118. path n. 路径，通路，轨道119. blank n. 空白，间隔120. open v. 打开，开启，断开121. add v. & n. 加，增加，添122. enable vt. 启动，恢复正常操作123. operation n. 操作，运算，动作124. erase v. 擦除，取消，删除125. filename n. 文件名126. search v. 检索，查询，搜索127. another a. 另一个，别的128. last a. & n. 最后(的)129. column n. 列，柱，栏130. after prep. & ad. 以后，后面131. prompt n. & v. 提示132. two n. & a. 二，两，双133. execute v. 实行，实施134. about ad. 关于，大约，附近135. escape v. 逃避，逸出，换码136. error n. 错误，误差，差错137. currently ad. 目前，现在138. extension n. 扩充，延伸139. same a. 同样的，相同的140. status n. 状态，态，状况141. run v. 运行，运转，操作142. argument n. 变元，自变量143. statement n. 语句，陈述，命题144. shift v. 转义，换档，移位145. store n. & vt. 存储，存储器146. scroll vt. 上滚(卷)；n. 纸卷147. replace vt. 替换，置换，代换148. macro n. 宏，宏功能，宏指令149. page n. 页面，页，版面150. quit v. 退出，结束151. define vt. 定义，规定，分辨152. reference n. & a. 参考；参考的153. other a. 别的，另外的154. while conj. 当…的时候155. pressing n. & a. 压制；紧急的156. restore vt. 恢复，复原157. top n. 顶，尖端158. how ad. 如何，怎样，多么159. color n. 颜色，色彩，(彩)色160. allow v. 允许，容许161. block n. (字，信息，数据)块162. decimal n. & a. 十进制；十进制的163. main a. 主要的164. definition n. 定义，确实，清晰度165. between prep. 在…之间，中间166. optional a. 任选的，可选的167. date n. 日期168. remove v. 除去，移动169. arrow n. 箭头，指针170. label n. 标签，标号，标识符171. within prep. 在…以内172. issue v. 发行，出版，流出173. different a. 不同的，各种各样的174. available a. 可用的175. returned a. 退回的176. associate v. 相联，联想，关联177. attribute n. 属性，标志，表征178. dos 磁盘操作系统179. before prep. 以前，前，先180. order n. & vt. 指令，次序；排序181. modify vt. 修改，改变，变址182. array n. 数组，阵列183. mouse n. 鼠标器184. note n. 注解，注释185. locate vt. 定位186. video n. 视频，电视187. printer n. 打印机，印刷机188. bar n. 条，杆，棒189. bottom n. & a. 底，基础；底下的190. carriage n. 滑架，托架191. content n. 含量，容量，内容192. either a. & pron. 任何一个，各193. ok ad. & a. 对，好；全对194. space n. 空格键，空间195. editor n. 编辑程序196. exist vi. 存在，生存，有197. scope n. 范围，显示器198. paragraph n. 段(落)，节，短讯199. multi (词头)多200. clear v. 清除，弄干净201. exit n. & vi. 出口；退出202. report vt. & n. 报告，报表203. execution n. 执行204. backup n. 备份，后备，后援205. version n. 版本206. find v. 寻找，发现207. pointer n. 指针，指示字208. subset n. 子集，子设备209. keyboard n. 键盘210. full a. & ad. & n. 全(的)，满211. check v. 校对，栓查，核算212. should v. & aux. 应当，该213. single a. & n. 单个的；一个，单214. positioning n. 定位215. provide v. 提供216. title n. 题目，标题217. expression n. 表达式218. through prep. & ad. 通过，直通219. toggle n. & v. 触发器；系紧220. code n. 码，代码，编码221. such a. & pron. 这样的，如此222. beginning n. 起点，初223. guide n. 向导，指南，入门224. tree n. 树，语法树225. environment n. 环境226. but 但是，可是，除非，不过227. device n. 设备，器件，装置228. highlight n. 增强亮度，提示区229. call v. 调用，访问，呼叫230. continue v. 连续，继续231. indicate vt. 指示，表示232. until prep. 到…为止，直到233. begin v. 开始，着手，开端234. place vt. 放，位，地点235. rename vt. 更名，改名236. swap v. 交换，调动237. work n. 工作238. remain vi. 剩下，留下，仍然239. close v. & a. 关闭，闭合；紧密的240. combination n. 结合，组合241. profile n. 简要，剖面，概貌242. unless conj. 除非243. so pron. & conj. 如此，这样244. except prep. 除…之外，除非245. turn v. & n. 转，转动；圈，匝246. back n. 背面，反向，底座247. sure a. & ad. 确实的；的确248. section n. 节，段，区域249. follow v. 跟随，跟踪250. split v. 分开，分离251. need v. 必须，需要252. access n. 存取，选取，接近253. additional a. 附加的，辅助的254. cancel v. 删除，取消，作废255. document n. 文献，资料，文件256. case n. 情况，场合257. numeric n. & a. 数字的，分数258. go vi. 运行，达到259. load n. & v. 装入，负载，寄存260. try n. (尝)试，试验261. size n. 尺寸，大小，容量262. entire a. & n. 完全的；总体263. leave v. 离开，留下264. history n. 历史265. second n. & a. 秒，第二(的) 266. reflow v. & n. 回流，逆流267. output n. 输出，输出设备268. out n. & a. 输入，在外269. both a. & ad. 两，双，都270. install vt. 安装271. source n. 源，电源，源点272. way n. 路线，途径，状态273. assign vt. 赋值，指定，分派274. support vt. 支援，支持，配套275. specific a. 特殊的，具体的276. join v. & n. 连接，并(运算) 277. expand v. 扩充，扩展，展开278. like a. 类似的，同样的279. diskette n. 软磁盘，软盘片280. skip v. 跳跃(定位)，跳过281. application n. 应用282. confirmation n. 认可283. whether conj. 无论，不管284. hold v. 保持285. click n. “卡搭”声，插销286. write v. 写，存入287. byte n. (二进制的)字节288. abbreviate vt. 缩写，省略289. show v. 显示，呈现，出示290. otherwise ad. & a. 另外291. working n. 工作，操作，作业292. delimiter n. 定界符，分界符293. location n. 定位，(存储器)单元294. perform v. 执行，完成295. graphic n. & a. 图形；图形的296. read v. 读，读阅297. confirm vt. 证实，确认298. sort v. 分类，排序299. clause n. 条款，项目，子句300. once ad. & n. 只一次，一旦301. however conj. 然而，可是302. extend v. 扩充303. look v. 看，查看304. starting a. 起始的305. now ad. & n. 此刻，现在306. original n. & a. 原文；原(初)始的307. correspond vi. 通信(联系)308. property n. 性(质)，特征309. several a. & n. 若干个，几个310. learn v. 学习，训练311. cause n. 原因，理由312. bracket n. (方)括号，等级313. omit vt. 省略，删去，遗漏314. running a. 运行着的，游动的315. sub-directory n. 子目录316. edge n. 棱，边，边缘，界限317. form n. 格式，表格，方式318. instruction n. 指令，指导319. ascii n. 美国信息交换标准码320. below a. & prep. 下列的；低于321. standard n. 标准322. occurrence n. 出现，发生323. lock n. & v. 锁，封闭；自动跟踪324. append vt. 附加，增补325. destination n. 目的地，接收站326. password n. 口令，保密字327. point n. 点，小数点，句号328. variety n. 变化，种类，品种329. many a. & n. 许多，多数330. buffer n. 缓冲器331. useful a. 有用的332. object n. 对象，目标，物体333. again ad. 再，又，重新，也334. operating a. 操作的，控制的335. carry v. 进位，带336. update v. 更新，修改，校正337. moving n. & a. 活动的，自动的338. coprocessor n. 协同处理器339. overlay v. 覆盖，重叠340. practice n. 实习，实践341. navigation n. 导航342. automatically ad. 自动地，机械地343. total n. & v. 总数；总计344. previous a. 早先的，上述的345. software n. 软件346. shortcut n. 近路，捷径347. long a. 长的，远的348. unique a. 唯一的，独特的349. part n. 部分，零件350. updated a. 适时的，更新的351. internal a. 内部的352. fill v. 填充353. basic n. & a. 基本；基本的354. math n. 数学355. since prep. 自从…以来356. determine v. 确定357. making n. 制造，构造358. center n. 中心，中央359. already ad. 已经，早已360. keyword n. 关键字(词)361. action n. 操作，运算362. condition n. 条件，情况；vt. 调节363. quick a. & ad. 快速的，灵敏的364. assigned a. 指定的，赋值的365. give vt. 给出，赋予，发生366. large a. (巨)大的，大量的367. chapter n. 章，段368. computer n. 计算机369. complete v. & a. 完成；完整的370. past a. 过去的，结束的371. match v. 比较，匹配，符合372. recover v. 恢复，回收373. always ad. 总是，一直，始终374. require v. 需要，要求375. opening n. 打开，断路，孔376. network n. & vt. 网络；联网377. sign n. 符号，信号，记号378. release vt. & n. 释放，核发，版379. three a. & n. 三(的)380. recall vt. 撤消，复活，检索381. deletion n. 删去(部分)，删除382. fixed a. 固定的，不变的383. amount vt. & n. 总计；合计384. alias n. 别名，代号，标记385. quote n. & v. 引号；加引号386. correct a. & vt. 正确的；改正387. else ad. & conj. 否则，此外388. maximum n. & a. 最大(的)，最高389. under prep. 在…下面(之下) 390. take v. 取，拿391. switching n. 开关，转接，交换392. element n. 元件，元素，码元393. modification n. 改变，修改394. modified a. 修改的，变更的395. input n. 输入，输入设备396. uppercase n. 大写字母397. plus prep. 加，加上，外加398. found v. 建立，创办399. debug vt. 调试400. force v. & n. 强制；压力，强度401. lowercase n. 下档，小写体402. just ad. 恰好403. undo vt. 取消，废除404. environ vt. 围绕，包围405. why ad. 为什么406. temporary a. 暂时的，临时的407. put v. 存放(记录)，放置408. instead ad. (来)代替，当作409. encounter v. & n. 遇到，碰到410. across prep. 交*，越过411. matching n. 匹配，调整412. wildcard n. 通配符413. spill v. 漏出，溢出，漏失414. level n. 水平，级，层次415. browse v. 浏览416. speech n. 说话，言语，语音417. occur vi. 发生，出现，存在418. memo n. 备忘录419. prior a. 先验的，优先的420. loaded a. 有负载的421. length n. (字，记录，块)长度422. round v. 舍入，四舍五入423. variant n. & a. 变体，易变的424. floppy n. 软磁盘425. machine n. 机器，计算机426. square n. & a. 正方形；方形的427. supply vt. & n. 电源，供给428. home n. & a. 家，出发点429. normal a. & n. 正常，标准430. onto prep. 向…，到…上431. during prep. 在…期间432. module n. 模块(程序设计)433. monochrome n. 单色434. assistance n. 辅助设备，帮助435. tell n. 讲，说，教，计算436. library n. (程序…)库，图书馆437. demonstration n. (公开)表演，示范438. stack n. 栈，堆栈，存储栈439. even a. & ad. 偶数的；甚至440. evaluate v. 估计，估算，求值441. times n. 次数442. previously ad. 以前，预先443. directly ad. 直接地，立即444. logical a. 逻辑的，逻辑“或”445. template n. 标准框，样板，模板446. calling n. 呼叫，调用，调入447. later a. 更后的，后面的448. driver n. 驱动器，驱动程序449. therefore ad. & conj. 因此，所以450. saving a. 保存的451. detail n. 元件，零件，细节452. linker n. 连接程序453. loop n. 圈，环；(程序)循环，回路454. process vt. 处理，进程，加工455. scheme n. 方案，计划，图456. every a. 每个，全体，所有的457. refer v. 访问，引用，涉及458. possible a. 可能的，潜在的459. above a. 在…之上，大于460. overview n. 综述，概要461. result n. 结果462. syntax n. 语法，文法，句法463. abbreviation n. 缩短，省略，简称464. bios n. 基本输入/输出系统465. hidden a. 隐藏的，秘密的466. null n. & a. 空(的)，零(的)467. send v. 发送468. private a. 专用的，私人的469. hard a. 硬的470. hardware n. 硬件471. say v. 说，显示，假定472. equal vt. & n. 等于，相等；等号473. pack n. 压缩，包裹474. minus a. & n. 负的；负数，减475. alternate a. 交替的，备用的476. collapse v. 崩溃，破裂477. corner n. 角，角落，转换478. present a. & v. 现行的；提供479. interpreter n. 解释程序，翻译机480. advance v. & n. 进步，提高；进展481. forward a. 正向的482. fast a. & ad. 快速的483. special a. 专用的，特殊的484. slash n. 斜线485. utility n. & a. 实用程序；实用性486. regardless a. 不注意的，不考虑的487. disable vt. 禁止，停用488. compatible a. 可兼容的，可共存的489. depend vi. 随…而定，取决于490. empty a. 空，零，未占用491. alphabetical a. 字母(表)的，abc的492. branch n. 分支，支线；v. 转换493. resume v. 重(新)开(始)494. multiple a. 多次的，复杂的495. monitor n. 监视器，监督程序496. configuration n. 配置497. replacement n. 替换，置换，更新498. required a. 需要的499. macros n. 宏命令(指令)500. table n. 表。