A Unified Tagging Approach to Text Normalization

合集下载

吉林省四校联考2024-2025学年高二上学期9月月考 英语试题

吉林省四校联考2024-2025学年高二上学期9月月考 英语试题

2024~2025(上)高二年级第一次月考英语全卷满分150分,考试时间120分钟。

注意事项:1.答题前,先将自己的姓名、准考证号填写在试卷和答题卡上,并将条形码粘贴在答题卡上的指定位置。

2.请按题号顺序在答题卡上各题目的答题区域内作答,写在试卷、草稿纸和答题卡上的非答题区域均无效。

3.选择题用2B 铅笔在答题卡上把所选答案的标号涂黑;非选择题用黑色签字笔在答题卡上作答;字体工整,笔迹清楚。

4.考试结束后,请将试卷和答题卡一并上交。

5.本卷主要考查内容:选择性必修第一册U2~U3。

第一部分听力(共两节,满分30分)第一节(共5小题;每小题1.5分,满分7.5分)听下面5段对话。

每段对话后有一个小题,从题中所给的A、B、C三个选项中选出最佳选项。

听完每段对话后,你都有10秒钟的时间来回答有关小题和阅读下一小题。

每段对话仅读一遍。

例:How much is the shirt?A.f19.15.B.f9.18.C.f9.15.答案是C。

1.Where does the conversation probably take place?A.In a hotel.B.At the airport.C.At a travel agency.2 .What time is it now?A.5:45.B.5:30.C.5:15.3.What does the woman mean?A.The man found his notes at last.B.The man always gets excellent grades.C.The4.How did better than she had expected. 4.How did the woman come to school today?A.ByB.OnC.By B.On foot. C.By bus.5.What topic does the man suggest for the report?A,Air pollutionB.Garbage B.Garbage (垃圾)sorting.C.Endangered species.第二节(共15小题;每小题1.5分,满分22.5分)听下面5段对话或独白。

阿尔卡特-朗讯OmniSwitch 2220 WebSmart千兆以太网交换机说明书

阿尔卡特-朗讯OmniSwitch 2220 WebSmart千兆以太网交换机说明书

Alcatel-LucentOmniSwitch 2220WebSmart Gigabit Ethernet LAN SwitchesThe Alcatel-Lucent OmniSwitch® 2220 Gigabit WebSmart family of switches provides a simple, secure, and smart business network at affordable prices.The OmniSwitch 2220 allows you to achieve a reliable business class networkperformance including security without paying for advanced network management features. These switches are a lower-priced alternative to managed switches for wired connectivity while maintaining performance, quality of service (QoS) and scalability using a simplified web management interface.The OmniSwitch 2220 family is embedded with the latest technology innovations, and offers maximum investment protection.Deployments benefiting from the OmniSwitch 2220 family are:• High speed desktop connectivity • Secure wireless connectivity• Unified communication connectivity (IP telephony, video and converged solutions)OS2220-8/-P8Features• 8-, 24- and 48-port, Power over Ethernet (PoE) and non-PoE models with fixed small form factor pluggable (SFP) 1G uplink interfaces• Optimized for energy efficiency• Simplified web-based management• Easy ACLsManagement• Web-based GUI (HTTP)• SNMP v1/v2/V3• RMON• Cable test diagnosticsSecurity• 802.1X RADIUS• MAC filtering/port security• Guest VLAN• Broadcast storm recoveryConvergence• Enhanced voice over IP (VoIP) VLAN• Auto-Vo IP• IEEE 802.3af/.at PoE for IP phones, WLAN access points and video camerasBenefitsSimple and easy configuration and managementThe OmniSwitch 2220s are designed to be easy to deploy and used by small businesses or partners. Performance and reliability OmniSwitch 2220 switches have been tested to deliver the high availability and performance you would expect from an Alcatel-Lucent Enterprise switch.Network securityOmniSwitch 2220 switches provide enhanced security and network management features such as 802.1x.IP telephony supportOmniSwitch 2220 switches include QoS features to prioritize delay-sensitive services such as voice and video while simplifying unified communications deployments.OmniSwitch 2220 8-, 24- and 48-port modelsTable 1. Available OmniSwitch 2220 models8/24/48 port modelsOS12220-882Internal AC N/AOS2220-24244Internal AC N/AOS2220-48484Internal AC N/AOS2220-P882Internal AC N/AOS2220-P24244Internal AC N/AOS2220-P48484Internal AC N/ADetailed product featuresSimplified management Configuration management interfaces• Web based GUI (HTTP)• Friendly port naming• Web management applet Monitoring and troubleshooting• Broadcast storm recovery• Event and error logging Facility • Port-based mirroring fortroubleshooting • Ping utility• SNMP v1/v2/V3 and associatedMIBs• RMON groups 1,2,3,9• Cable test diagnosticsAdvanced securityAccess control• RADIUS Client• 802.1X RADIUS usage guidelines• Guest VLANQoS• Co S WRR/WRED• Mapping of CoS to queue• Auto Vo IPLayer-2 and MulticastLayer-2 switching• Up to 16k MACs• Up to 64 VLANs• Latency: < 4 μs• Max Frame: 9216 bytes (jumbo)Multicast• IGMPv1/v2/v3 snooping tooptimize multicast trafficTechnical specifications8/24/48 port modelsRJ-4510/100/1000ports88242448481G SFP uplinkports222222RJ45/SFPcombo ports000022PoE ports08024048 802.3af/at ports 08024048 Switchingcapacity20 Gb/s20 Gb/s56 Gb/s56 Gb/s104 Gb/s104 Gb/sWidth33.0 cm(13.00 in)33.0 cm(13.00 in)44.0 cm(17.32 in)44.0 cm(17.32 in)44.0 cm(17.32 in)44.0 cm(17.32 in)Height 4.4 cm(1.73 in)4.4 cm(1.73 in)4.4 cm(1.73 in)4.4 cm(1.73 in)4.4 cm(1.73 in)4.4 cm(1.73 in)Depth17.5 cm( 6.9 in)17.5 cm( 6.9 in)25.7 cm(10.12 in)25.7 cm(10.12 in)34.5 cm(13.58)34.5 cm(13.58)Operating temperature 0° C to +50° C32° F to +122° F0° C to +50° C32° F to +122° F0° C to +50° C32° F to +122° F0° C to +50° C32° F to +122° F0° C to +50° C32° F to +122° F0° C to +50° C32° F to +122° FStorage temperature -20° C to +70° C-4° F to +158° F-20° C to +70° C-4° F to +158° F-20° C to +70° C-4° F to +158° F-20° C to +70° C-4° F to +158° F-20° C to +70° C-4° F to +158° F-20° C to +70° C-4° F to +158° FHumidity(operating andstorage)5% - 95%5% - 95%5% - 95%5% - 95%5% - 95%5% - 95%Fan (variablespeed)*Fan-less Fan-less Fan-less 2 fans 3 fans 3 fans Acoustic (dB)0 db (A)<0db (A)<40db (A)<40db (A)<40db (A)<40db (A) MTBF (hours)1,209,148 1,104,816987,125656,251487,878425,531 PoE powerbudget (watts)N/A75N/A192N/A390* 48 port models are future development Alcatel-Lucent and the Alcatel-Lucent Enterprise logo are trademarks of Alcatel-Lucent. To view other trademarks used by affiliated companies of ALE Holding, visit: /en/legal/trademarks-copyright . All other trademarks are the property of their respective owners. The IndicatorsSystem LEDs• Po wer LED • Reset butto nPer-port LEDs• 10/100/1000: PoE, link/activity • SFP: Link/activityCompliance and certificationsCommercial• EMI/EMC • CE, UL, cUL, CB• IEC 60825-1 Laser, IEC 60825-2 Laser • CDRH LaserSupported standardsIEEE standards• IEEE 802.1AB — LLDP • IEEE 802.1D — Spanning Tree • IEEE 802.1p — Ethernet Prioritywith User Mapping • IEEE 802.1Q — Virtual LANs w/ Port-based VLANs • IEEE 802.1S — Multiple Spanning Tree • IEEE 802.1W — Rapid Spanning Tree • IEEE 802.1X — Port Based Authentication • IEEE 802.3ac — VLAN Tagging • IEEE 802.3ad — Link Aggregation (w/LACP) • IEEE 802.3x — Flow ControlIETF RFCs• RFC 1534 — Interop. between BootP and DHCP • RFC 2030 — Simple Network Time Protocol (SNTP) V4 • RFC 2131 — DHCP Client • RFC 2865 — RADIUS Client • RFC 3580 — 802.1X RADIUS Usage Guidelines • RFC 951 — BootPOrdering informationOS2220-8Gigabit Ethernet WebSmart chassis in a 1U form factor with 8 x 10/100/1000 Base-T ports, 2 fixed SFP (1G)ports with internal power supply.OS2220-P8Gigabit Ethernet WebSmart chassis in a 1U form factor with 8 x 10/100/1000 PoE Base-T ports, 2 fixed SFP (1G) ports with internal power supply (75 W power budget).OS2220-24Gigabit Ethernet WebSmart chassis in a 1U form factor with 24 x 10/100/1000 Base-T ports, 2 fixed SFP (1G) with internal power supply.OS2220-P24Gigabit Ethernet WebSmart chassis in a 1U form factor with 24 x 10/100/1000 PoE Base-T ports, 2 fixed SFP (1G) with internal power supply (192 W PoE budget).OS2220-48*Gigabit Ethernet WebSmart chassis in a 1U form factor with 48 x 10/100/1000 Base-T ports, 2 RJ45/SFP combo and 2 fixed SFP ports (1G) with internal power supply.OS2220-P48*Gigabit Ethernet WebSmart chassis in a 1U form factor with 48 x 10/100/1000 PoE Base-T ports, 2 RJ45/SFP combo and 2 fixed SFP ports (1G) with internal power supply (390 W PoE budget).* 48 port models are future developmentAll models above include an AC power supply with a country-specific power cord, user manuals access card and hardware for mounting in a 19” rack for 24 and 48 port models.。

2024届山东省实验中学高三下学期一模英语试题

2024届山东省实验中学高三下学期一模英语试题

绝密★启用并使用完毕前山东省实验中学2024届高三第一次模拟考试英语试题2024.04(本试卷共10页, 共三部分: 全卷满分120分, 考试用时100分钟)注意事项:1. 答卷前, 先将自己的姓名、准考证号填写在试卷和答题纸上。

2. 选择题的作答: 每小题选出答案后, 用2B铅笔把答题卡上对应题目的答案标号涂黑。

如需改动, 用橡皮擦干净后, 再选涂其他答案标号。

3. 非选择题的作答: 用0.5mm黑色签字笔直接答在答题卡上对应的答题区域内, 写在试卷、草稿纸和答题卡上的非答题区域均无效。

第一部分阅读理解(共两节, 满分50分)第一节(共15小题; 每小题2.5分, 满分37.5分)阅读下列短文, 从每题所给的A、B、C、D四个选项中选出最佳选项。

AIntroduction to Drama ExamsOur exams inspire and enable learners across the globe to be confident communicators. Exams are open to anyone looking to gain confidence and experience in speech, communication and performance. There are no age restrictions. As one of the UK's oldest and most respected drama schools and awarding organizations, we examine over 100,000candidates and deliver exams both online and in person in many countries across the globe.Now we are pleased to offer free, online "Introduction to Examinations" information session. Booking is now opening for events until Summer 2024.The 1.5-hour session will begin with an Introduction to Examinations, their history and the format of assessment. Work will then focus on the subjects available to take, and will end with a Q&A phase where participants will be invited to write in their questions to the host organizer.Ifyouhaveanyquestionsregardingthis,********************************.ukandwewillbehappytohelp. Looking forward to seeing you online at this event.1. What is an advantage of the drama exam?A. It is free of charge.B. It offers flexible schedules.C. It suits a wide range of people.D. It puts restrictions on nationality.2. What is required to register for the sessions?A. Payment in advance.B. Contact information.C. Education background.D. Performance experience.3. What should you do if you have a question during the online session?A. Email it to the drama school.B. Write it down before the session.C. Propose it at the beginning of the session.D. Send it to the host organizer in Q&A phase.BCafeterias have been filled with challenges—right from planning, purchasing, and preparing, to reducing waste, staying on budget, managing goods, and training staff. Through the tedious process, restaurateurs lacked a unified platform for efficient management. To bring consistency to the unorganised catering(餐饮)industry, childhood friends Arjun Subramanian and Raj Jain, who shared a passion for innovation, decided to partner in 2019 to explore opportunities in the cafeteria industry.In May 2020, they co-founded Platos, a one-stop solution for restaurants with a custom technology kit to streamline all aspects of cafeteria management. The company offers end-to-end cafeteria management, staff selection and food trials to ensure smooth operations and consistent service. "We believe startups solve real problems and Platos is our shot at making daily workplace food enjoyable again. We aim to simplify the dining experience, providing a convenient and efficient solution that benefits both restaurateurs and customers and creating a connected ecosystem, "says Subramanian, CEO and co-founder.Platos guarantees that a technology-driven cafeteria allows customers to order, pay, pick up, and provide ratings and feedback. It also offers goods and menu management to effectively perform daily operations. Additionally, its applications connect all shareholders for a smart cafeteria experience. "We help businesses that are into catering on condition that they have access to an industrial kitchen setup where they' re making food according to certain standards," Jain states.Since the beginning, Platos claims to have transformed 45 cafeterias across eight cities in the country. Currently, it has over 45,000 monthly users placing more than 200,000 orders. Despite facing challenges in launching cafeterias across major cities in the initial stages, Platos has experienced a 15% increase in its month-over-month profits.As for future plans, the startup is looking to raise $1 million from investors as strategic partners, bringing in capital, expertise, and networks. "Finding the right lead investor is the compass that points your startup toward success," Subramanian says.4. What does the underlined word "tedious" in Paragraph 1 mean?A. Time-consuming.B. Breath-taking.C. Heart-breaking.D. Energy-saving.5. What is the purpose of founding Platos?A. To connect customers with a greener ecosystem.B. To ensure food security and variety in cafeterias.C. To improve cafeteria management with technology.D. To make staff selection more efficient and enjoyable.6. What can we learn from the statistics in Paragraph 4?A. Platos has achieved its ultimate financial goal.B. Platos has gained impressive marketing progress.C. Challenges in food industry can be easily overcome.D. Tech-driven cafeterias have covered most urban areas.7. What is Subramanian's future plan for Platos?A. To reduce costs.B. To increase profits.C. To seek investment.D. To innovate technology.CWith a brain the size of a pinhead, insects possess a great sense of direction. They manage to locate themselves and move through small openings. How do they do this with their limited brain power? Understanding the inner workings of an insect's brain can help us in our search towards energy-efficient computing, physicist Elisabetta Chicca of the University of Groningen shows with her most recent result: a robot that acts like an insect.It's not easy to make use of the images that come in through your eyes when deciding what your feet or wings should do. A key aspect here is the apparent motion of things as you move. "Like when you're on a train,” Chicca explains. "The trees nearby appear to move faster than the houses far away." Insects use this information to infer how far away things are. This works well when moving in a straight line, but reality is not that simple. To keep things manageable for their limited brain power, they adjust their behaviour: they fly in a straight line, make a turn, then make another straight line.In search of the neural mechanism(神经机制)that drives insect behaviour, PhD student Thorben Schoepe developed a model of its neuronal activity and a small robot that uses this model to find the position. His model is based on one main principle: always head towards the area with the least apparent motion. He had his robot drive through a long passage consisting of two walls and the robot centred in the middle of the passage, as insects tend to do. In other virtual environments, such as a space with small openings, his model also showed similar behaviour to insects.The fact that a robot can find its position in a realistic environment is not new. Rather, the model gives insight into how insects do the job, and how they manage to do things so efficiently. In a similar way, you could make computers more efficient.In the future, Chicca hopes to apply this specific insect behaviour to a chip as well. "Instead of using a general-purpose computer with all its possibilities, you can build specific hardware; a tiny chip that does the job, keeping things much smaller and energy-efficient." She comments.8. Why is "a train" mentioned in Paragraph 2?A. To illustrate the principle of train motion.B. To highlight why human vision is limited.C. To explain how insects perceive distances.D. To compare the movement of trees and houses.9. What does Paragraph 3 mainly talk about concerning Schoepe's model?A. Its novel design.B. Its theoretical basis.C. Its possible application.D. Its working mechanism.10. What do the researchers think of the finding?A. Amusing.B. Discouraging.C. Promising.D. Contradictory.11. What will Chicca's follow-up study focus on?A. Inventing insect-like chips.B. Studying general-purpose robots.C. Creating insect-inspired computers.D. Developing energy-efficient hardware.DWith the help from an artificial language(AL)model, MIT neuroscientists have discovered what kind of sentences are most likely to fire up the brain's key language processing centers. The new study reveals that sentences that are more complex, because of either unusual grammar or unexpected meaning, generate stronger responses in these language processing centers. Sentences that are very straightforward barely engage these regions, and meaningless orders of words don't do much for them either.In this study, the researchers focused on language-processing regions found in the left hemisphere(半球)of the brain. By collecting a set of 1,000 sentences from various sources, the researchers measured the brain activity of participants using functional magnetic resonance imaging(fMRI)while they read the sentences. The same sentences were also fed into a large language model, similar to ChatGPT, to measure the model's activation patterns. Once the researchers had all of those data, they trained the model to predict how the human language network would respond to any new sentence based on how the artificial language network responded to these 1,000 sentences.The researchers then used the model to determine 500 new sentences that would drive highest brain activity and sentences that would make the brain less active, and their findings were confirmed in subsequent human participants. To understand why certain sentences generate stronger brain responses, the model examined the sentences based on 11 different language characteristics. The analysis revealed that sentences that were more surprising resulted in greater brain activity. Another linguistic(语言的)aspect that correlated with the brain's language network responses was the complexity of the sentences, which was determined by how well they followed English grammar rules and bow logically they linked with each other.The researchers now plan to see if they can extend these findings in speakers of languages other than English. They also hope to explore what type of stimuli may activate language processing regions in the brain's right hemisphere.12. What sentences make our brain work harder?A. Lengthy.B. Logical.C. Straightforward.D. Complicated.13. What is the function of the AL model in the research?A. To examine language network.B. To reduce language complexity.C. To locate language processing area.D. To identify language characteristics.14. How did the researchers carry out their study?A. By conducting interviews.B. By collecting questionnaires.C. By analyzing experiment data.D. By reviewing previous studies.15. Which of the following is a suitable title for the text?A. AL Model Stimulates Brain ActivitiesB. AL Model Speeds Up Language LearningC. AL Model Reveals the Secrets of Brain ActivationD. AL Model Enhances Brain Processing Capacity第二节(共5小题; 每小题2.5分, 满分12.5分)根据短文内容, 从短文后的选项中选出能填入空白处的最佳选项。

Azure Stack HCI 超融合解决方案概览说明书

Azure Stack HCI 超融合解决方案概览说明书

Azure Stack HCI 超融合解决方案概览GPS CSA Tianhao CaoAzure HybridInnovation anywhere with AzureSingle control plane with Azure ArcExtend to the edge with Azure IoTBring Azure services to any infrastructure Modernize datacenters with Azure Stack了解更多:Azure Stack Edge云托管设备Azure Stack HCI超融合基础架构Azure Stack Hub云原生集成系统Azure Stack产品线的拓展Azure Stack HCI超融合混合的Azure 服务交付广泛硬件支持此前…行业标准级x86服务器+高速存储与网络组件SAN / NAS存储连接虚拟机管理器网络设备交换机专门用途的Host操作系统最新的Azure虚拟机管理程序,内置软件定义存储与网络针对虚拟作业优化,组件更少,结构更简单适合远程管理需求,极尽轻简的本地用户界面Azure Stack HCI 软件定义数据中心Hyper-V SDS SDN认证过的硬件( 2节点~16节点)Windows/Linux VMsVM 加密SnapshotClone自动或手动VM 漂移自动Fail OverVM Affinity喜好规则两份或三份数据复本磁盘、服务器容错存储复制(同步或异步)数据去重,压缩和加密In-Memory 缓存百万级IOPS基于VM的IOPS QoSVirtual NetworkVNet Peering负载均衡ACLUDR用户自定义路由VPN Gateway动态IP或静态IPWAC 管理中心基于延伸群集实现本地灾难恢复建立跨房间、跨建筑甚至跨城市的Azure Stack HCI群集同步或异步存储复制、可选加密方案、站点本地弹性拉伸超融合群集北京上海Azure Stack HCI超融合高性能的超融合架构混合的Azure 服务交付广泛硬件支持一键快速配置HCI虚拟机容灾到Azure云内置Azure Stack HCI 云监控基于Azure Stack HCI的Azure Kubernetes Service•使用Azure Stack HCI运行Linux和Windows容器•可使用Kubernetes Dashboard和kubectl等熟悉的工具•集成Arc Enabled K8S 服务•在本地使用Arc Enabled Data Service 和Arc Enabled ML Service•微软负责端到端的安全保护,包括对应OSS组件的安全更新•集成Active Directory(AD)、Azure ADNEW组织在本地使用ARM 群组和标签进行注册和组织计费根据核心/时间通过订阅付费支持通过Azure 门户请求技术支持正在预览Kubernetes 使用托管Kubernetes运行容器化应用策略将Azure 策略配置分配给主机和VM 恢复通过云备份和站点恢复功能保护虚拟机已发布联网安全地实现本地到Azure 虚拟网络的连接Azure Stack HCI 集成Azure混合云服务监控涵盖众多细分指标和警报的全局视图Azure 自助服务分配访问权限,方便其他Azure 用户创建VM开发中即将发布*注册到Azure激活HCIAzure Stack HCI超融合高性能的超融合架构混合的Azure服务交付根据你的需求扩展2节点起步,随时扩展,不影响现有业务Up to 16 servers per cluster Up to 4,000 TB storage capacity per cluster Up to 896 physical cores per cluster Up to 48TB memory per cluster Beyond 1,000servers with cluster sets Automatic VM load balancing Automatic storage rebalancing添加磁盘到添加服务器到集群中登录/HCI查看优选供应商列表并选择适合的硬件设备200+解决方案25+合作伙伴集成系统与认证节点示例What does Azure Stack HCI charge for?Cores Days 80706050403020100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30Physical coresCores from running virtual machines$10/month per physical processor core 1Predictable Doesn’t vary with consumption, more VMs doesn’t cost moreSimple No math with memory, storage, or network ingress/egress Rewards Those who virtualize efficiently, with higher v-to-p density1Price is subject to change before launch (but we think this is right)国内版2.1元/天31天=>65.1元per physical processor core 1How does it work: License guests and apps separately As little or as much Windows Server as you need, like other HCI platformsWhat you want to run:Linux applicationsOpen source softwareWhat you buy:OEM HW+Azure Stack HCI+Nothing else from MicrosoftList price freeWhat you want to run:A few Windows Server rolesor applicationsWhat you buy:Validated OEM HW+Azure Stack HCI+Windows Server 2019 Standard(s)List price 882 USD per 2 guests*What you want to run:Unlimited Windows Serverroles or applicationsWhat you buy:Validated OEM HW+Azure Stack HCI+Windows Server 2019 DatacenterList price 6,155 USD for unlimited guests* Host: SubscriptionGuest: PerpetualEnded微软Microsoft ,其超融合软件与其公有云Azure有着天然的支持。

Emulex VFA5.2 网络适配器家族产品指南说明书

Emulex VFA5.2 网络适配器家族产品指南说明书

Emulex OCe14100B VFA5.2 Adapter FamilyProduct GuideThe Emulex Virtual Fabric Adapter 5 Generation 2 (VFA5.2) Network Adapter Family, which includes the OCe14104B-NX 10 GbE adapter builds on the foundation of the previous generation of Emulex VFA5 adapters by delivering performance enhancements, including RoCE v2 support and management enhancements with improved support for Lenovo XClarity. These new features help reduce complexity, reduce cost, and improve performance with Lenovo servers.The following figure shows the Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ Adapter.Figure 1. Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ AdapterDid you know?The Emulex VFA5.2 adapters support three methods to virtualize I/O, out of the box: Virtual Fabric Mode (vNIC1), Switch Independent mode (vNIC2) and Unified Fabric Port (UFP) mode. With Virtual Fabric, up to eight virtual network ports (vNICs) can be created with a single two-port 10 GbE network adapter.On System x adapters, storage protocols such as iSCSI and FCoE are also supported by Features on Demand upgrades. Using a common hardware infrastructure for Ethernet and SAN and by virtualizing your network adapter, you can reduce your infrastructure capital expense.Click here to check for updatesTable 3. Optical cablesPart number Feature code DescriptionLC-LC OM3 Fiber Optic Cables (requires transceivers)00MN499ASR5Lenovo 0.5m LC-LC OM3 MMF Cable00MN502ASR6Lenovo 1m LC-LC OM3 MMF Cable00MN505ASR7Lenovo 3m LC-LC OM3 MMF Cable00MN508ASR8Lenovo 5m LC-LC OM3 MMF Cable00MN511ASR9Lenovo 10m LC-LC OM3 MMF Cable00MN514ASRA Lenovo 15m LC-LC OM3 MMF Cable00MN517ASRB Lenovo 25m LC-LC OM3 MMF Cable00MN520ASRC Lenovo 30m LC-LC OM3 MMF CableSFP+ 10Gb Active Optical Cables00YL634ATYX Lenovo 1m SFP+ to SFP+ Active Optical Cable 00YL637ATYY Lenovo 3m SFP+ to SFP+ Active Optical Cable 00YL640ATYZ Lenovo 5m SFP+ to SFP+ Active Optical Cable 00YL643ATZ0Lenovo 7m SFP+ to SFP+ Active Optical Cable 00YL646ATZ1Lenovo 15m SFP+ to SFP+ Active Optical Cable 00YL649ATZ2Lenovo 20m SFP+ to SFP+ Active Optical CableThe following table lists the supported direct-attach copper (DAC) cables.Table 4. Copper cablesPart number Feature code DescriptionSFP+ Passive DAC Cables00D6288A3RG0.5m Passive DAC SFP+ Cable90Y9427A1PH1m Passive DAC SFP+ Cable00AY764A51N 1.5m Passive DAC SFP+ Cable00AY765A51P2m Passive DAC SFP+ Cable90Y9430A1PJ3m Passive DAC SFP+ Cable90Y9433A1PK5m Passive DAC SFP+ CableSFP+ Active DAC Cables00VX111AT2R Lenovo 1m Active DAC SFP+ Cables00VX114AT2S Lenovo 3m Active DAC SFP+ Cables00VX117AT2T Lenovo 5m Active DAC SFP+ CablesSFP28 25Gb Passive DAC Cables7Z57A03557AV1W Lenovo 1m Passive 25G SFP28 DAC Cable 7Z57A03558AV1X Lenovo 3m Passive 25G SFP28 DAC Cable 7Z57A03559AV1Y Lenovo 5m Passive 25G SFP28 DAC CableThe following figure shows the Emulex VFA5.2 2x10 GbE SFP+ PCIe Adapter.Figure 2. Emulex VFA5.2 2x10 GbE SFP+ PCIe AdapterFigure 3. ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-Port SFP+ Ethernet AdapterTable 6. Server support - ThinkSystem (Part 2 of 3)Part NumberDescriptionDense V32S Intel V2AMD V1Dense V24S V28SEthernet only; No upgrade to FCoE and iSCSI00AG570Emulex VFA5.2 2x10 GbE SFP+ PCIe AdapterN N N N N N N N N N N N N N N N N N NY7ZT7A00493ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-Port SFP+ Ethernet AdapterN N N N N N N N N N N N N N N N N N NY Table 7. Server support - ThinkSystem (Part 3 of 3)Part NumberDescription 4S V11S Intel V12S Intel V1Dense V1Ethernet only; No upgrade to FCoE and iSCSI00AG570Emulex VFA5.2 2x10 GbE SFP+PCIe AdapterY Y Y N Y Y Y Y Y Y Y Y Y Y N N N N N7ZT7A00493ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-Port SFP+ Ethernet AdapterY Y Y N N N Y Y Y Y Y Y Y Y N N N N NS D 665 V 3 (7D 9P )S D 665-N V 3 (7D A Z )S D 650 V 3 (7D 7M )S D 650-I V 3 (7D 7L )S T 650 V 2 (7Z 75 / 7Z 74)S R 630 V 2 (7Z 70 / 7Z 71)S R 650 V 2 (7Z 72 / 7Z 73)S R 670 V 2 (7Z 22 / 7Z 23)S R 635 (7Y 98 / 7Y 99)S R 655 (7Y 00 / 7Z 01)S R 655 C l i e n t O S S R 645 (7D 2Y / 7D 2X )S R 665 (7D 2W / 7D 2V )S D 630 V 2 (7D 1K )S D 650 V 2 (7D 1M )S D 650-N V 2 (7D 1N )S N 550 V 2 (7Z 69)S R 850 V 2 (7D 31 / 7D 32)S R 860 V 2 (7Z 59 / 7Z 60)S R 950 (7X 11 / 7X 12)S R 850 (7X 18 / 7X 19)S R 850P (7D 2F / 2D 2G )S R 860 (7X 69 / 7X 70)S T 50 (7Y 48 / 7Y 50)S T 250 (7Y 45 / 7Y 46)S R 150 (7Y 54)S R 250 (7Y 52 / 7Y 51)S T 550 (7X 09 / 7X 10)S R 530 (7X 07 / 7X 08)S R 550 (7X 03 / 7X 04)S R 570 (7Y 02 / 7Y 03)S R 590 (7X 98 / 7X 99)S R 630 (7X 01 / 7X 02)S R 650 (7X 05 / 7X 06)S R 670 (7Y 36 / 7Y 37)S D 530 (7X 21)S D 650 (7X 58)S N 550 (7X 16)S N 850 (7X 15)Server support - System xThe Emulex VFA5.2 adapter family is supported in the System x servers that are listed in the following tables.Support for System x and dense servers with Xeon E5/E7 v4 and E3 v5 processorsTable 8. Support for System x and dense servers with Xeon E5/E7 v4 and E3 v5 processorsPartnumber Description00AG570Emulex VFA5.2 2x10 GbE SFP+ PCIe Adapter Y Y Y Y Y Y N 00AG580Emulex VFA5.2 2x10 GbE SFP+ Adapter and FCoE/iSCSI SW N N Y Y Y Y N 00AG560Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ Adapter N N Y Y Y Y N 01CV770Emulex VFA5.2 ML2 2x10 GbE SFP+ Adapter and FCoE/iSCSISWN N Y Y Y Y N7ZT7A00493ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-Port SFP+ Ethernet Adapter N N N N N N N x325M6(3943)x325M6(3633)x355M5(8869)x365M5(8871)x385X6/x395X6(6241,E7v4)nx36M5(5465,E5-26v4)sd35(5493)Support for servers with Intel Xeon v3 processors Table 3. Support for servers with Intel Xeon v3 processorsPart number Description00AG570Emulex VFA5.2 2x10 GbE SFP+ PCIe AdapterN N Y Y Y Y Y 00AG580Emulex VFA5.2 2x10 GbE SFP+ Adapter and FCoE/iSCSI SW N N Y Y Y Y Y 00AG560Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ AdapterN N N Y Y Y Y 01CV770Emulex VFA5.2 ML2 2x10 GbE SFP+ Adapter and FCoE/iSCSI SWN N N N N Y Y 7ZT7A00493ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-Port SFP+Ethernet AdapterNNNNNN NSupport for servers with Intel Xeon v2 processors Table 9. Support for servers with Intel Xeon v2 processorsPart number Description00AG570Emulex VFA5.2 2x10 GbE SFP+ PCIe AdapterN N N N N N Y Y 00AG580Emulex VFA5.2 2x10 GbE SFP+ Adapter and FCoE/iSCSI SWN N N N N N Y Y 00AG560Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ Adapter N N N N N N Y Y 01CV770Emulex VFA5.2 ML2 2x10 GbE SFP+ Adapter and FCoE/iSCSI SWN N N N N N N Y 7ZT7A00493ThinkSystem Emulex OCe14104B-NX PCIe 10Gb 4-PortSFP+ Ethernet AdapterNNNNNNNN Modes of operationx 3100 M 5 (5457)x 3250 M 5 (5458)x 3500 M 5 (5464)x 3550 M 5 (5463)x 3650 M 5 (5462)x 3850 X 6/x 3950 X 6 (6241, E 7 v 3)n x 360 M 5 (5465)x 3300 M 4 (7382)x 3500 M 4 (7383, E 5-2600 v 2)x 3550 M 4 (7914, E 5-2600 v 2)x 3630 M 4 (7158, E 5-2400 v 2)x 3650 M 4 (7915, E 5-2600 v 2)x 3650 M 4 B D (5466)x 3750 M 4 (8753)x 3850 X 6/x 3950 X 6 (6241, E 7 v 2)In pNIC mode,the adapter operates as a standard dual-port 10 Gbps Ethernet adapter, and itfunctions with any 10 GbE switch. In pNIC mode, with the Emulex FCoE/iSCSI License, the card operates in a traditional Converged Network Adapter (CNA) mode with two Ethernet ports and two storage ports (iSCSI or FCoE) available to the operating system.The following table compares the three virtual fabric modes.Table 10. Comparison of virtual fabric modesFunction Virtual Fabric Mode(vNIC1)Switch Independent Mode(vNIC2)UFP ModeDescription Intelligence in the NetworkingOS working with selectEmulex adapters. VLANbased.Intelligence in the adapter,independent of the upstreamnetworking device.Intelligence in the adapter,independent of theupstream networkingdevice.Supported switches G8124E, G8264, G8264T,G8264CSAll 10 GbE switches G8272, G8264CS, andG8264 (NOS 7.9 or later)Number of vNICs per physical 10 Gb port 4 (3 if storage functions areused to provide a vHBA)4 (3 if storage) 4 (3 if storage)MinimumvNICbandwidth100 Mb100 Mb100 MbServer-to-switch bandwidth limit per vNIC Yes No Yes, maximum burstallowed and minimumguaranteeSwitch-to-server bandwidth limit per vNIC Yes No Yes, maximum burstallowed and minimumguaranteeIEEE 802.1q VLAN tagging Optional Required (untagged traffic will betagged with LPVID which isconfigured in UEFI on a per-vNICbasis)Optional for Trunk orTunnel mode; notsupported for accessmode.Isolated NICteamingfailover pervNICYes No Yes (NOS 7.9 or later)Switch modes Tunnel mode Access or Trunk Mode (two vNICwhich are part of the same physicalport cannot carry the same VLAN)Access, Trunk, Tunnel, and FCoE modesUplink connectivity Dedicated Share Dedicated for Tunnelmode; Shared for othermodesiSCSI/FCoEsupportYes Yes Yes Physical specificationsTable 11. Operating system support for Emulex VFA5.2 2x10 GbE SFP+ PCIe Adapter, 00AG570 (Part 1of 2)Operating systemsMicrosoft Windows Server 2008 R2N N N N N N N N N N N N N N N Microsoft Windows Server 2012N N N N N N N N N N N N N N N Microsoft Windows Server 2012 R2N N N N N N N N N N N N N N N Microsoft Windows Server 2016Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server 2019Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server 2022N N N Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server version 1709N N N N N N N N N N N N N N N Microsoft Windows Server version 1803N N N N N N N N N N N N N N N Red Hat Enterprise Linux 6 Server x64 Edition N N N N N N N N N N N N N N N Red Hat Enterprise Linux 6.10N N N N N N N N N N N N N N N Red Hat Enterprise Linux 6.9N N N N N N N N N N N N N N N Red Hat Enterprise Linux 7.3N N N N N N N N N N N N N N N Red Hat Enterprise Linux 7.4N N N N N N N N N N N N N N N Red Hat Enterprise Linux 7.5Y Y Y N N N N N N N N N N N N Red Hat Enterprise Linux 7.6N N N Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.7Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.8Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.9Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.5Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.6Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 9.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 11 SP4N N N N N N N N N N N N N N N SUSE Linux Enterprise Server 11 SP4 with Xen N N N N N N N N N N N N N N N SUSE Linux Enterprise Server 12 SP2N N N N N N N N N N N N N N N SUSE Linux Enterprise Server 12 SP2 with Xen N N N N N N N N N N N N N N N SUSE Linux Enterprise Server 12 SP3Y Y Y N N N N N N N N Y N N N SUSE Linux Enterprise Server 12 SP3 with Xen Y Y Y N N N N N N N N Y N N N SUSE Linux Enterprise Server 12 SP4N N N Y Y Y Y Y Y Y Y Y Y Y YS R 150S R 250S T 250S D 530 (X e o n G e n 2)S R 530 (X e o n G e n 2)S R 550 (X e o n G e n 2)S R 570 (X e o n G e n 2)S R 590 (X e o n G e n 2)S R 630 (X e o n G e n 2)S R 650 (X e o n G e n 2)S R 850 (X e o n G e n 2)S R 850P (X e o n G e n 2)S R 860 (X e o n G e n 2)S R 950 (X e o n G e n 2)S T 550 (X e o n G e n 2)SUSE Linux Enterprise Server 12 SP4 with Xen N N N Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP1 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP2 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP3 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP4 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere 5.1 (ESXi)N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 5.5N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.0 U3N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.5N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.5 U1N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.5 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7Y Y Y N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.7 U1N N N Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y YOperating systemsS R 150S R 250S T 250S D 530 (X e o n G e n 2)S R 530 (X e o n G e n 2)S R 550 (X e o n G e n 2)S R 570 (X e o n G e n 2)S R 590 (X e o n G e n 2)S R 630 (X e o n G e n 2)S R 650 (X e o n G e n 2)S R 850 (X e o n G e n 2)S R 850P (X e o n G e n 2)S R 860 (X e o n G e n 2)S R 950 (X e o n G e n 2)S T 550 (X e o n G e n 2)Table 12. Operating system support for Emulex VFA5.2 2x10 GbE SFP+ PCIe Adapter, 00AG570 (Part 2of 2)Operating systemsMicrosoft Windows Server 2008 R2N N N N N N N N N N N Y N N N Y Y Y Y N Microsoft Windows Server 2012N N N N N N N N N N N N Y Y Y Y Y N Y N Microsoft Windows Server 2012 R2Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y Y Y Microsoft Windows Server 2016Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y YMicrosoft Windows Server 2019Y Y Y Y Y Y Y Y Y Y Y N Y Y N N N Y N Y Microsoft Windows Server 2022Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N Microsoft Windows Server version 1709Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server version 1803Y N N N N Y Y Y Y Y N N N N N N N Y N Y Red Hat Enterprise Linux 6 Server x64EditionN N Y N N N N N N N N N Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 6.10Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 6.9Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.3Y Y Y N N Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.5Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.6Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.7Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.8Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.9Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.0Y Y Y Y Y Y Y Y Y Y Y N Y N N N N N N N Red Hat Enterprise Linux 8.1Y Y Y Y Y Y Y Y Y Y Y N Y N N N N N N N Red Hat Enterprise Linux 8.2Y Y Y Y Y Y Y Y Y Y Y N Y N N N N N N N Red Hat Enterprise Linux 8.3Y Y Y Y Y Y Y Y Y Y Y N Y N N N N N N N Red Hat Enterprise Linux 8.4Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N Red Hat Enterprise Linux 8.5Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N Red Hat Enterprise Linux 8.6Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N Red Hat Enterprise Linux 9.0Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N SUSE Linux Enterprise Server 11 SP4Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y N Y N SUSE Linux Enterprise Server 11 SP4 with XenY Y Y Y Y Y Y Y Y Y Y N Y N Y Y Y N Y N SUSE Linux Enterprise Server 12 SP2Y Y Y N N Y Y Y N Y YYYY Y Y Y Y Y YS D 530 (X e o n G e n 1)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)S T 550 (X e o n G e n 1)x 3850/3950 X 6 (6241, E 7 v 3)x 3850/3950 X 6 (6241, E 7 v 4)x 3250 M 6 (3633)n x 360 M 5 (5465)x 3500 M 5 (5464)x 3550 M 5 (5463)x 3550 M 5 (8869)x 3650 M 5 (5462)x 3650 M 5 (8871)11SUSE Linux Enterprise Server 12 SP2 with XenY Y Y N N Y Y Y N Y Y Y Y N Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP3 with XenY Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP4 with XenY Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5 with XenY Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y SUSE Linux Enterprise Server 15Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y SUSE Linux Enterprise Server 15 SP1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y SUSE Linux Enterprise Server 15 SP1 with XenY Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y SUSE Linux Enterprise Server 15 SP2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y SUSE Linux Enterprise Server 15 SP2 with XenY Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y SUSE Linux Enterprise Server 15 SP3Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N SUSE Linux Enterprise Server 15 SP3 with XenY Y Y Y Y Y Y Y Y Y Y N N N N N N N N N SUSE Linux Enterprise Server 15 SP4Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N SUSE Linux Enterprise Server 15 SP4 with XenY Y Y Y Y Y Y Y Y Y Y N N N N N N N N N SUSE Linux Enterprise Server 15 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N N Y N Y VMware vSphere 5.1 (ESXi)N N N N N N N N N N N N N N Y Y Y N Y N VMware vSphere Hypervisor (ESXi) 5.5N N N N N N N N N N N N N Y Y Y Y N Y N VMware vSphere Hypervisor (ESXi) 6.0 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5Y Y Y N N Y Y Y N Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y N Y N Y VMware vSphere Hypervisor (ESXi) 6.7 U1Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y N Y N Y VMware vSphere Hypervisor (ESXi) 6.7 U2Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y N Y N Y VMware vSphere Hypervisor (ESXi) 6.7 U3Y Y Y Y Y Y Y Y Y Y YNYY Y Y N Y N YOperating systemsS D 530 (X e o n G e n 1)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)S T 550 (X e o n G e n 1)x 3850/3950 X 6 (6241, E 7 v x 3850/3950 X 6 (6241, E 7 v x 3250 M 6 (3633)n x 360 M 5 (5465)x 3500 M 5 (5464)x 3550 M 5 (5463)x 3550 M 5 (8869)x 3650 M 5 (5462)x 3650 M 5 (8871)VMware vSphere Hypervisor (ESXi) 7.0Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N VMware vSphere Hypervisor (ESXi) 7.0 U1Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N VMware vSphere Hypervisor (ESXi) 7.0 U2Y Y Y Y Y Y Y Y Y Y Y N N N N N N N N N VMware vSphere Hypervisor (ESXi) 7.0 U3Y Y Y Y Y Y Y Y Y Y YNNN N N N N N NOperating systems[in box driver support only]Table 13. Operating system support for Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ Adapter, 00AG560(Part 1 of 2)Operating systemsMicrosoft Windows Server 2008 R2N N N N N N N N N N N N N N N N N N N Microsoft Windows Server 2012N N N N N N N N N N N N N N N N N N N Microsoft Windows Server 2012 R2N N N N N N N N N N Y Y Y Y Y Y Y Y Y Microsoft Windows Server 2016Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server 2019Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server 2022Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Microsoft Windows Server version 1709N N N N N N N N N N Y Y N Y Y Y Y Y Y Microsoft Windows Server version 1803N N N N N N N N N N N N N N Y Y Y Y Y Red Hat Enterprise Linux 6 Server x64 Edition N N N N N N N N N N N Y N N N N N N N Red Hat Enterprise Linux 6.10N N N N N N N N N N Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 6.9N N N N N N N N N N Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.3N N N N N N N N N N Y Y N N Y Y Y N Y Red Hat Enterprise Linux 7.4N N N N N N N N N N Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.5N N N N N N N N N N Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.6Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.7Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.8Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.9Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y YS D 530 (X e o n G e n 1)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)S T 550 (X e o n G e n 1)x 3850/3950 X 6 (6241, E 7 v x 3850/3950 X 6 (6241, E 7 v x 3250 M 6 (3633)n x 360 M 5 (5465)x 3500 M 5 (5464)x 3550 M 5 (5463)x 3550 M 5 (8869)x 3650 M 5 (5462)x 3650 M 5 (8871)1S R 530 (X e o n G e n 2)S R 550 (X e o n G e n 2)S R 570 (X e o n G e n 2)S R 590 (X e o n G e n 2)S R 630 (X e o n G e n 2)S R 650 (X e o n G e n 2)S R 850 (X e o n G e n 2)S R 850P (X e o n G e n 2)S R 860 (X e o n G e n 2)S R 950 (X e o n G e n 2)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)Red Hat Enterprise Linux 8.1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.5Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.6Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 8.7Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Red Hat Enterprise Linux 9.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 11 SP4N N N N N N N N N N Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 11 SP4 with Xen N N N N N N N N N N Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP2N N N N N N N N N N Y Y N N Y Y Y N Y SUSE Linux Enterprise Server 12 SP2 with Xen N N N N N N N N N N Y Y N N Y Y Y N Y SUSE Linux Enterprise Server 12 SP3N N N N N N N Y N N Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP3 with Xen N N N N N N N Y N N Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP4 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 12 SP5 with Xen Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP1 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP2 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP3 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP4Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 SP4 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y SUSE Linux Enterprise Server 15 with Xen Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere 5.1 (ESXi)N N N N N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 5.5N N N N N N N N N N N N N N N N N N N VMware vSphere Hypervisor (ESXi) 6.0 U3N N N N N N N N N N Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5N N N N N N N N N N Y Y N N Y Y Y N Y VMware vSphere Hypervisor (ESXi) 6.5 U1N N N N N N N N N N Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.5 U3Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y YOperating systems S R 530 (X e o n G e n 2)S R 550 (X e o n G e n 2)S R 570 (X e o n G e n 2)S R 590 (X e o n G e n 2)S R 630 (X e o n G e n 2)S R 650 (X e o n G e n 2)S R 850 (X e o n G e n 2)S R 850P (X e o n G e n S R 860 (X e o n G e n 2)S R 950 (X e o n G e n 2)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)VMware vSphere Hypervisor (ESXi) 6.7N N N N N N N N N N Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7 U1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7 U2Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 6.7 U3Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U1Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U2Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y VMware vSphere Hypervisor (ESXi) 7.0 U3Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y YOperating systemsTable 14. Operating system support for Emulex VFA5.2 ML2 Dual Port 10GbE SFP+ Adapter, 00AG560(Part 2 of 2)Operating systemsMicrosoft Windows Server 2008 R2Y N N Y Y Y N Microsoft Windows Server 2012N Y Y Y N Y N Microsoft Windows Server 2012 R2N Y Y Y Y Y Y Microsoft Windows Server 2016Y Y Y Y Y Y YMicrosoft Windows Server 2019N Y N N Y N Y Microsoft Windows Server 2022N N N N N N N Microsoft Windows Server version 1709Y Y Y Y Y Y Y Microsoft Windows Server version 1803N N N N Y N Y Red Hat Enterprise Linux 6 Server x64 Edition N Y Y Y Y Y Y Red Hat Enterprise Linux 6.10N Y Y Y Y Y Y Red Hat Enterprise Linux 6.9N Y Y Y Y Y Y Red Hat Enterprise Linux 7.3Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.4Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.5Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.6Y Y Y Y Y Y Y Red Hat Enterprise Linux 7.7YYY Y Y Y YS R 530 (X e o n G e n 2)S R 550 (X e o n G e n 2)S R 570 (X e o n G e n 2)S R 590 (X e o n G e n 2)S R 630 (X e o n G e n 2)S R 650 (X e o n G e n 2)S R 850 (X e o n G e n 2)S R 850P (X e o n G e n S R 860 (X e o n G e n 2)S R 950 (X e o n G e n 2)S R 530 (X e o n G e n 1)S R 550 (X e o n G e n 1)S R 570 (X e o n G e n 1)S R 590 (X e o n G e n 1)S R 630 (X e o n G e n 1)S R 650 (X e o n G e n 1)S R 850 (X e o n G e n 1)S R 860 (X e o n G e n 1)S R 950 (X e o n G e n 1)x 3850/3950 X 6 (6241, E 7 v 3)x 3850/3950 X 6 (6241, E 7 v 4)n x 360 M 5 (5465)x 3550 M 5 (5463)x 3550 M 5 (8869)x 3650 M 5 (5462)x 3650 M 5 (8871)11。

Optimizing Operations in New Retail Management

Optimizing Operations in New Retail Management

Optimizing Operations in New RetailManagementIntroduction:New retail management has emerged as a transformative force in the retail industry, blending traditional brick-and-mortar stores with advanced digital technologies. This innovative approach requires a shift in operations to effectively meet the expectations of tech-savvy customers and maintain a competitive edge. In this response, we will explore key strategies to optimize operations in the new retail management era.1. Integration of Online and Offline Channels:To thrive in new retail management, retailers must seamlessly integrate their online and offline channels. This involves creating a unified customer experience by implementing strategies such as click-and-collect, where customers can order online and pick up in-store, and offering an online inventory that reflects real-time stock availability. By connecting the online and offline worlds, retailers can enhance customer convenience and increase sales.2. Data-Driven Decision Making:Data plays a crucial role in new retail management. Retailers should leverage advanced analytics tools to gain insights into customer behavior, market trends, and inventory management. By analyzing data, retailers can make informed decisions about inventory replenishment, personalized promotions, and pricing strategies. Furthermore, data-driven decision making allows retailers to optimize operations by identifying areas for improvement and maximizing efficiency.3. Implementing Automated Systems:Automation is a key driver in optimizing retail management operations. Retailers should leverage technologies such as RFID (Radio Frequency Identification) tagging andbarcode scanning to automate inventory management and improve accuracy. Automated systems can track inventory levels, reduce manual labor, prevent stock-outs, and streamline supply chain operations. Additionally, automation can be extended to other areas such as checkout processes, enabling retailers to provide a seamless shopping experience.4. Personalization and Customer Engagement:In new retail management, personalized communication plays a vital role in customer engagement. Retailers should leverage customer data to deliver personalized recommendations, offers, and promotions. By understanding customer preferences, retailers can create targeted marketing campaigns that resonate with their target audience and increase customer loyalty. Implementing loyalty programs and offering rewards for customer engagement can also help optimize retail operations by driving repeat business.5. Enhancing Fulfillment Services:Efficient and timely order fulfillment is crucial in the new retail management landscape. Retailers should evaluate their fulfillment processes and invest in technologies such as robotics and artificial intelligence to streamline operations. Implementing automated picking and packing systems can improve order accuracy and speed up the fulfillment process. Additionally, retailers should optimize their delivery services by partnering with logistics providers that offer flexible and fast delivery options, including same-day and next-day delivery.6. Employee Training and Empowerment:Optimizing operations in new retail management also involves investing in employee training and empowerment. Retailers should provide comprehensive training programs to educate employees about new technologies, customer service best practices, and inventory management techniques. Additionally, empowering employees by granting them decision-making authority and creating a positive work environment can improve operational efficiency and enhance customer satisfaction.Conclusion:Embracing new retail management requires retailers to adapt their operations to meet the demands of the digital age. By integrating online and offline channels, leveraging data-driven decision making, implementing automated systems, personalizing customer engagement, enhancing fulfillment services, and investing in employee training, retailers can optimize their operations and stay competitive in this evolving landscape. Successfully optimizing operations in new retail management will not only drive business growth but also create a seamless and personalized shopping experience for customers.。

天津市静海区北京师范大学静海附属学校2024-2025学年高一上学期第一次月考英语试题

天津市静海区北京师范大学静海附属学校2024-2025学年高一上学期第一次月考英语试题

天津市静海区北京师范大学静海附属学校2024-2025学年高一上学期第一次月考英语试题一、听力选择题1.What are the speakers going to do this afternoon?A.Play tennis.B.See a movie.C.Arrange a party. 2.What is most probably the man?A.A postman.B.A policeman.C.A repairman.3.Where does the conversation take place?A.On a bus.B.On a plane.C.On a train.4.How do the Scots feel about moving to Paris?A.Nervous.B.Happy.C.Uncertain.5.What does the man think of his cake?A.It’s not soft.B.It’s not fresh.C.It’s not sweet.听下面一段较长对话,回答以下小题。

6.How long was the man in the shower?A.For about 10 minutes.B.For about 30 minutes.C.For about 60 minutes. 7.What do we know about the man?A.He tried to save water.B.He is out of work now.C.He paid the water bill last month.8.What is the most probable relationship between the speakers?A.Brother and sister.B.Host and guest.C.Teacher and student.听下面一段较长对话,回答以下小题。

德尔网络Z9000产品说明说明书

德尔网络Z9000产品说明说明书

Highly-available, high-performance Active Fabric spineThe Dell Networking Z9000 is a high-performance, efficient switch-router product designed to meet the requirements for high density 10/40GbE aggregation in a data center core network. The Z9000 switch is designed to address the East/West traffic patterns of modern data centers, providing higher performance and bandwidth across the data center for server to server communications. The Z9000 fabric switch can support 32 ports of 40GbE QSFP+ or 128 ports of 10GbE SFP+ realized through breakout cables. Supporting a full suite of Ethernet switching and routing protocols in the hardened Dell Networking OS, the Z9000 fabric switch can enable an Active Fabric™ via Layer 2 or Layer 3 protocols.An Active Fabric design with Z9000 switches can be built outto create scalable, high-performance 10/40GbE data center networks. The resiliency of an Active Fabric is superior to legacy, centralized core architectures, since the failure of a single node within a CLOS network cannot bring down the entire switching fabric. A single switching element can be restarted or replaced in the event of a failure versus an entire chassis reboot required in a centralized design.The Z9000 is supported with Active Fabric Manager (AFM), which helps automate design and deployment of multi-tier fabrics. AFM helps customers manage multiple fabrics from a single console, enabling a unified view of the entire fabric, when combinedwith Dell OMNM and other management solutions. With AFM, over 25 templates can be customized for specific workloadand deployment scenarios, easily delivering active-active L2or L3 designs for 1/10/40G with Z9000 to rack (with top-of-rack switches including Dell S4810/S4820T, S6000) and blade infrastructures (including Dell MXL).Key applications• Containerized data centers and prover-hosted data centers• Enterprise DC core aggregating 10/40GbE, cloud computing, high-performance cores• High-performance SDN/OpenFlow 1.0 enabled with ability to inter-operate with industry standard OpenFlow controllers Key features• 2RU high-density 10/40GbE fabric/core switch with 32 x 40GbE ports expandable to 128 x 10GbE ports using QSFP+ to SFP+ breakout cables• 2.5Tbps (full-duplex) non-blocking, fabric delivers line-rate performance under full load• Virtual link trunking (VLT) and enhanced VLT for layer 2 multipathing• Modular Dell Networking OS software delivers inherent stability as well as advanced monitoring and serviceability functions • Supported with Active Fabric design and Active Fabric Manager to reduce design, configuration and management for active/active deployments• Total aggregated packet buffer memory of 54MB for line-rate processing• 128 link aggregation groups with up to eight members per group, using advanced hashing with random seed values• Reversible front-to-back or back-to-front airflow• Supports jumbo frames for high-end server connectivity• Redundant, hot-swappable power supplies and fans• Low power consumption• Supports OpenFlow 1.0 in hybrid mode• Supports new QSFP+ PSM4, SR and ESR transceiver/cablesDell Networking Z9000Data center core fabric switchHigh-density 32-port 40GbE core router/switch in 2RU form factor; line rate, non-blocking, low-latency and lower power, enabling a greener, faster data center; feature-rich Dell Networking OS.High-performance,efficient fabric switchfor modern data centertraffic.© 2013 Dell, Inc. All rights reserved. Dell and the DELL logo are trademarks of Dell, Inc. All other company names are trademarks of their respective holders.Information in this document is subject to change without notice. Dell Inc. assumes no responsibility for any errors that may appear in this document.Learn more at /NetworkingNovember 2013 | Version 2.1dell-networking-Z9000-spec sheetSpecifications: Z9000 data center core switchProductZ9000, 32 x 40GbE QSFP+, 1 x AC PSU, 4 x Fans, I/O Panel to PSU AirflowZ9000, 32 x 40GbE QSFP+, 1 x AC PSU, 4 x Fans, PSU to I/O Panel AirflowZ9000, 32 x 40GbE QSFP+, 1 x DC PSU, 4 x Fans, I/O Panel to PSU AirflowZ9000, 32 x 40GbE QSFP+, 1 x DC PSU, 4 x Fans, PSU to I/O Panel AirflowRedundant power supplyZ9000, AC Power Supply, I/O Panel to PSU Airflow Z9000, AC Power Supply, PSU to I/O Panel Airflow Z9000, DC Power Supply, I/O Panel to PSU Airflow Z9000, DC Power Supply, PSU to I/O Panel Airflow OpticsTransceiver, QSFP+, 40GbE, SR Optics, 850nm Wavelength, 100–150m Reach on OM3/OM4Transceiver, QSFP+, 40GbE, ESR OpticsTransceiver, QSFP+, 40GbE PSM4 (2km reach), 1m, 5m, 15m Transceiver, QSFP+, 40GbE, LR4, 10Km reach CablesCable, 40GbE QSFP+, Active Fiber Optic, 10m, 50mCable, 40GbE QSFP+, Direct Attach Cable, 0.5m, 1m, 3m, 5m, 7m Cable, 40GbE MTP to 4xLC Optical Breakout Cable, 1m, 3m, 5m, 7m (optics not included)Cable, 40GbE QSFP+ to 4xSFP+, Direct Attach Breakout Cable, 0.5m, 1m, 3m, 5m, 7mCable, 40GbE MTP Fiber over OM3, 1m, 3m, 5m, 7m, 10m, 25m, 50m, (75m and 100m in 2014)Cable Management Kit, Z9000 MTP to LC (1RU 48-port LC)SoftwareDell Networking OS Software, Layer3Note: In-field change of airflow direction not supported.Physical32 line-rate 40 Gigabit Ethernet QSFP+ ports1 RJ45 console/management port with RS232 signaling 1 RJ45 10/100/1000 Base-T management port 1 x USB 2.0 type A storage port 1 x USB 2.0 type B console portSize: 2 RU, 3.48 x 17.32 x 24” (8.8 x 44 x 61 cm) (H x W x D)Weight: 39 lbs (1 power supply, 4 fan trays)Power supply: 100–240V AC 50/60 Hz, -40 to -60V DC Max. thermal output: 2692 BTU/h Max. current draw per system:8A at 100/120V AC, 4A at 200/240V AC 16.5A at -48V DCMax. power consumption: 789W Max. operating specifications:Operating temperature: 0°C to 40°COperating humidity: 10 to 85% (RH), non-condensing Max. non-operating specifications:Storage temperature: –40°F to 158°F (–40°C to 70°C)Storage humidity: 5 to 95% (RH), non-condensing Reliability: MTBF 135,744 hoursRedundancyHot swappable redundant power Hot swappable redundant fansPerformanceMAC addresses:128K IPv4 routes: 16KIPv6 routes:8K (shared cam space with IPv4)Switch fabric capacity: 2.56Tbps (full-duplex)Forwarding capacity 1.9BppsQueues per port: 8 COS queues L2 VLANs: 4096ACLs: 8K ingress, 4k egressLAGs:128 with up to 8 members per LAG LAG load balancing: Based on Layer 2, IPv4 headers Packet buffer memory:54MBIEEE compliance802.1AB LLDP802.1D Bridging, STP 802.1p L2 Prioritization 802.1Q V LAN Tagging, Double VLAN Tagging, GVRP 802.1s MSTP802.3ad Link Aggregation with LACP 802.3ae 10 Gigabit Ethernet (10GBase-X)802.3ba 40 Gigabit Ethernet (40GBase-SR4, 40GBase-LR4) on optical ports802.3uFast Ethernet (100BASE-TX) on manatement ports802.3x Flow Control Force10 PVST+MTU 12,000 bytesRFC and I-D ComplianceGeneral Internet protocols768 UDP 793 TCP 854 Telnet 959 FTP 1321 MD5 1350 TFTP 2474 Differentiated Services 3164 SyslogGeneral IPv4 protocols791 IPv4792 ICMP 826 ARP 1027 Proxy ARP 1035 DNS (client)1042 Ethernet Transmission 1191 Path MTU Discovery 1305 NTPv31519 CIDR 1812 Routers 1858 IP Fragment Filtering 2131 DHCP (relay)2338 VRRP 3021 31-bit Prefixes 3046 DHCP Option 823069 Private VLAN 3128 Tiny Fragment Attack ProtectionRIP1058 RIPv12453RIPv2OSPF2154 MD5 1587 NSSA 2328 OSPFv2 2370 Opaque LSA 2740 OSPFv3 4552 OSPFv3 IPsec authenticationBGP1997 Communities 2385 MD52439 Route Flap Damping 2796 Route Reflection 2842 Capabilities 2918 Route Refresh 3065 Confederations 4360 Extended Communities 4893 4-byte ASN 5396 4-byte ASN Representations 4271 BGPv42545 BGp.4 Multiprotocol Extensions for IPv6 Inter-Domain Routing Draft Graceful Restart Draft BGP Add PathMulticast1112 IGMPv1 2236 IGMPv23376 IGMPv3 3569 SSM for IPv44541 IGMP 4601PIM-SMSnoopingSDN/OpenflowOpenflow standard 1.0 with extensionsNetwork management1155 SMIv1 1156 Internet MIB 1157 SNMPv1General IPv6 protocols2460 IPv6 1858 IP FragmentFiltering 2461 Neighbor Discovery 2675 Jumbograms (partial) 3587 Global Unicast 2462 Stateless Address Address Format Autoconfiguration (partial) 4291 Addressing2463 ICMPv6 1981 IPv6 Path MTU 4861 IPv6 Host for Management DiscoveryPortIS-ISRFC 1195 Routing IPv4 with IS-IS RFC 5308 Routing IPv6 with IS-IS 2461 Neighbor Discovery1212 Concise MIB Definitions 1215 SNMP T raps 1493 Bridges MIB 1850 OSPFv2 MIB 1901 Community-Based SNMPv22011 IP MIB 2012 TCP MIB 2013 UDP MIB 2096 IP Forwarding Table MIB 2570 SNMPv32571 Management Frameworks 2572 Message Processing and Dispatching 2576 Coexistence Between SNMPv1/v2/v32578 SMIv22579 Textual Conventions for SMIv22580 Conformance Statements for SMIv22618 RADIUS Authentication MIB 2665 Ethernet-Like Interfaces MIB 2674 Extended Bridge MIB 2787 VRRP MIB 2819 RMON MIB (groups 1, 2, 3, 9)2863 Interfaces MIB 2865 RADIUS 3273 RMON High Capacity MIB 3416 SNMPv23418 SNMP MIB 3434 RMON High Capacity Alarm MIB 5060 PIM MIB ANSI/TIA-1057 LLDP-MED MIB draft-ietf-idr-bgp4-mib-06 BGP MIBv1IEEE 802.1AB LLDP MIB IEEE 802.1AB LLDP DOT1 MIB IEEE 802.1AB LLDP DOT3 MIB ruzin-mstp-mib-02 MSTP MIB (traps) sFlowv5 MIB (version 1.3)FORCE10-BGP4-V2-MIB F orce10 BGP MIB(draft-ietf-idr-bgp4-mibv2-05)FORCE10-IF-EXTENSION-MIB FORCE10-LINKAGG-MIBFORCE10-COPY-CONFIG-MIB FORCE10-PRODUCTS-MIB FORCE10-SS-CHASSIS-MIB FORCE10-SMIFORCE10-SYSTEM-COMPONENT-MIB FORCE10-TC-MIBFORCE10-TRAP-ALARM-MIBFORCE10-FORWARDINGPLANE-STATS-MIBRegulatory complianceSafetyUL/CSA 60950-1, Second EditionEN 60950-1, Second EditionIEC 60950-1, Second Edition Including all National Deviations and Group DifferencesEN 60825-1 Safety of Laser Products Part 1: Equipment Classification Requirements and User’s GuideEN 60825-2 Safety of Laser Products Part 2: Safety of Optical Fibre Communication Systems FDA Regulation 21 CFR 1040.10 and 1040.11EmissionsAustralia/New Zealand: AS/NZS CISPR 22: 2008, Class A Canada: ICES-003:2004, Class AEurope: EN 55022: 2006+A1:2007 (CISPR 22: 2008), Class A Japan: VCCI V-3/2010.04 Class AUSA: FCC CFR 47 Part 15, Subpart B:2011, Class AImmunityEN 300 386 V1.4.1:2008 EMC for Network EquipmentEN 55024: 1998 + A1: 2001 + A2: 2003EN 61000-3-2: Harmonic Current Emissions EN 61000-3-3: Voltage Fluctuations and Flicker EN 61000-4-2: ESDEN 61000-4-3: Radiated Immunity EN 61000-4-4: EFT EN 61000-4-5: SurgeEN 61000-4-6: Low Frequency Conducted ImmunityRoHSAll Z-Series components are EU RoHS compliant.CertificationsTAA (T rade Agreement Act) compliant models also available。

  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

A Unified Tagging Approach to Text NormalizationConghui ZhuHarbin Institute of TechnologyHarbin, Chinachzhu@ Jie TangDepartment of Computer Science Tsinghua University, China jietang@Hang Li Microsoft Research AsiaBeijing, China hangli@Hwee Tou NgDepartment of Computer Science National University of Singapore, Singapore nght@.sgTie-Jun ZhaoHarbin Institute of TechnologyHarbin, Chinatjzhao@AbstractThis paper addresses the issue of text nor-malization, an important yet often over-looked problem in natural language proc-essing. By text normalization, we meanconverting ‘informally inputted’ text intothe canonical form, by eliminating ‘noises’in the text and detecting paragraph and sen-tence boundaries in the text. Previously,text normalization issues were often under-taken in an ad-hoc fashion or studied sepa-rately. This paper first gives a formaliza-tion of the entire problem. It then proposesa unified tagging approach to perform thetask using Conditional Random Fields(CRF). The paper shows that with the in-troduction of a small set of tags, most ofthe text normalization tasks can be per-formed within the approach. The accuracyof the proposed method is high, becausethe subtasks of normalization are interde-pendent and should be performed together.Experimental results on email data cleaningshow that the proposed method signifi-cantly outperforms the approach of usingcascaded models and that of employing in-dependent models.1IntroductionMore and more ‘informally inputted’ text data be-comes available to natural language processing, such as raw text data in emails, newsgroups, fo-rums, and blogs. Consequently, how to effectively process the data and make it suitable for natural language processing becomes a challenging issue. This is because informally inputted text data is usually very noisy and is not properly segmented. For example, it may contain extra line breaks, extra spaces, and extra punctuation marks; and it may contain words badly cased. Moreover, the bounda-ries between paragraphs and the boundaries be-tween sentences are not clear.We have examined 5,000 randomly collected emails and found that 98.4% of the emails contain noises (based on the definition in Section 5.1).In order to perform high quality natural lan-guage processing, it is necessary to perform ‘nor-malization’ on informally inputted data first, spe-cifically, to remove extra line breaks, segment the text into paragraphs, add missing spaces and miss-ing punctuation marks, eliminate extra spaces and extra punctuation marks, delete unnecessary tokens, correct misused punctuation marks, restore badly cased words, correct misspelled words, and iden-tify sentence boundaries.Traditionally, text normalization is viewed as an engineering issue and is conducted in a more or less ad-hoc manner. For example, it is done by us-ing rules or machine learning models at different levels. In natural language processing, several is-sues of text normalization were studied, but were only done separately.This paper aims to conduct a thorough investiga-tion on the issue. First, it gives a formalization ofthe problem; specifically, it defines the subtasks of the problem. Next, it proposes a unified approach to the whole task on the basis of tagging. Specifi-cally, it takes the problem as that of assigning tags to the input texts, with a tag representing deletion, preservation, or replacement of a token. As the tagging model, it employs Conditional Random Fields (CRF). The unified model can achieve better performances in text normalization, because the subtasks of text normalization are often interde-pendent. Furthermore, there is no need to define specialized models and features to conduct differ-ent types of cleaning; all the cleaning processes have been formalized and conducted as assign-ments of the three types of tags.Experimental results indicate that our method significantly outperforms the methods using cas-caded models or independent models on normali-zation. Our experiments also indicate that with the use of the tags defined, we can conduct most of the text normalization in the unified framework.Our contributions in this paper include: (a) for-malization of the text normalization problem, (b) proposal of a unified tagging approach, and (c) empirical verification of the effectiveness of the proposed approach.The rest of the paper is organized as follows. In Section 2, we introduce related work. In Section 3, we formalize the text normalization problem. In Section 4, we explain our approach to the problem and in Section 5 we give the experimental results. We conclude the paper in Section 6.2Related WorkText normalization is usually viewed as an engineering issue and is addressed in an ad-hoc manner. Much of the previous work focuses on processing texts in clean form, not texts in informal form. Also, prior work mostly focuses on processing one type or a small number of types of errors, whereas this paper deals with many different types of errors.Clark (2003) has investigated the problem of preprocessing noisy texts for natural language processing. He proposes identifying token bounda-ries and sentence boundaries, restoring cases of words, and correcting misspelled words by using a source channel model.Minkov et al. (2005) have investigated the prob-lem of named entity recognition in informally in-putted texts. They propose improving the perform-ance of personal name recognition in emails using two machine-learning based methods: Conditional Random Fields and Perceptron for learning HMMs. See also (Carvalho and Cohen, 2004).Tang et al. (2005) propose a cascaded approach for email data cleaning by employing Support Vec-tor Machines and rules. Their method can detect email headers, signatures, program codes, and ex-tra line breaks in emails. See also (Wong et al., 2007).Palmer and Hearst (1997) propose using a Neu-ral Network model to determine whether a period in a sentence is the ending mark of the sentence, an abbreviation, or both. See also (Mikheev, 2000; Mikheev, 2002).Lita et al. (2003) propose employing a language modeling approach to address the case restoration problem. They define four classes for word casing: all letters in lower case, first letter in uppercase, all letters in upper case, and mixed case, and formal-ize the problem as assigning class labels to words in natural language texts. Mikheev (2002) proposes using not only local information but also global information in a document in case restoration. Spelling error correction can be formalized as a classification problem. Golding and Roth (1996) propose using the Winnow algorithm to address the issue. The problem can also be formalized as that of data conversion using the source channel model. The source model can be built as an n-gram language model and the channel model can be con-structed with confusing words measured by edit distance. Brill and Moore, Church and Gale, and Mayes et al. have developed different techniques for confusing words calculation (Brill and Moore, 2000; Church and Gale, 1991; Mays et al., 1991). Sproat et al. (1999) have investigated normaliza-tion of non-standard words in texts, including numbers, abbreviations, dates, currency amounts, and acronyms. They propose a taxonomy of non-standard words and apply n-gram language models, decision trees, and weighted finite-state transduc-ers to the normalization.3Text NormalizationIn this paper we define text normalization at three levels: paragraph, sentence, and word level. The subtasks at each level are listed in Table 1. For ex-ample, at the paragraph level, there are two sub-tasks: extra line-break deletion and paragraph boundary detection. Similarly, there are six (three) subtasks at the sentence (word) level, as shown in Table 1. Unnecessary token deletion refers to dele-tion of tokens like ‘-----’ and ‘====’, which are not needed in natural language processing. Note that most of the subtasks conduct ‘cleaning’ of noises, except paragraph boundary detection and sentence boundary detection.Table 1. Text Normalization SubtasksAs a result of text normalization, a text is seg-mented into paragraphs; each paragraph is seg-mented into sentences with clear boundaries; and each word is converted into the canonical form. After normalization, most of the natural language processing tasks can be performed, for example, part-of-speech tagging and parsing.We have manually cleaned up some email data (cf., Section 5) and found that nearly all the noises can be eliminated by performing the subtasks de-fined above. Table 1 gives the statistics.1. i’m thinking about buying a pocket2. pc device for my wife this christmas,.3. the worry that i have is that she won’t4. be able to sync it to her outlook express5. contacts…Figure 1. An example of informal textI’m thinking about buying a Pocket PC device for my wife this Christmas.// The worry that I have is thatshe won’t be able to sync it to her Outlook Express contacts.//Figure 2. Normalized textFigure 1 shows an example of informally input-ted text data. It includes many typical noises. From line 1 to line 4, there are four extra line breaks at the end of each line. In line 2, there is an extra comma after the word ‘Christmas’. The first word in each sentence and the proper nouns (e.g., ‘Pocket PC’ and ‘Outlook Express’) should be capitalized. The extra spaces between the words ‘PC’ and ‘device’ should be removed. At the end of line 2, the line break should be removed and a space is needed after the period. The text should be segmented into two sentences.Figure 2 shows an ideal output of text normali-zation on the input text in Figure 1. All the noises in Figure 1 have been cleaned and paragraph and sentence endings have been identified.We must note that dependencies (sometimes even strong dependencies) exist between different types of noises. For example, word case restoration needs help from sentence boundary detection, and vice versa. An ideal normalization method should consider processing all the tasks together.4 A Unified Tagging Approach4.1ProcessIn this paper, we formalize text normalization as a tagging problem and employ a unified approach to perform the task (no matter whether the processing is at paragraph level, sentence level, or word level). There are two steps in the method: preprocess-ing and tagging. In preprocessing, (A) we separate the text into paragraphs (i.e., sequences of tokens), (B) we determine tokens in the paragraphs, and (C) we assign possible tags to each token. The tokens form the basic units and the paragraphs form the sequences of units in the tagging problem. In tag-ging, given a sequence of units, we determine the most likely corresponding sequence of tags by us-ing a trained tagging model. In this paper, as the tagging model, we make use of CRF.Next we describe the steps (A)-(C) in detail and explain why our method can accomplish many of the normalization subtasks in Table 1.(A). We separate the text into paragraphs by tak-ing two or more consecutive line breaks as the end-ings of paragraphs.(B). We identify tokens by using heuristics. There are five types of tokens: ‘standard word’, ‘non-standard word’, punctuation mark, space, and line break. Standard words are words in natural language. Non-standard words include several general ‘special words’ (Sproat et al., 1999), email address, IP address, URL, date, number, money, percentage, unnecessary tokens (e.g., ‘===‘ and‘###’), etc. We identify non-standard words by using regular expressions. Punctuation marks in-clude period, question mark, and exclamation mark. Words and punctuation marks are separated into different tokens if they are joined together. Natural spaces and line breaks are also regarded as tokens. (C). We assign tags to each token based on the type of the token. Table 2 summarizes the types of tags defined.Token Type TagDescriptionPRV Preserve line breakRPA Replace line break by spaceLine break DELDelete line breakPRV Preserve spaceSpaceDEL Delete spacePSBPreserve punctuation mark and view itas sentence endingPRVPreserve punctuation mark withoutviewing it as sentence endingPunctuationmark DELDelete punctuation markAUCMake all characters in uppercaseALCMake all characters in lowercaseFUCMake the first character in uppercase WordAMCMake characters in mixed case PRVPreserve the special tokenSpecial tokenDELDelete the special tokenTable 2. Types of tagsFigure 3. An example of taggingFigure 3 shows an example of the tagging proc-ess. (The symbol ‘ ’ indicates a space). In the fig-ure, a white circle denotes a token and a gray circle denotes a tag. Each token can be assigned several possible tags.Using the tags, we can perform most of the text normalization processing (conducting seven types of subtasks defined in Table 1 and cleaning 90.55% of the noises).In this paper, we do not conduct three subtasks, although we could do them in principle. These in-clude missing space insertion, missing punctuationmark insertion, and misspelled word correction. Inour email data, it corresponds to 8.81% of the noises. Adding tags for insertions would increase the search space dramatically. We did not do that due to computation consideration. Misspelled word correction can be done in the same framework eas-ily. We did not do that in this work, because the percentage of misspelling in the data is small.We do not conduct misused punctuation mark correction as well (e.g., correcting ‘.’ with ‘?’). It consists of 0.64% of the noises in the email data. To handle it, one might need to parse the sentences. 4.2CRF ModelWe employ Conditional Random Fields (CRF) as the tagging model. CRF is a conditional probability distribution of a sequence of tags given a sequence of tokens, represented as P(Y|X) , where X denotes the token sequence and Y the tag sequence (Lafferty et al., 2001).In tagging, the CRF model is used to find the sequence of tags Y * having the highest likelihood Y * = max Y P (Y |X ), with an efficient algorithm (the Viterbi algorithm).In training, the CRF model is built with labeled data and by means of an iterative algorithm based on Maximum Likelihood Estimation.Transition Featuresy i-1=y’, y i =y y i -1=y’, y i =y , w i =w y i -1=y’, y i =y , t i =t State Featuresw i =w , y i =yw i-1=w , y i =y w i-2=w , y i =y w i-3=w , y i =y w i-4=w , y i =y w i +1=w , y i =y w i +2=w , y i =y w i +3=w , y i =y w i +4=w , y i =y w i -1=w’, w i =w , y i =y w i +1=w’, w i =w , y i =y t i =t , y i =y t i -1=t , y i =y t i -2=t , y i =y t i -3=t , y i =y t i -4=t , y i =y t i +1=t , y i =y t i +2=t , y i =y t i +3=t , y i =y t i +4=t , y i =y t i -2=t’’, t i -1=t’, y i =y t i -1=t’, t i =t , y i =yt i =t , t i +1=t’, y i =y t i +1=t’, t i +2=t’’, y i =y t i -2=t’’, t i -1=t’, t i =t , y i =y t i -1=t’’, t i =t , t i +1=t’, y i =y t i =t , t i +1=t’, t i +2=t’’, y i =yTable 3. Features used in the unified CRF model4.3FeaturesTwo sets of features are defined in the CRF model: transition features and state features. Table 3 shows the features used in the model.Suppose that at position i in token sequence x, w i is the token, t i the type of token (see Table 2), and y i the possible tag. Binary features are defined as described in Table 3. For example, the transition feature y i-1=y’, y i=y implies that if the current tag is y and the previous tag is y’, then the feature value is true; otherwise false. The state feature w i=w, y i=y implies that if the current token is w and the current label is y, then the feature value is true; otherwise false. In our experiments, an actual fea-ture might be the word at position 5 is ‘PC’ and the current tag is AUC. In total, 4,168,723 features were used in our experiments.4.4Baseline MethodsWe can consider two baseline methods based on previous work, namely cascaded and independent approaches. The independent approach performs text normalization with several passes on the text. All of the processes take the raw text as input and output the normalized/cleaned result independently. The cascaded approach also performs normaliza-tion in several passes on the text. Each process car-ries out cleaning/normalization from the output of the previous process.4.5AdvantagesOur method offers some advantages.(1) As indicated, the text normalization tasks are interdependent. The cascaded approach or the in-dependent approach cannot simultaneously per-form the tasks. In contrast, our method can effec-tively overcome the drawback by employing a uni-fied framework and achieve more accurate per-formances.(2) There are many specific types of errors one must correct in text normalization. As shown in Figure 1, there exist four types of errors with each type having several correction results. If one de-fines a specialized model or rule to handle each of the cases, the number of needed models will be extremely large and thus the text normalization processing will be impractical. In contrast, our method naturally formalizes all the tasks as as-signments of different types of tags and trains a unified model to tackle all the problems at once. 5Experimental Results5.1Experiment SettingData SetsWe used email data in our experiments. We ran-domly chose in total 5,000 posts (i.e., emails) from12 newsgroups. DC, Ontology, NLP, and ML arefrom newsgroups at Google (/groups). Jena is a newsgroup at Ya-hoo (/group/jena-dev). Wekais a newsgroup at Waikato University (https://list. ). Protégé and OWL are from a project at Stanford University (/). Mobility, WinServer, Windows, and PSS are email collections from a company.Five human annotators conducted normalizationon the emails. A spec was created to guide the an-notation process. All the errors in the emails were labeled and corrected. For disagreements in the annotation, we conducted “majority voting”. For example, extra line breaks, extra spaces, and extra punctuation marks in the emails were labeled. Un-necessary tokens were deleted. Missing spaces and missing punctuation marks were added and marked. Mistakenly cased words, misspelled words, and misused punctuation marks were corrected. Fur-thermore, paragraph boundaries and sentence boundaries were also marked. The noises fell intothe categories defined in Table 1.Table 4 shows the statistics in the data sets. From the table, we can see that a large number of noises (41,407) exist in the emails. We can also seethat the major noise types are extra line breaks, extra spaces, casing errors, and unnecessary tokens.In the experiments, we conducted evaluations in terms of precision, recall, F1-measure, and accu-racy (for definitions of the measures, see for ex-ample (van Rijsbergen, 1979; Lita et al., 2003)). Implementation of Baseline MethodsWe used the cascaded approach and the independ-ent approach as baselines.For the baseline methods, we defined several basic prediction subtasks: extra line break detec-tion, extra space detection, extra punctuation mark detection, sentence boundary detection, unneces-sary token detection, and case restoration. We compared the performances of our method with those of the baseline methods on the subtasks.Data Set NumberofEmailNumberofNoisesExtraLineBreakExtraSpaceExtraPunc.MissingSpaceMissingPunc.CasingErrorSpellingErrorMisusedPunc.Unnece-ssaryTokenNumber ofParagraphBoundaryNumber ofSentenceBoundaryDC 100 702 476 31 8 3 24 53 14 2 91 457 291 Ontology 100 2,731 2,132 24 3 10 68 205 79 15 195 677 1,132 NLP 60 861 623 12 1 3 23 135 13 2 49 244 296 ML 40 980 868 17 0 2 13 12 7 0 61 240 589 Jena 700 5,833 3,066 117 42 38 234 888 288 59 1,101 2,999 1,836 Weka 200 1,721 886 44 0 30 37 295 77 13 339 699 602 Protégé 700 3,306 1,770 127 48 151 136 552 116 9 397 1,645 1,035 OWL 300 1,232 680 43 24 47 41 152 44 3 198 578 424 Mobility 400 2,2961,292 64 22 35 87 495 92 8 201 891 892 WinServer 400 3,487 2,029 59 26 57 142 822 121 21 210 1,232 1,151 Windows 1,000 9,293 3,416 3,056 60 116 348 1,309291 67 630 3,581 2,742 PSS 1,000 8,965 3,348 2,880 59 153 296 1,331276 66 556 3,411 2,590 Total5,000 41,407 20,586 6,474 293645 1,4496,2491,418265 4,028 16,654 13,580Table 4. Statistics on data setsFor the case restoration subtask (processing on token sequence), we employed the TrueCasing method (Lita et al., 2003). The method estimates a tri-gram language model using a large data corpus with correctly cased words and then makes use of the model in case restoration. We also employed Conditional Random Fields to perform case restoration, for comparison purposes. The CRF based casing method estimates a conditional probabilistic model using the same data and the same tags defined in TrueCasing.For unnecessary token deletion, we used rules as follows. If a token consists of non-ASCII charac-ters or consecutive duplicate characters, such as ‘===‘, then we identify it as an unnecessary token. For each of the other subtasks, we exploited the classification approach. For example, in extra line break detection, we made use of a classification model to identify whether or not a line break is a paragraph ending. We employed Support Vector Machines (SVM) as the classification model (Vap-nik, 1998). In the classification model we utilized the same features as those in our unified model (see Table 3 for details).In the cascaded approach, the prediction tasks are performed in sequence, where the output of each task becomes the input of each immediately following task. The order of the prediction tasks is: (1) Extra line break detection: Is a line break a paragraph ending? It then separates the text into paragraphs using the remaining line breaks. (2) Extra space detection: Is a space an extra space? (3) Extra punctuation mark detection: Is a punctuation mark a noise? (4) Sentence boundary detection: Is a punctuation mark a sentence boundary? (5) Un-necessary token deletion: Is a token an unnecessary token? (6) Case restoration. Each of step s (1) to (4) uses a classification model (SVM), step (5) uses rules, whereas step (6) uses either a language model (TrueCasing) or a CRF model (CRF).In the independent approach, we perform the prediction tasks independently. When there is a conflict between the outcomes of two classifiers, we adopt the result of the latter classifier, as de-termined by the order of classifiers in the cascaded approach.To test how dependencies between different types of noises affect the performance of normali-zation, we also conducted experiments using the unified model by removing the transition features. Implementation of Our MethodIn the implementation of our method, we used the tool CRF++, available at /~taku /software/CRF++/. We made use of all the default settings of the tool in the experiments.5.2Text Normalization ExperimentsResultsWe evaluated the performances of our method (Unified) and the baseline methods (Cascaded and Independent) on the 12 data sets. Table 5 shows the five-fold cross-validation results. Our method outperforms the two baseline methods.Table 6 shows the overall performances of text normalization by our method and the two baseline methods. We see that our method outperforms the two baseline methods. It can also be seen that the performance of the unified method decreases when removing the transition features (Unified w/o Transition Features).We conducted sign tests for each subtask on the results, which indicate that all the improvements of Unified over Cascaded and Independent are statis-tically significant (p << 0.01).Detection Task Prec. Rec. F1Acc.Independent 95.16 91.52 93.3093.81Cascaded 95.16 91.52 93.3093.81Extra LineBreakUnified 93.87 93.63 93.7594.53Independent 91.85 94.64 93.2299.87Cascaded 94.54 94.56 94.5599.89Extra Space Unified 95.17 93.98 94.5799.90Independent 88.63 82.69 85.5699.66Cascaded 87.17 85.37 86.2699.66ExtraPunctuation Mark Unified 90.94 84.84 87.7899.71Independent 98.46 99.62 99.0498.36Cascaded 98.55 99.20 98.8798.08SentenceBoundaryUnified 98.76 99.61 99.1898.61Independent 72.51 100.0 84.0684.27Cascaded 72.51 100.0 84.0684.27UnnecessaryTokenUnified 98.06 95.47 96.7596.18Independent 27.32 87.44 41.6396.22Case Restoration(TrueCasing) Cascaded 28.04 88.21 42.5596.35Independent 84.96 62.79 72.2199.01Cascaded 85.85 63.99 73.3399.07CaseRestoration (CRF) Unified 86.65 67.09 75.6399.21Table 5. Performances of text normalization (%)Text Normalization Prec. Rec. F1 Acc.Independent (TrueCasing) 69.54 91.33 78.9697.90Independent (CRF)85.05 92.52 88.6398.91Cascaded (TrueCasing) 70.29 92.07 79.7297.88Cascaded (CRF)85.06 92.70 88.7298.92Unified w/o TransitionFeatures86.03 93.45 89.5999.01Unified 86.46 93.92 90.0499.05Table 6. Performances of text normalization (%) DiscussionsOur method outperforms the independent method and the cascaded method in all the subtasks, espe-cially in the subtasks that have strong dependen-cies with each other, for example, sentence bound-ary detection, extra punctuation mark detection, and case restoration.The cascaded method suffered from ignorance of the dependencies between the subtasks. For ex-ample, there were 3,314 cases in which sentence boundary detection needs to use the results of extra line break detection, extra punctuation mark detec-tion, and case restoration. However, in the cas-caded method, sentence boundary detection is con-ducted after extra punctuation mark detection and before case restoration, and thus it cannot leveragethe results of case restoration. Furthermore, errors of extra punctuation mark detection can lead to errors in sentence boundary detection.The independent method also cannot make use of dependencies across different subtasks, because it conducts all the subtasks from the raw input data. This is why for detection of extra space, extra punctuation mark, and casing error, the independ-ent method cannot perform as well as our method. Our method benefits from the ability of model-ing dependencies between subtasks. We see from Table 6 that by leveraging the dependencies, our method can outperform the method without using dependencies (Unified w/o Transition Features) by 0.62% in terms of F1-measure.Here we use the example in Figure 1 to show the advantage of our method compared with the inde-pendent and the cascaded methods. With normali-zation by the independent method, we obtain:I’m thinking about buying a pocket PC device for my wife this Christmas, The worry that I have is that she won’t be able to sync it to her outlook express contacts.//With normalization by the cascaded method, weobtain:I’m thinking about buying a pocket PC device for my wife this Christmas, the worry that I have is that she won’t be able to sync it to her outlook express contacts.//With normalization by our method, we obtain:I’m thinking about buying a Pocket PC device for my wife this Christmas.// The worry that I have is that she won’t be able to sync it to her Outlook Express contacts.//The independent method can correctly deal with some of the errors. For instance, it can capitalize the first word in the first and the third line, remove extra periods in the fifth line, and remove the four extra line breaks. However, it mistakenly removes the period in the second line and it cannot restore the cases of some words, for example ‘pocket’ and ‘outlook express’.In the cascaded method, each process carries out cleaning/normalization from the output of the pre-vious process and thus can make use of the cleaned/normalized results from the previous proc-ess. However, errors in the previous processes will also propagate to the later processes. For example, the cascaded method mistakenly removes the pe-riod in the second line. The error allows case resto-ration to make the error of keeping the word ‘the’ in lower case.。

相关文档
最新文档