2014.6.27 김우주 연세대학교 정보산업공학과 목차 I. 빅데이터 시대와 정보의 홍수 II. 빅데이터 활용 사례 III. 빅데이터의 한계와 극복 방안 IV. Linked Data의 구축과 활용 V. LOD 2 - 시맨틱 기술의 미래 2 3 An Instrumented Interconnected World 30 billion RFID 12+ TBs camera phones world wide 100s of millions of GPS enabled data every day ? TBs of of tweet data every day tags today (1.3B in 2005) 4.6 billion devices sold annually 25+ TBs of 2+ billion log data every day 76 million smart meters in 2009… 200M by 2014 people on the Web by end 2011 Information Overflow on the Web Growth of the Web The amount of information available on the Web grows so fast. The February 2014 survey shows there exist at least 920,120,079 sites (http://news.netcraft.com/archives/category/web-server-survey/). 5 Information Overflow on the Web The Indexed Web contains at least 19.8 billion pages (Sunday, 02 March, 2014). http://www.worldwidewebsize.com/ 6 빅데이터란? 빅데이터란? (07/11/2013, European Commission) Every minute the world generates 1.7 million billion bytes of data, equivalent to 360,000 standard DVDs. The big data sector is growing at a rate of 40% a year. 무엇이 빅데이터를 중요하게 하는가? Big data is already affecting all areas of the economy. Data-driven decision making leads to 5-6% efficiency gains in the different sectors observed. Intelligent processing of data is also essential for addressing societal challenges. 7 IBM의 예측: 2014년 6대 빅데이터 트렌드 직감보다는 더 분석적인 경영 방식 Companies will grow increasingly data driven and willing to apply analyticsderived insights to key business operations. 빅데이터 프라이버시와 보안 문제 Organizations will make a greater effort to build security, privacy, and governance policies into their big data processes. 빅데이터에 대한 투자 확대 CDO(Chief Data Officer)의 등장 More organizations will bring a chief data officer (CDO) on board. 보다 유용한 빅데이터 응용 시스템 외부 데이터에 대한 관심 증대 8 LOD를 말하다! 9 구글의 독감 트렌드 ‘독감’ 관련 검색어 분석을 통한 독감 예보 가능성 확인 구글 검색 사이트에 사용자가 남긴 검색어의 빈도를 조사, 독감 환자의 분포 및 확산 정보 제공 10 샌프란시스코, 범죄 예방 시스템 과거 범죄 발생 지역과 시각 패턴 분석을 통한 경찰 인력 배치 과거 발생한 범죄 패턴을 분석하여 후속 범죄 가능성 예측 과거 데이터에서 범죄자 행동을 분석하여 사건 예방을 위한 해법 제시 11 미국 국세청, 탈세 방지 시스템 빅데이터 분석을 통한 탈세 및 사기 범죄 예방 시스템 구축 사기 방지 솔루션, 소셜 네트워크 분석, 데이터 통합 및 마이닝 등 활용 세금 누락 및 불필요한 세금 환급 절감의 효과 발생 12 KT, 서울특별시 – 빅데이터 기반 심야버스 노선 정책 지원 심야버스 노선 결정을 위한 유동인구 분석 및 노선 분석 서울시의 교통 환경(정류장/전용차로/환승)기반 지역별 최적 정류장 위치를 도출하고 KT의 CDR데이터 기반 심야시간 유동인구 및 목적지 통계를 융합하여 노선 검증 13 비씨카드, 점포 평가 서비스 소상공인 창업 성공률 제고를 위한 상가데이터 및 신용카드거래데이터 기반의 빅데이터 분석 점포이력, 상권분석, 업종추천 등이 이루어지는 과거현황분석, 추천 업종 또는 사용자 선택 업종 매출예측, 수익예측 등의 서비스 제공 14 15 Information Overflow Problems Problems How to cover all available information? - Recall How to find the relevant information? - Precision Not data (search), but integration, analysis and insight, leading to decisions and discovery 16 Example Query to Google ‘iPad’ 검색 사례 17 Information Silo Problem Stove-piped Systems and Poor Content Aggregation Semantic Interoperability To cope with the problems mentioned in the preceding slide, we need Semantic Interoperability. Semantics “The meaning or the interpretation of a word, sentence, or other language form.” What is Semantic Interoperability? “Processing or Integration of resources based on the understanding what’s intended or expressed by other systems or parties.’’ 19 Front-endedness? 20 What if I want to ... Move my content from one place to another? RSS ? Not enough Aggregate my data An open FriendFeed? Re-use my Flickr friends on Twitter? Invite. Again and again ... The Semantic Web and Ontology can help ! By providing a common framework to interlink data from various providers in an open way. 21 How is it Possible? Ontology: Agreement with Common Vocabulary & Domain Knowledge Semantic Annotation: metadata (manual & automatic metadata extraction) Reasoning: semantics enabled search, integration, analysis, mining, discovery 22 Semantic Web Layer Cake 23 Three Technical Building Block Basic Building Block URIs for unambiguous names for resources, RDF for common data model for expressing metadata, Ontology(OWL) for common vocabularies. Semantic Web becomes: web of data/things/concepts • What is a Thing/Concept? It can be anything in the world - a movie, a person, a disease, a location… • Machines will be able to understand the concept behind a html page. • This page is talking about ‘Barack Obama’, He is a ‘Person’ and he is the ‘President of USA’ ? 24 Who borrows this Idea? Facebook Facebook Open Graph Protocol and Graph Search Google Knowledge Graph Twitter Real-time Semantic Web with Twitter Annotations 25 LOD를 말하다! 26 Linked Data Building a “Web of Data” to enhance the current Web The Linking Open Data (LOD) project: http://linkeddata.org/ Translating existing datasets into RDF and linking them together. • For example, DBpedia (Wikipedia) and GeoNames, Freebase, BBC programmes, etc. Government data also available as Linked Data • DATA.gov • DATA.gov.uk 27 The LOD cloud 2007 2008 28 The LOD cloud 2008 2009 29 Web of Data 30 Web of Data (Statistics) The size of the Web of Data The size of the Web of Data can be estimated based on the data set statistics that are collected by the LOD community in the ESW wiki. According to these statistics, the Web of Data currently consists of 31 billion RDF triples, which are interlinked by around 500 million RDF inter-links (09/19/2011). 31 Types of Linked Data Applications Linked Data의 활용 방안 32 Semantic Search Engines Top 7 Semantic Search Engines as An Alternative to Google Kngine Hakia Kosmix: now is part of @WalmartLabs DuckDuckGo Evri: specialized for iPad and iPhone Powerset: now is part of Bing Truevert: focus only on environmental concerns. 33 LOD를 말하다! 34 LOD2 : What is LOD2? LOD2(Linked Open Data) LOD2 is the large-scale integrating project co-funded by the European Commission within the FP7 Information and Communication Technologies Work Programme. • Started in September 2010 Partners • 14 partners (11 European Country) 35 LOD2 : Objectives of LOD2 LOD2 Project Objectives Achieving visualization, deployment, sharing, accessibility for linked open data by software technology. • Increase visibility of Linked Data activities [Visualization] • Support deployment Linked Data components [Deployment] • Improve information sharing between Linked Data components so that publishing Linked Data is eased. [Sharing] • Improve access to the content: the online Linked Open Data [Accessibility] • Improve the software technology which support it [By software technology] 36 LOD2 Stack : Overview LOD2 Stack LOD2 project provides LOD2 Stack for the sake of easy access to linked data software. the LOD2 software stack is an integrated distribution of aligned tools supporting the life-cycle of Linked Data from extraction, authoring/creation over enrichment, interlinking, fusing to visualization and maintenance 37 LOD2 Stack 3.0 38 LOD2 Stack : The overview of tools Apache Stanbol In the LOD2 Stack, Apache Stanbol can be used for NLP services which rely on the stack internal knowledge bases, such as named entity recognition and text classification. CubeViz CubeViz is a facetted browser for statistical data utilizing the RDF Data Cube vocabulary which is the state-of-the-art in representing statistical data in RDF. 39 LOD2 Stack : The overview of tools Dbpedia Spotlight DBpedia Spotlight is a tool for automatically annotating mentions of DBpedia resources in text, providing a solution for linking unstructured information sources to the Linked Open Data cloud through DBpedia. D2RQ D2RQ is a system for accessing relational databases(RDBMS) as virtual RDF graphs. 40 LOD2 Stack : The overview of tools DL-Learner The DL-Learner software learns concepts in Description Logics (DLs) from user-provided examples. (Supervised-learning) ORE The ORE (Ontology Repair and Enrichment) tool allows for knowledge engineers to improve an OWL ontology by fixing inconsistencies and making suggestions for adding further axioms to it. 41 LOD2 Stack : The overview of tools Poolparty The PoolParty Extractor (PPX) offers an API providing text mining algorithms based on semantic knowledge models. 42 LOD2 Stack : The overview of tools SemMap SemMap allows to visualize knowledge bases having a spatial dimension. Silk The Silk Link Discovery Framework supports data publishers in accomplishing the second task. Using the declarative Silk - Link Specification Language (Silk-LSL), developers can specify which types of RDF links should be discovered between data sources as well as which conditions data items must fulfill in order to be interlinked. 43 LOD2 Stack : The overview of tools Sieve Sieve allows Web data to be filtered according to different data quality assessment policies and provides for fusing Web data according to different conflict resolution methods. LIMES LIMES is a link discovery framework for the Web of Data. It implements time-efficient approaches for large-scale link discovery based on the characteristics of metric spaces. 44 Silk : Link Discovery Framework Interlinking and Fusion Stage Component of LOD2 Stack Can be used by data providers to generate RDF links between data sets on the web of data • Especially, to set explicit RDF links between data items within different data sources “Data publishers can use Silk to set RDF links from their data sources to other data sources on the Web” 45 Silk : Silk – Link Specification Language Example Aggregation Example: Combines multiple confidence values into a single value (average) Confidence value is the average of two compared weight Numeric differences between parameters 46 DL-Learner Introduction The goal of DL-Learner is to provide a DL/OWL based machine learning tool to solve supervised learning tasks. The DL-Learner software learns concepts in Description Logics (DLs) from examples. DL-Learner : Features Learning Problems Positive and Negative Examples (=previous example) Class Learning • Find out Class Expression for given class • father ≡ hasChild 𝐬𝐨𝐦𝐞 male 𝐨𝐫 female 𝐚𝐧𝐝 𝐧𝐨𝐭 female Demo of SDT Plug-in to Protégé 49 SWCL - Sample Example Country PopulationValue ? hasPart Province positiveInteger PopulationValue positiveInteger 𝑥. 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑉𝑎𝑙𝑢𝑒 = 𝑦. 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑉𝑎𝑙𝑢𝑒, for all 𝑦 ∈ 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑥∈ℎ𝑎𝑠𝑃𝑎𝑟𝑡 − . 𝑦 50 𝐶 Constraints Representation in SWCL Target Constraint 𝑥∈ℎ𝑎𝑠𝑃𝑎𝑟𝑡 − . 𝑦 𝑥. 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑉𝑎𝑙𝑢𝑒 = 𝑦. 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑉𝑎𝑙𝑢𝑒, for all 𝑦 ∈ 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 Corresponding SWCL Code <swcl:Constraint rdf:ID=”numberOfPopulation"> <swcl:qualifier> <swcl:Variable rdf:id="y"> <swcl:bindingClass rdf:resource="#Country"/> </swcl:Variable> </swcl:qualifier> <swcl:hasLHS> <swcl:TermBlock rdf:ID="termBlock_1"> <swcl:sign rdf:resource="&swcl;plus"/> <swcl:aggregateOperator rdf:resource="&swcl;Sigma"/> <swcl:parameter> <swcl:Variable rdf:id="x"> <rdfs:subClassOf> <owl:Restriction> <owl:onProperty rdf:resource="#partOf"/> <owl:hasValue rdf:resource="#y"/> </owl:Restriction> </rdfs:subClassOf> </swcl:Variable> </swcl:parameter> <swcl:factor> <swcl:FactorAtom> <swcl:bindingClass rdf:resource="#x"/> 51 𝐶 Our Direction to the Future Directions Open, Share your data, whenever and wherever you want Semantic, Enhance your data, to make more sense of it An example: LinkedGeoData.org We need an integrated framework to enhance communication and information sharing in GeoData. 52 Q&A 53