Bikesh - Beginning Smart Phone Web Development
Check out this SlideShare Presentation:
Open Source Tools for Creating Mashups with Government Datasets MOSC2010
Mohammed Firdaus, Muhd Sharuzzamal Bakri
Malaysia Open Source Conference 2010
Open Source Tools for Creating Mashups with Government Datasets MOSC2010 - Presentation Transcript
Open Source Tools for Creating Mashups with Government Datasets Mohammed Firdaus, Muhd Sharuzzamal Bakri June 29, 2010 Mohammed Firdaus, Muhd Sharuzzamal Bakri
Introduction About the Speakers About the Speakers Mohammed Firdaus bin Mohammed Ab Halim (@ﬁrdaus halim) and Muhd Sharuzzamal Bakri (@amai) Founders of Persada Terbilang Sdn Bhd - We have no relationship whatsoever to any fertilizer supplier
Introduction What are Mashups? Mashups A mashup is a web page or application that uses and combines data, presentation or functionality from two or more sources to create new services. (Source: Wikipedia) Data mashups combine similar types of media and information from multiple sources into a single representation. (Source: Wikipedia)
Challenges Data Sets are Not Available in Machine Readable Form Data Sets are Not Available in Machine Readable Form Nothing useful here: ﬁletype:csv site:.gov.my ﬁletype:xml site:.gov.my ﬁletype:rdf site:.gov.my We have to resort to web scraping.
Challenges No Data Dictionaries No Data Dictionaries Since the data sets that are available were meant for humans to consume rather machines they are usually published without any type of data dictionary. This means that an application developer will have to make assumptions about the structure of each ﬁeld e.g. whether it’s unique, whether it’s a multi-value ﬁeld, which ﬁelds are mandatory/option. These assumptions may or may not turn out be correct as you see more and more data in the data set.
Challenges New Data Sets Constantly Become Available New Data Sets Constantly Become Available This is a not a bad thing. However, our code, database and schema must be ﬂexible enough to deal with future data sets that we might want to use in our applications.
Challenges Lack of Standards Across Agencies Lack of Standards Across Agencies Diﬀerent identiﬁers for referring to the same entity. The lack of common identiﬁers makes it tedious to combine data sets together which maybe describing the same entity. MyCoID and MyID are steps in the right direction.
Challenges Summary In Summary Because of these challenges, we need an agile method for modeling, storing and processing these government datasets in our application. The purpose of this presentation is to show how representing your data as a graph both help you deal with these challenges and at the same time help make compelling data mashups. ===
Graphs Introduction to Graphs What is a Graph? A data structure that consists of a collection of vertices and the connections between those vertices, called edges. Vertices are sometimes called nodes or dots. Edges are sometimes called relationships or edges. The terminology diﬀers between software packages.
Graphs Types of Graphs Types of Graphs A directed graph (or digraph) is one where the edges have a direction (i.e. there’s an outgoing and incoming vertex). A multigraph is one where multiple edges can exist between two vertices. An edge-labeled graph is a graph where edges have labels. Similarly, a vertex-labeled graph is one in which the vertices have labels. An attributed graph is one in which the vertices and edges can have attributes (key-value pairs). A graph can have more than one of these properties e.g. a multi digraph is one which multiple directed edges can exist between two vertices. Mohammed Firdaus, Muhd Sharuzzamal Bakri
Graphs Types of Graphs Examples - Social Graphs Source: http://www.ﬂickr.com/photos/greenem/11696663/ Undirected Graph - Vertices represent people and edges represents friendship. Mohammed Firdaus, Muhd Sharuzzamal Bakri
Graphs Types of Graphs Examples - Web Graph http://en.wikipedia.org/wiki/File:WorldWideWebAroundWikipedia.png Multi-digraph - Vertices represent web pages and directed edges represent links between pages.
Graphs Property Graphs Property Graphs ’Property graph’ is another term for attributed labeled multi-digraph. Property graphs are ﬂexible enough to support most types of graph data. Other types of graphs (with the exception of hypergraphs) can be built on top of property graphs by removing features or using features of the property graph in certain ways. The tools that we are covering in this presentation deal primarily with property graphs.
Graphs Property Graphs Property Graphs Source: http://wiki.github.com/tinkerpop/gremlin/defining-a-property-graph
Data Sets Treasury Procurement Data Treasury - Tenders Awarded Source: http://myprocurement.treasury.gov.my/index.php/en/list-keputusan-tender
Data Sets Treasury Procurement Data Fields Tajuk Tender (Title of Tender) Nombor Tender (Tendor Number) Kategori Perolehan (Procurement Category) Kementerian (Ministry) Petender Berjaya (Winner of Tender) No Pendaftaran Dengan ROB/ROS/ROC (Registration Number with ROB/ROS/ROC) No Pendaftaran Dengan MOF/PKK (Registration Number with MOF/PKK) Harga Setuju Terima (Agreed Upon Value)
Data Sets Treasury Procurement Data Code and Data in Machine Readable Form For this presentation we are using data that we scraped form this site on 2010-04-26 The source code for our scraper and the CSV dump from 2010-04-26 is available at http://mfirdaus.com/mosc-paper/ The dump contains 2615 records.
Data Sets Treasury Procurement Data The Dump
Data Sets Issues with this Data Sets Missing Fields Out of the 2615 records in the dump 510 records were missing a tender number 472 records were missing a category 1836 records were missing a ROB/ROS/ROC number 510 records were missing a MOF no
Data Sets Issues with this Data Sets Tender Numbers are Not Unique 32 records have the same tender number and title as another record 23 records have the same tender number as another record In some cases these appear to be duplicate records since the ﬁelds all match up. In other cases, one or two ﬁelds are slightly diﬀerent indicating that there was a probably a typo (erroneous record was not deleted). In some cases, the other ﬁelds are completely diﬀerent which leads us to think that it’s possible for there to be multiple winners of a tender (need some government oﬃcials to verify this for us).
Data Sets Issues with this Data Sets Format of Tender Numbers Examples of tender numbers: 8/2009 PL.(T).08.2009(JKP) X0141110101090021 128/2009 KBS.S.4-14/69 (T.26/2009) Probably not a good idea to write code that attempts to parse the tender number.
Data Sets Issues with this Data Sets Format of the ”Petender Berjaya” Field SYARIKAT PROSPECTRUM SDN BHD TELEKOM SMART SCHOOL SDN BHD NO.45-8, LEVEL 3, BLOCK C, PLAZA DAMANSARA, JALAN MEDAN SETIA 1, BUKIT DAMANSARA 50490 KUALA LUMPUR 1. GLOBAL AEROSPACE SDN BHD (A002) 2. SYSTEM ALLIANCE TECHNOLOGY SDN. BHD.(A003) 3. KARISMA WIRA SDN. BHD. (A004) 4. KESUMA TECHNOLOGY SDN. BHD (A005) A QUALITY REPUTATION SDN BHD B PRIMABUMI SDN BHD
Data Sets Modeling Modeling this Data Set as a Property Graph One way to model this data as a graph is to: Vertices to represent tenders, ministries and companies/businesses. An ”awarded by” labeled edge to associate a tender with a ministry. An ”awarded to” labeled edge to associate a tender with the winner of the tender (the company/business). Attributes on tender vertices for the tender title, number, value, category Attributes on company/business vertices for the company/business name, ROB/ROC/ROS registration number and MOF registration number. Attributes on ministry vertices from the name of the ministry.
Data Sets Modeling Example
Graph Databases and Neo4j Neo4j - Introduction Neo4j Neo4j is a graph database. Persists data in graph form. Property graph data model with the exception of vertex labels. In Neo4j terms, vertices are nodes, edges are relationships and attributes are properties. Property values can be a String or any Java primitive (arrays of these types are supported as well). Licensed under the AGPLv3. Which basically means that you don’t need a license if your application is released under a compatible free software license. For other uses, you need a commercial license from them.
Graph Databases and Neo4j Neo4j - Introduction Neo4j Written in Java. Bindings available for Python, Ruby, Clojure, Erlang, Groovy, Scalan and PHP. We will be using the Python bindings in this talk. An embedded database, meaning that it runs in the same process space as the application. There’s a standalone REST server for those who prefer it.
Graph Databases and Neo4j Inserting into Neo4j Initializing the Database import neo4j db = neo4j.GraphDatabase("db")
Graph Databases and Neo4j Inserting into Neo4j Creating the Nodes ministry node = db.node(name=ministry, type="ministry") entity node = db.node(name=entity name, no=entity no, mof no=entity mof no, type="business entity") tender node = db.node(no=tender no, title=tender title, category=tender category, value=tender value, type="tender")
Graph Databases and Neo4j Inserting into Neo4j Creating the Relationships tender node.awarded by(ministry node) tender node.awarded to(entity node) ===
Graph Databases and Neo4j Inserting into Neo4j Indexing Nodes ministries = db.index("ministries", create=True) business entities = db.index("business entities", create=True) tenders by no = db.index("tenders by no", create=True) tenders by title = db.index("tenders by title", create=True) tenders by no[tender no] = tender node tenders by title[tender title] = tender node
Graph Databases and Neo4j Inserting into Neo4j The Result
Graph Traversals Traversing the Graph Traversing is the process of walking around the graph.
Graph Traversals Graph Traversal Options Graph Traversal Framework Gremlin SPARQL Manual traversal
Graph Traversals Problem Lets use graph traversal to ﬁnd all the companies who have been awarded contracts by Kementerian Kesihatan.
Graph Traversals Graph Around Kementerian Kesihatan
Graph Traversals Traversal Framework Deﬁning the Traversal # Companies who have gotten contracts from a particular ministry # The start node is a ministry class Contractors(neo4j.Traversal): types = [neo4j.Incoming.awarded by, neo4j.Outgoing.awarded to] order = neo4j.DEPTH FIRST stop = neo4j.STOP AT END OF GRAPH def isReturnable(self, position): if position["type"] == "business entity": return True else: return False
Graph Traversals Traversal Framework Using the Traversal with db.transaction: moh = ministries["KEMENTERIAN KESIHATAN"] contractors = Contractors(moh) for c in contractors: print c["name"]
Graph Traversals Traversal Framework Output RAF SYNERGY SDN BHD PRIMABUMI SDN BHD AVERROES PHARMACEUTICALS SDN BHD QUALITY REPUTATION SDN BHD UNISENDO SDN BHD PRESTIGE PHARMA SDN BHD PHARMANIAGA LOGISTICS SDN BHD IDAMAN PHARMA SDN BHD PHARMASERV ALLIANCES SDN BHD
Graph Traversals Traversing Graphs with Gremlin Gremlin Gremlin is a graph based programming language. Can express complex graph traversals concisely. Available at http://wiki.github.com/tinkerpop/gremlin/
Graph Traversals Traversing Graphs with Gremlin Traversing the Graph with Gremlin $ ./gremlin.sh ,,,/ (o o) --–-oOOo-( )-oOOo--–- gremlin> $ := g:key(”ministries”, ”KEMENTERIAN KESIHATAN”) ==>v gremlin> ./inE[@label=”awarded by”]/outV/ outE[@label=”awarded to”][email protected] ==>PHARMASERV ALLIANCES SDN BHD ==>IDAMAN PHARMA SDN BHD ==>PHARMANIAGA LOGISTICS SDN BHD ==>PRIMABUMI SDN BHD ==>PRESTIGE PHARMA SDN BHD ==>UNISENDO SDN BHD ==>PRIMABUMI SDN BHD ==>QUALITY REPUTATION SDN BHD ==>AVERROES PHARMACEUTICALS SDN BHD ==>PRIMABUMI SDN BHD .....
Graph Traversals Traversing Graphs with Gremlin Explanation ./inE[@label=”awarded by”]/outV/outE[@label=”awarded to”][email protected] inE - incoming edges outV - outgoing vertices outE - outgoing edges inV - incoming vertices
Graph Traversals Traversing Graphs with Gremlin Explanation ./inE[@label=”awarded by”]/outV/outE[@label=”awarded to”][email protected]
Graph Traversals Traversing Graphs with Gremlin Explanation ./inE[@label=”awarded by”]/outV/outE[@label=”awarded to”][email protected] Get current object (.) (the ’KEMENTERIAN KESIHATAN’ node). Get the incoming edges labeled ”awarded by” (inE[@label=”awarded by”]). Get the outgoing vertices of those edges (outV) (the contract nodes). Get the outgoing ”awarded to” edges of the contract nodes (outE[@label=”awarded to”]). Get the incoming vertices of those edges (inV) (the business entity vertices). Get the name attributes of those vertices (@name).
Graph Visualizations Gephi Gephi Photoshop for graphs. Supports for various graph layout algorithms. Graph metrics supported - clustering coefficient. pagerank, diameter, betweeness centrality, closeness centrality File formats supported - csv, graphml, gexf etc.. http://www.gephi.org
Graph Visualizations Gephi
Mashing Up Adding External Data Sources Mashing Up Lets add shareholding data from Suruhanjaya Syarikat Malaysia (SSM) to the graph so that we can show the tenders that have been awarded to Telekom Malaysia BERHAD and any of its subsidiaries/associate companies.
Mashing Up Adding External Data Sources Connecting Telekom Malaysia Berhad and Telekom Smart School Sdn Bhd telekom = business entities["TELEKOM MALAYSIA BERHAD"] telekom smart school = business entities["TELEKOM SMART SCHOOL SDN BHD"] telekom multi media = db.node( name="TELEKOM MULTI-MEDIA SDN BHD", no="345420-H", text="TELEKOM MULTI-MEDIA SDN BHD", type="business entity") telekom.shareholder in(telekom multi media, units=1650000) telekom multi media.shareholder in(telekom smart school, units=7650000)
Mashing Up Adding External Data Sources Graph Centered at Telekom Malaysia Berhad
Mashing Up Adding External Data Sources Graph Centered at Telekom Smart School Sdn Bhd
Mashing Up Traversing to Find Direct/Indirect Awards The Traverser class AllTendersDirectIndirect(neo4j.Traversal): types = [neo4j.Incoming.awarded to, neo4j.Outgoing.shareholder in] order = neo4j.DEPTH FIRST stop = neo4j.STOP AT END OF GRAPH def isReturnable(self, position): if position["type"] == "tender": return True else: return False
Mashing Up Traversing to Find Direct/Indirect Awards Executing the Traverser and the Output Executing the Traversal Deﬁnition telekom = business entities["TELEKOM MALAYSIA BERHAD"] tenders = AllTendersDirectIndirect(telekom) for tender in tenders: print tender["no"] Output 30/2009 35/2009 8/2009 162/2009 JASA/OP/1/2009
Wrapup Making this Easier