�ݺ�ߣ

distilling the Web of Data
drop by drop (with Java)

Sourcesense UK ��Last Wednesday�� - Davide Palmisano @dpalmisano

Wednesday, June 29, 2011

the shortest introduction
ever to the Web o f Data

Web pages markup technologies are
intended for human consumption

they let machines to present raw
data to humans

extracting valuable data may
require fancy scraping techniques

scraping: one size doesn��t fit all



<div>
<div> Canon Rebel T2i (EOS 550D) $899 </div>
<div> The Rebel T2i EOS 550D is Cannon's
top-of-the-line consumer digital SLR
camera. It can shoot up

<div> AN_UCC-13: 013803123784 </div>
<div> price: 899 USD </div>

</div>
</div>



<div>

<div> AN_UCC-13: 013803123784 </div>
what does this
</div> is this a tag mean?
</div> currency or what?


��meaning��, Joseph Kosuth, The Panza Collection, MART - Rovereto, Italy


Microformats

��Microformats are a way of adding simple markup
to human-readable data items such as events,
contact details or locations, on web pages��
Andy Mabbett

- community driven initiative
- largely adopted
- quick & dirty
- scarcely extensibility


Microformats

<div class=��hlisting item��>
<div> Canon Rebel T2i (EOS 550D) $899< /div>
<div class=��description��> The Rebel T2i EOS
550D is Cannon's
<div> AN_UCC-13: 013803123784 </div>
<div class=��price��> price: 899 USD </div>
</div>
</div>


RDFa: RDF in attribute

model your data as they were Web pages
connected with named links and properties
and embed them in your (X)HTML using
@attributes

- RDF, graph-based model
- W3C Recommandation
- highly extensible

i.e GoodRelations[1], a fully flavored
vocabulary for the e-commerce



model your data

http://mystore.com/product/5642

ex:price ex:value 899

ex:producer
ex:currency

ex:description
USD
http://canon.co.uk

The Rebel T2i EOS
550D blah blah


and then embed them in your
(X)HTML pages
<div about=��http://mystore.com/product/5642��>
<div>Canon Rebel T2i (EOS 550D) $899</div>
<div property=��gr:description��>The Rebel T2i EOS 550D
is Cannon's blah blah</div>

<div rel=��gr:hasPriceSpecification��>
<span> price:
<span property=��gr:hasCurrencyValue��>899</span>
<span property=��gr:hasCurrency��>USD</span>
</span>
</div>
</div>

HTML5: Microdata

Microdata allows nested groups of name-value
pairs to be added to HTML documents, in
parallel with the existing content

- W3C Working draft
- native of HTML5 specification
- serializable in RDF

- Google, Yahoo! and Bing endorsed Schema.org
- large adoption expected


HTML5: Microdata

<div itemscop itemtype=��http://schema.org/Offer��>
<div itemprop=��name��> Canon Rebel T2i (EOS 550D) $899
</div>
<div itemprop=��description��> The Rebel T2i EOS 550D
is Cannon's blah blah</div>

<div>
<span> price:
<span itemprop=��price��> 899 </span>
<span itemprop=��priceCurrency��> USD </span>
</span>
</div>
</div>


% of marked up Web pages

3.5

3

2.5

2

1.5

1

RDFa 0.5
hCard
adr
09/2008 xfn 0
03/2009 hReview
10/2010
data from Yahoo! [2]


tie ��em all together

uniform, reconciled and
unified RDF representation


a drop-by-drop distiller

Anything To Triples (any23) is an open source,
Apache-licensed:

- Java library,
- Web service and
- a command-line tool

able to distill RDF triples from a
variety of semantically marked up Web
documents

http://developers.any23.org


live demo http://any23.org

Web site with ~5000 products description with
GoodRelations using RDFa


use Any23 in your Java
programs
Any23 runner = new Any23();
runner.setHTTPUserAgent("test-user-agent");
HTTPClient httpClient = runner.getHTTPClient();
DocumentSource source = new HTTPDocumentSource(
? ? ?httpClient,
? ? ?"http://test.com/index.html"
? );
ByteArrayOutputStream out = new
ByteArrayOutputStream();
TripleHandler handler = new NTriplesWriter(out);
runner.extract(source, handler);
String n3 = out.toString("UTF-8");


Any23: Command-Line tool
any23-core/bin$ ./any23

usage: any23 [-e <arg>] [-f <arg>] [-l <arg>] [-n] [-o <arg>]
? ? ? ?[-p] [-s] [-t] [-v] {<url>|<file>}
?-e <arg> ? ? ? ? ? ?comma-separated list of extractors, e.g.
? ? ? ? ? ? ? ? ? ? ?rdf-xml,rdf-turtle
?-f,--format <arg> ? Output format [turtle (default),
ntriples, rdfxml, quad, uris]
?-l,--log <arg> ? ? ?logging, please specify a file
?-n,--nesting ? ? ? ?disable production of nesting triples
?-o,--output <arg> ? ouput file (defaults to stdout)
?-p,--pedantic ? ? ? validates and fixes HTML content
detecting commons issues
?-s,--stats ? ? ? ? ?print out statistics of Any23


Any23: Web Service
blacky:~ davide$ curl http://any23.org/any23/?format=nquads&url=http://
www.bbc.co.uk/programmes/b00kygwh&report=on

<response>
<extractors>
<extractor>rdf-xml</extractor>
</extractors>
<report>
...
<validationReport>
<ruleActivations></ruleActivations>
...
</validationReport>
</report>
<data>
<rdf:Description rdf:about="http://www.bbc.co.uk/programmes/
b00kygwh#programme">
<rdf:type rdf:resource="http://purl.org/ontology/po/Episode"/>
<po:pid>b00kygwh</po:pid>
<dc:title>The Terminator</dc:title>
</rdf:Description>
</data>
</response>

Apache Tika
mimetype detection

Cyber Neko HTML
DOM extraction

Rule Fix
Validator

Microdata RDFa hListing hReview hCalendar hCard
Extractor Extractor

Microformat Extractors

Sesame RDF/XML NQuads JSON
Writer Writer Writer
ExtractionResult


extractor
public interface Extractor<Input> {

/**
* Executes the extractor. Will be invoked only once, extractors are
* not reusable.
*
* @param in The extractor's input
* @param documentURI The document's URI
* @param out Sink for extracted data
* @throws IOException On error while reading from the input stream
* @throws ExtractionException On other error, such as parse errors
*/
void run(Input in, URI documentURI, ExtractionResult out)
throws IOException, ExtractionException;

/**
* Returns a {@link org.deri.any23.extractor.ExtractorDescription} of
* this extractor.
*/
ExtractorDescription getDescription();

}


validate and fix
public interface Rule {

String getHRName();

boolean applyOn(
DOMDocument document,
RuleContext context,
ValidationReportBuilder validationReportBuilder
);
}

public interface Fix {

String getHRName();

void execute(Rule rule, RuleContext context, DOMDocument document);

}

void addRule(Class<? extends Rule> rule, Class<? extends Fix> fix);


plugins
@PluginImplementation
? @Author(name="Michele Mostarda (mostarda@fbk.eu)")
? public class HTMLScraperPlugin implements ExtractorPlugin {

? ? private static final Logger logger =
LoggerFactory.getLogger(HTMLScraperPlugin.class);

? ? @Init
? ? public void init() {
? ? ? ? logger.info("Plugin initialization.");
? ? }

? ? @Shutdown
? ? public void shutdown() {
? ? ? ? logger.info("Plugin shutdown.");
? ? }

? ? public ExtractorFactory getExtractorFactory() {
? ? ? ? return HTMLScraperExtractor.factory;
? ? }

? }


roadmap
incoming 0.6.0 release
- support for Microdata
- support for CSV
- support for RDFa 1.1 prefix mechanism
- improved app configuration
- bug fixing

Apache (pre) Incubation process
- http://wiki.apache.org/incubator/Any23Proposal
- supporters and mentors (thanks guys!)
Simone Tripodi (@stripodi)
Tommaso Teofili (@tteofili)
- we��re looking for mentors


closing credits

active committers

Giovanni Tummarello ( @jccq )
Michele Mostarda ( @micmos )
Davide Palmisano ( @dpalmisano )
Richard Cyganiak ( @cygri )

thanks to the whole Semantic Web community,
especially those who tirelessly challenge us
with bugs and features requests


References

[1] http://purl.org/goodrelations/v1

[2] http://tripletalk.wordpress.com/2011/01/25/
rdfa-deployment-across-the-web/


�ݺ�ߣ

distilling the Web of Data drop by drop (with Java)

Recommended

More Related Content

Similar to distilling the Web of Data drop by drop (with Java) (20)

More from Davide Palmisano (6)

Recently uploaded (20)

distilling the Web of Data drop by drop (with Java)