Semi-structured data
and the Web
COMP 60372

Uli Sattler and Bijan Parsia

{sattler•bparsia@cs.man.ac.uk}

Goals of the course

The Web is

a very large (by several metrics) information structure

Indexable Web estimated size (Jan 2005): 11.5 billion pages

a complex, distributed, often interactive computing platform

and very large

How do we put information on it?
This course will acquaint you with the representation of data on (or inspired by) the Web

semi-structured data
XML, HTML, and RDF/OWL
data, documents, and (Web) applications

Newish Course!

The XML part has been taught for two years
The HTML/CSS part was taught last year
The RDF & OWL stuff is new (but has been taught before elsewhere)
In both areas, change is rapid

Change comes from the top (from big corps and standards bodies)
Change comes from the bottom (from small communities or individuals)
Separate hype and flash from fundamentals

Who are we?

Uli Sattler

Professor in the IMG group
Developed key algorithms which made practical the logics underlying the Web Ontology Language (OWL); member of OWL working group

Bijan Parsia

Lecturer in the IMG group
Participated in several Web standardization efforts, e.g., SPARQL, OWL 2, WSDL, WS-Policy

Data on the Web

Databases are designed to store relations/tuples w.r.t. a schema

follows a client/server architecture

Often data on the Web

does not come in tuples, but often in “documents” with less structure: semi-structured

is shared:

needs to be self-describing

to cope with uncoordinated reuse, evolution, etc.

in a multi-tier and decentralized architecture

Semi-structured Data

A semi-structured data format:

partially self-describing, can be enhanced by schemas

person:

    {name: {first: "Uli", last: "Sattler"},
   tel: 56176,
   office: “Rm 2.24”},

person:
    {name: {first: "Bijan", middle: "John"},
   tel: 56199,
   tel: 50690,
   office: “2.32”}

How do we

describe the commonality
cope with the variation
prescribe certain aspects

without overprescribing!

SSD and XML

XML is a W3C standard for data exchange on the Web

based on embedded markup <b>..</b> to indicate structure

a textual representation of data
e.g., <person> <name><first>Uli</first> <last>Sattler</last></name> <tel> 56176</tel> <email>sattler@cs.man.ac.uk</tel>
</person>

Schema Langauges for XML

XML schema languages support structural constraints e.g.,

a <tel> contains a number
a <person> may contain one or more <tel>s

Like a grammar for XML documents
We can validate an XML document against a schema

i.e., check the (loose) structure against a schema (or several schemas)
need to describe tree structures with "holes" and "extra stuff"

We discuss XML schema languages and their validation problems (automata-based):

DTDs

XML Schema & RELAX NG

Query Languages

We want to retrieve data encoded in XML

Unlike RDBMSs, the data are:

tree shaped
irregular

Path queries common

Supported by XPath

We also want to process XML

XSLT, a transformation langauge
XQuery, general purpose, query oriented language

We also discuss tree algebras for query languages

(X)HTML and (X)HTML 5

HTML, the language of the Web

A hypertext markup language
Originally, page oriented and static; informal description
XHTML...the XMLizaiton of HTML 4
XHTML 2, a break from the past

HTML 5

Dump the XHTML approach
Key Goal: support for Web applications
Key constraint: Don't break the Web
DOM based language

Consistent Javascript behavior

CSS

Markup languages: Separate structure/semantics and style

But what about the style?

Cascading Style Sheets

Make documents and data presentable to people

An Alternative Data Language

XML is

...tree oriented
...has no notion of objects
...is a bit distinct from the Web

The RDF alternative:

Graphs (the Web is a Graph)
Links everywhere (URIs down to the bottom)
Triples: reified web links

How to integrate with the HTML Web?

Logic Based Approaches

Until now, we've taken a data structure view

We are talking about semi-structured data after all
However, SSD aims for more flexible models

I.e., models closer to the domain

Logic based ontology languages support this

OWL (the Web Ontology Language)

Represent "knowledge"
Richer models of your domain

and thus your data

Inference services

Good segue to COMP60462

Outcomes

A firm grounding in an array of SSD languages and supporting technologies

Make sense of the alphabet soup!
Understanding of the underlying, unifying, and contrasting principles
Able to assess applicability, utility, and tradeoffs

Acquaintance with applications

Designing formats and data
Designing for extensibility/usability
Examine some existing formats (e.g., SVG, XHTML, MathML)

Understanding of the formal principles

Grammar & automata theory
Tree Algebras
Logics

Prerequisites

Good familiarity with relational database theory
Some familiarity with grammars and automata
Reasonable familiarity with first order logic
You all have written web pages, right?
This module will look at topics in a rigorous and formal way

Practicalities

Period 3

Jan. 29 - March 2, 2009
Mondays, 9:00 - 17:00

Assessment

50% exam

Covers conceptual issues, simple problems

50% course work

Encode data into various formats
Transform, query, style it
Some analysis

Lectures & Topics

Introduction: Semi-structured data.
XML: core concepts

DTDs, a simple schema language for XML

XPath, a navigation language for XML

XML namespaces: a concept ignored so far

XSLT, a transformation language for XML
DOM and SAX, APIs for XML
XML Schema and RELAX NG, more expressive schema languages
XQuery, a query language for XML
Tree Algebras for XML

Lectures & Topics (cont.)

HTML 5: A new language for the web

text/html vs. application/xml+xhtml

Validation of HTML 5

Practical validation (including Schematron and code)

CSS and the DOM

Web Data vs. Web Documents vs Web Applications
Styling and transforming

RDF: A graph language for Web data (and a bit of knowledge)

SPARQL: A query language
GRDDL & RDFa: HTML integration
"Linked Data"

OWL and OWL 2: How inference can help data