exhibitlightweight structured data publishing

david huynh + david karger + rob miller

MIT COMPUTER SCIENCE AND ARTIFICIAL INTELLIGENCE LABORATORY

2

good ol’ days ... early 1990s

3

sort

filter

search

4

5

early 1990s → 2007

6

PRESENTATIONHTML

Web Browser

File System

Static Files

Web Server

ImagesJavascript CSS

7

PRESENTATION

DATA

HTML

Web Browser

Database

File System

Static Files

Web Server

Images

MySQL / Postgres / Oracle

Javascript CSS

8

PRESENTATION

LOGIC

DATA

HTMLJavascript CSS

XML XSLT

SQL

Web Browser

XmlHttp

Database

File System

Static Files

Application Server Web Server

ASP

ASP.NETCGI

JSP/Java

PHP

Images

MySQL / Postgres / Oracle

9

publishing data is hard.

10

can Semantic Web technologies help?

11

duh!

obviously!

SW technologies are supposed to help!

12

what people want

what SW lets them do

sortfilter

search

13

14

outline✓problem: publishing data is too hard✓demo: using Exhibit to publish data in 10 min

•implementation: how Exhibit works•real world uses + discussion•related work•future work•conclusion

15

PRESENTATION

LOGIC

DATA

HTMLJavascript CSS

XML XSLT

SQL

Web Browser

XmlHttp

Database

File Systems

Static Files

Application Server Web Server

ASP

ASP.NETCGI

JSP/Java

PHP

Images

MySQL / Postgres / Oracle

16

PRESENTATION

LOGIC

DATA

HTMLJavascript CSS

XML XSLT

SQL

Web Browser

XmlHttp

Database

File Systems

Static Files

Application Server Web Server

ASP

ASP.NETCGI

JSP/Java

PHP

Images

MySQL / Postgres / Oracle

17

PRESENTATION

LOGIC

DATA

HTML

Javascript CSS

Web Browser

XmlHttp

File Systems

Static Files

Web Server

Images

Exhibit API

18

data

Exhibit API

database

expression languagelocalization

imagescss

viewslens

templatefacets exporters

importers

HTML+

Images+

CSS+JS

web browser

HTML

JS

DOM dataexports

<script src= “...... /exhibit-api.js”></script>

19

data

Exhibit API

web browser

presentation

sorting filtering maps

timelines

my users

20

•JSON as default format

•http:// simile . mit . edu / babel /•Bibtex•Excel spreadsheets•Tab separated values•RDF/XML, N3

•Dynamic importers

data formats

JSON files

21

22

JSONP data feedgdata.io.handleScriptLoaded({ ... "entry": [ { "id":{ "$t":"http://spreadsheets.google.com/feeds/list/.../od6/public/basic/cokwr" }, "updated":{"$t":"2007-04-16T18:41:56.378Z"}, "category":[ { "scheme":"http://schemas.google.com/spreadsheets/2006", "term":"http://schemas.google.com/spreadsheets/2006#list" } ], "title": { "type":"text", "$t":"Lord of the Rings: The Return of the King" }, "content": { "type": "text", "$t": "{type}: Movie, {genre}: Drama; Epic, {plot:single}: The former Fellowship of the Ring prepare for the final battle for Middle Earth, while Frodo \u0026 Sam approach Mount Doom to destroy the One Ring., {rating:number}: 4" }, ... }, ... ]})

23

•Javascript is slow, not designed for implementing DBs

•Recommended for < 500 items•Some people have been brave: 2733

items or more

•Not a limitation per se•Exhibit is intended for small data sets

scalability

24

outline✓problem: publishing data is too hard✓demo: using Exhibit to publish data in 10 min

✓implementation: how Exhibit works•real world uses + discussion•related work•future work•conclusion

25

26

27

28

29

30

31

32

33

oops!

34

35

36

37

38

39

someone is planning a wedding using Exhibit

40

presentationscompany members

software toolsrestaurants3 recipes

radio albumsinstalled fonts

hotels near a dance eventdogs for adoption

lego setsdances, costumes, performances

breweries and distillerieskansai dialect field study data

world conflictswedding attendees

41

presentationscompany members

software toolsrestaurants3 recipes

radio albumsinstalled fonts

hotels near a dance eventdogs for adoption

lego setsdances, costumes, performances

breweries and distillerieskansai dialect field study data

world conflictswedding attendees

If Semantic Web researchers were tobuild a web site with data,

what topic would the data be about?

42

scientific papers

43

pub

licati

on

s

The Long Tail

information topics

quantity or

popularitymerchandises

moviesphotos

newsevents

software

lego setsisrael folk dance videos

breweries and distilleries

in Ontario 1914 - 1915

free laborin addition to grad students

✓ ✓dormant data publishers

44

reuse withoutscraping

have fun!

45

outline✓problem: publishing data is too hard✓demo: using Exhibit to publish data in 10 min

✓implementation: how Exhibit works✓real world uses + discussion•related work•future work•conclusion

46

HTML

Ruby on Rails

Flickr

Google BaseDabbleDBFreeBase

customizedSemantic MediaWiki extension

wiki, blog

Semantic MediaWiki extension

circle size = amount of effort

flexibility of presentation

flexibility of data

modeling

Related Work

custom 3-tier web app

Exhibit

47

Exhibit

personal

Semantic MediaWiki extension

group world

Freebase

data ownership

personal blogpersonal web space

wiki Wikipedia

unstructured

structured

Related Work

DBPediaYAGO

DabbleDB Google Base

48

•database in Javascript•TimBL’s Tabulator

•generic browsing interface•for data consumers to do mash-up

•Exhibit•customizable publishing framework

•for data publishers

related work

49

•feature requests•more views: calendar, histogram, ...•more flexible layouts•visual synchronization, e.g., color

coding•value formats, e.g., $(6,000)•localization

•if there will be a lot of exhibits, let people...•search over them•merge them together

future work

50

•authoring interface•HTML got us so far...•WYSIWYG editors got us further

•Exhibit will get us so far...•A front-end to Exhibit will get us

further

future work

51

conclusion•many dormant data publishers in the long tail

•... with few resources to publish data

•Exhibit•answer real world needs of publishing data

•as easy and expressive as HTML•tap the free labor in the long tail

•produce data that doesn’t have to be scraped

•build a Data Web representative of the Web

52

google for “exhibit”