Machine Tags

28
This work is licensed under the Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/3.0/us/ Machine Tags Dr. Harry Chen CMSC 491S/691S February 27, 2008

description

http://tinyurl.com/2flvsk

Transcript of Machine Tags

Page 1: Machine Tags

This work is licensed under the Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/3.0/us/

Machine Tags

Dr. Harry Chen

CMSC 491S/691S

February 27, 2008

Page 2: Machine Tags

Agenda

What’s Machine TagsDifferent ways to use Machine TagsDeficiency of Machine TagsMachine Tags implementation in gnizr

Page 3: Machine Tags

Traditional tags

This is how your “grandfather” uses tags

Page 4: Machine Tags

Tags are used for labeling

Page 5: Machine Tags

Tagging Model

Remember this diagram from our previous discussion?

User

Resources Tags

Page 6: Machine Tags

What else can you say about “photo-1293”?

Resources

Relates to a YouTube video(link?)

John Smith is in this picture(homepage?)

Taken on 2007/12/03 @ 10:00PM

Johnny is the photographer(homepage?)

Relates to a family trip(trip’s homepage?)

Location Piccadilly Square(latitude/longitude?)

Page 7: Machine Tags

Describing and relating resources

The amount of user-generated content is growing on the Web (flickr, youtube, last.fm, upcoming, etc.).

New problem How to describe “ad-hoc” properties about the

resources that we have created How to relate multiple resources that we have

created across multiple Web sites

What’s an “ad-hoc” property?

Page 8: Machine Tags

Machine Tags

Tags with a special syntax that allow users to describe “ad-hoc” properties and relate distinctive resources across multiple Web sites.

+ =

http://www.flickr.com/groups/api/discuss/72157594497877875/

Page 9: Machine Tags

Triple Model

Machine Tag syntax is based on the Triple model, similar to which of the RDF language.

Subject

Predicate(Property) Object

(Value)

JohnnyPhoto-1923

takenBy

Page 10: Machine Tags

Machine Tag Syntax

There is no one standard syntax. Different systems may have variations of the original syntax proposed by Flickr.

[namespace] : [predicate] = [object]

Subject

Predicate(Property) Object

(Value)

Where is the “Subject” in the syntax?

Page 11: Machine Tags

Machine Tag example

Resource

Predicate

Object

Johnny SmithPhoto-1923

foo:takenBy

[namespace] : [predicate] = [object]

foo : takenBy = “Johnny Smith”

Page 12: Machine Tags

Describing location information

Subject

Predicate

Object

Photo-1923

geo:locality

Piccadilly Square

[namespace] : [predicate] = [object]

geo:locality=“piccadilly square”

Page 13: Machine Tags

Relating resources

What if I want to say “photo-1923” relates to “event-1393” on upcoming.yahoo.com…

Subject

Predicate

Object

Photo-1923

upcoming:event

http://upcoming.yahoo.com/event/1393

upcoming:event=1393

Page 14: Machine Tags

I machine-tagged, but what use?

Machine Tags can be used for Geotagging: putting your photos on a map Field-based search: search with prefix URL shortcuts: include image and video in your

blogs without much HTML editing Semantic Web experiments: describe properties

that link to RDF or OWL documents …

What’s your experience with Machine Tags?

Page 15: Machine Tags

Interesting use of Machine Tags

What camera did u use?

About the airplane

Page 16: Machine Tags

Machine Tags for selling books

http://adactio.com/journal/1274/

amazon:asin=1234567890

book:isbn=1234567890

http://learningtheworld.eu/2007/amazon-machine-tags/

Page 17: Machine Tags

Relating Flickr photos with a blog post

http://adactio.com/journal/1274/

Collective intelligence?

Page 18: Machine Tags

Machine Tag for field-based search

There are overlaps between the Machine Tag syntax and field-based search.

Benefit: learn one syntax and use it in two different ways

Page 19: Machine Tags

Machine Tag for building taxonomy?

Google Code allows administrator to define “Issue Labels”.

Page 20: Machine Tags

Relating resources to RDF/OWL

Geonames.ORG

http://sws.geonames.org/5352844/about.rdf

geonames:feature=5352844

http://www.geospatialsemanticweb.com/2007/03/28/geonames-machine-tags

Page 21: Machine Tags

Shortcomings of the Machine Tags

Lack of control vocabulary Who says you should or should not use what tags? Do you use “geonames:feature” or “gn:f=”

Difficult to express “object” of a complex value geo:lat/lng=23.00,39.23 (ugly, parser is need) geo:lat=23.00, geo:lng=39.23 (okay) What if I want to express multiple coordinates: geo:lat=23,00,

geo:lng=39.23, geo:lat=33.00, geo:lng=-30.00, ... (How do I interpret the key/value pairs?)

Difficult to describe certain semantic information we will see an example.

Page 22: Machine Tags

Where is your kid?

Tagged: - person=john- person=mary- person=david- landmark=power_station

Who is John?

Power Station?

Page 23: Machine Tags

Machine Tags in Gnizr

How “tags” are storedWhat does gnizr do when a “machine tag”

is detectedHow to add new implementation to

process new types of “machine tag”

Page 24: Machine Tags

DB Table: “tag”

id tag count

Page 25: Machine Tags

BookmarkManager

BookmarkManager

GeonamesTagListener

ForUserListener

geonames:[location]

for:[username]

bookmark

added

updateddeleted

notify

notify

Page 26: Machine Tags

GeonamesTagListener.java

Find all “machine tags” of gn:geonames=[location]

Get the latitude and longitude pair for[location] from Geonames.org.

Add POINT(lng,lat) about a bookmark to the DB.

Page 27: Machine Tags

Listen for new Machine Tags

Extend from BookmarkListener

Add new listeners to the BookmarkManager

Page 28: Machine Tags

Summary

Machine Tags are tags with a special syntax that allow users to describe “ad-hoc” properties and relate distinctive resources across multiple Web sites.

Use in geotagging, field-based search, foring URL shortcuts and experimenting with Semantic Web technology

Machine Tags have certain shortcomings in expressing rich semantic information.