Up
TagFinder
Sites
Sign-Up
Demo
Contact

 

 

What is the Zingo TagFinder?

Why is Zingo TagFinder good?

Current Status & Demo site

How does Zingo TagFinder compare to similar services?

The Zingo TagFinder Service API

How does it work?

Credits

 

What is the Zingo TagFinder?

The Zingo TagFinder is a Web service that automatically extracts tags from a piece of text.  The tags are chosen based on both statistical and linguistic analysis of the original text. For example, given the  text:

Three more countries said on Tuesday they had detected cases of deadly bird flu in wild swans, with Germany, Iran and Austria the latest to find the virus that has killed 91 people worldwide.

Zingo TagFinder returns the following tags:

bird flu wild
deadly bird flu
austria
germany
virus
wild

^ back to top ^

Why is Zingo TagFinder good?

A couple of months ago, I have created Newzingo.com - a web site that displays a tag cloud of current news stories from Google News. Newzingo uses Yahoo! ’s Term Extraction service to extract tags from news stories. When trying to enhance Newzingo, I came across several limitations imposed by Yahoo’s TES:
  • Single language support - English only .
  • Licensing – non-commercial license only
  • Rate limiting -license allows a maximum 5,000 requests a day
  • Poor results in the long tail – news stories that contain non frequent terms were getting non relevant tags.
  • Non learning system.

Zingo TagFinder is built to overcome all these limitations.

^ back to top ^

 

Current Status & Demo site

Zingo TagFinder is in early development stages and is currently in use only on newzingo.com.

Live demonstration

On http://zingosoft.com/tagger/term/extract you'll find a simple form that lets one experiment with the Zingo TagFinder index for the Newzingo tagging service. Enter any news related body of texts and let Zingo TagFinder tag it for you.

The Zingo TagFinder API

currently in closed beta testing. In order to paricipate in the API beat program, please send an e-mail to guy.tavor@zingosoft.com.

^ back to top ^

How does Zingo TagFinder compare to similar services?

Curently there are two term extraction services online: Yahoo!'s term extraction service, and TagyU

Below is a comparison between the three services. The comparison was created automatically. There is no “human editing” envolved. MSNBC news stories were used for this comparison.

Zingo TagFinder

Yahoo!

TagyU

Iraq death squad claims probed.

The Iraqi Interior Ministry has launched an investigation into claims that a police death squad has been operating in the country, an official said Thursday. Meanwhile, attacks around the country killed 10 people.

death squad
investigation
country
ministry
police death squad
iraqi interior ministry
has launched investigation
death

 

death squad
interior ministry
iraq

Iraq

collateral

war

politics

iyad allawi

 

Note:  Yahoo! Gives the best results here.  TagyU gives very poor results and TagFinder is second.  We are currently working on improving this by eliminating irrelevant results (has launched investigation, country, police death squad).

 

Anger over new Iraq abuse photos New images of naked prisoners, some bloodied and lying on the floor, threatened to revive public anger over abuse by U.S. guards at Abu Ghraib prison at a time when tensions with the West are already running high in the Middle East.

abu ghraib prison
anger
floor
prison time
middle east
revive
prisoners
iraq abuse
running high
already running
abuse photos

 

public anger
naked prisoners
abu ghraib prison
lying
middle east
iraq
running
images
photos
abu ghraib

northview

website

prison

study

msphere

phychology

nature

humans

humanrights

Note:  Here the best results are provided by TagFinder and Yahoo! TagFinder has slightly better relevance scores.

 

Pakistan keeps protest in check

Thousands of Pakistanis  shouting burned effigies of the Danish prime minister in the country’s fourth day of protests over cartoons of Prophet Muhammad.

danish prime minister
effigies
pakistan
protest
keeps
cartoons prophet muhammad
prophet muhammad

danish prime minister
prophet muhammad
effigies
pakistanis
protests
cartoons
pakistan
prime minister

religion

goodnewssites

politics

Note: Yahoo! And TagFinder are equivalent, tagyU is bad.

 

Stocks rise after strong Hewlett-Packard profit

hewlett packard
profit
stocks

hewlett packard
stocks

programming

scheme

video

reference

sicp

lisp

Note: Yahoo! And TagFinder are equivalent.

 

MySpace users to get their own cell phones

The social networking site MySpace, hugely successful among teenagers and twenty-somethings, is about to become more ubiquitous with the launch of a cellular service that will let users read and post to the site for free.

myspace
twenty
launch
cellular service
social networking
let users

 

cellular service
social networking
myspace
cell phones
ubiquitous
teenagers
service

social

myspace

wow

blog

guild

networking

friends

community

^ back to top ^

The Zingo TagFinder Service API

Zingo TagFinder implements a REST interface for easy programatical usage. In order to use it, you must obtain a developer_id.

SOAP and XML-RPC interfaces are available on request.

Request

•     Term extraction requests must be submitted using POST protocol.

•     http://zingosoft.com/tagger/api/v1/term/extract

•     Parameters

 

Parameter

Type

Notes

Description

developer_id

string

required

A unique user identifier issued on the Zingo TagFinder site on registration

title

string

optional

The title of the document. (UTF-8)

text

string

required

text for term extraction, UTF-8 encoded

 

Response

Response is an XML with the following structure:

XML tag

Description

ResultSet

Contains all results

Result

A single result

 

Example Response

<?xml version="1.0" encoding="UTF-8"?>

<ResultSet>

  <Result>george w bush</Result>

  <Result>desert storm</Result>

  <Result>gulf war</Result>

</ResultSet>

Errors

HTTP CODE

Description

200

Everything is okay. Result contains XML with tags (see above)

500

Text or developer_id are mandatory but were not supplied

403

The developer_id that was supplied is unauthorized

503

Developer must contact ZingoSoft for more details

How does it work?

Zingo TagFinder's engine uses a crawler that builds fast-access index for runtime automated tag extraction. In runtime mode, the system is doing just one thing: extracting tags.

In training mode, the system reads a given body of documents and runs various statistical and linguistics algorithms to create a preliminary partition of corpus-space into tags.

This “training process” creates a database that is later used in runtime-mode to extract terms.

For you own application, you could either use a “web-wide” index – or a domain specific index tailored for your needs. The current demo database for Newzingo is an example of such an index, which contains only news stories.

^ back to top ^

Credits

The ZingoTagger engine was developed by Guy Tavor based on advice and guidance from Frank Smadja Phd, and Grigory Begelman Phd student, Both provided help and support on various stages of this version.
^ back to top ^

Home | TagFinder | Sites | Sign-Up | Demo | Contact
(C) Copyright ZingoSoft and Guy Tavor. All rights reserved.
Last updated: 02/26/06.