Google Blogoscoped


Google News Cloud  (View post)

/pd [PersonRank 10]

Wednesday, May 10, 2006
17 years ago3,937 views

I like the collation of the various assositative tags ... Tag clouds are becoming meta data viewpoints.

Erin O'Brien Owner's Manual [PersonRank 1]

17 years ago #

The cloud represents our virtual lightness of being.

This is the truth and the sound of falling water.

Splasho [PersonRank 10]

17 years ago #

Even better IMO would be different shades of red for different relatednesses

/pd [PersonRank 10]

17 years ago #

Splasho : How will your define "relatedness" – I think thats the delimemma that faces all experts on the 'relevancy' subject!!

Thomas Hofmann Online [PersonRank 2]

17 years ago #

An ingenious idea! The most wonderfull thing were, if this were to do with CSS only.

Philipp Lenssen [PersonRank 10]

17 years ago #

I would define tag relatedness as significant overlap of neighboring tags.
For example when story 1 has these tags:
____ A B
and story 2 has these
____ B C

____ President Bush
____ Bush on Iran

then we can assume that A and C have a relation. The more of those matches, the stronger the relation factor – this way we could match two different words which are yet related, like e.g. "bush" and "iran". We could then use a treshold of N to use only the top N (like 4) hits found in the data structure...

I'm not sure I did the same at , though we do have a "Related module" feature – take a look at this page and the link below the AdSense:

/pd [PersonRank 10]

17 years ago #

wiat a second Philipp!!

what is the relevnace between A and C ? just because of the common denomintor of tag "B" ?? The issues here is ;

Story (A) = > tag "bush", "iran"
Story(B) = > tag "iran", "china"

What would be the relevance to the two stories in terms of

Story(a) where Bush and the amercian doctrine of Iran
Story)b) Where Iran is importing rice from China

So the dilemma-- " What I would find useful is something that helps me find relevant new content – content related to what I actually read. "

in this case, {"Iran and Bush"} and not {Iran and rice from china}

Niraj Sanghvi [PersonRank 10]

17 years ago #

/pd: As Philipp mentioned, "The more of those matches, the stronger the relation factor"

You may get a few false positives like you mentioned, but you'll get many more valid combinations.

1200 matches "bush" and "iran"
800 matches "bush" and "iraq"
50 matches "bush" and "china"

By adjusting your threshold, you will throw out the junk matches and only save ones that appeared often across stories. (in the simple example above, the threshold would be 2)

/pd [PersonRank 10]

17 years ago #

yeah from a mathemtical function, this is ok. Its pattern matching techniques. Hwoever- the issue is "relevancy" based on "tags". No two people use tags in the same way..!!

Hashim [PersonRank 10]

17 years ago #

I love this! This should be what all tag clouds makers should incorporate.

Forum home


Blog  |  Forum     more >> Archive | Feed | Google's blogs | About


This site unofficially covers Google™ and more with some rights reserved. Join our forum!