I think I agree with your Yahoo example in that you take the post and
remove common words. What your left with is a selection of words which
you can then add to the users profile. You can priorities capitalised
words and keep a count of the number of times the word is used. In a
perfect world you could then accociate the users account with various
posts which match his/hers profile, even emphasizing the words which
they have entered more often. But as you say this requires volume and
which is alright in a limited forum but with a lot of potential users it
could be BIG! One possible counter to the large data volume is to
remove older keywords and only keep the more recent ones.. I think
whats there is cool though because it isnt about the "user", its about
the board so its a different type of tagging.
The mightyv.com (http://www.mightyv.com/alltags/) is a perfect example
of where tagging is really useful because the data is limited and your
audience is already tailored to a specific genre. For me "direct users
to multiple threads that they might be interested in (Jem's post)" means
direct them based upon something we know about them as opposed to here's
a load of tags which people are talking about and hopefully they'll find
something they like.
If your post was a message on the board we could see some good profiling
and potentially link you to some other related posts even though we have
only one posting from you. I suspect also that the following words would
grab your attention more than "smeg" :)
tag - 6 times
Yahoo - 4 times (capitalised)
API - 1 time (capitalised)
It generally falls down when you consider the other words in the post
which are difficult to remove all the time.
Best
Jim.
Tom Loosemore wrote:
I like the personalised tagging concept, but you need serious tag volume
for it to work - and generating serious user-submitted tag volumes from
message board postings might prove hard. What's the end-user benefit to
tagging a message board posting? They're not usually things you want to
go back and find later (unlike photos or links).
I'm intrigued by auto-tag generation using term extraction software or
APIs such as Yahoo's
http://developer.yahoo.net/search/content/V1/termExtraction.html , or
http://www.tagyu.com
I've used the Yahoo Term extractor API on an internal project to
auto-generate tags from programme descriptions, and it worked a treat on
such keyword-dense text (see http://www.tomski.com/yahoo_demo.html )
You need to prune the tags to get rid of noise (such as "programme", in
my example), but it's quick, dirty and seems to work OK. Much Better
Than Nothing. You can always let users augment auto-generated tags with
their own.
I bet running a term extractor over the England message bodies would tag
a discussion rather well, provided you tuned & pruned tags at the
beginning... And would it might scale better in terms of cost. Hmm, just
tried a really clumsy experiment with a thread on the Kent board
http://www.tomski.com/yahoo_bbc_demo.html
Could be worse... I quite fancy seeing all messages posted on
bbc.co.uk/england which are tagged "smeg"...
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of James
Sent: 01 November 2005 12:00
To: [email protected]
Subject: Re: [backstage] Programme Archive/Message Board beta
Would it not be more effective by basing the results on what the
individual has posted about before ? The same tagging concept could
apply but then the results are limited to the tags found in what the
user has shown as a previous interest. It's difficult to predict
interest just based upon popularity of tags as I imagine there is going
to be a huge user base with diverse interests.
Design wise: For me it's too tight. The main content body is too small
and the amount of scrolling is going to be huge.
Jim
Jem Stone wrote:
The challenge they were trying to solve was how to direct users to
multiple threads that they might be interested in across nearly 50
message boards up and down (in this case) the English regions and
cities. We're hoping that thread/topic tagging might be the way to go.
-
Sent via the backstage.bbc.co.uk discussion group. To unsubscribe,
please visit
http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Unofficial list archive:
http://www.mail-archive.com/[email protected]/
-
Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please
visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Unofficial list archive: http://www.mail-archive.com/[email protected]/
-
Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please
visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
Unofficial list archive: http://www.mail-archive.com/[email protected]/
