Guy Berger
(Revised version of presentation
given at conference on
Journalism Education & Training:
The Challenges
17 October 2008, Stellenbosch)
In an ever-crowded online
universe, the distinctiveness of a
given journalistic product can be
dwarfed or even missed completely
 Especially because much online
consumption is search, rather than
brand, driven. Also: think RSS.
Find/Populate content in trusted places
Especially where your target people go…
But don’t tread on toes.
Seize the moment – topicality talks.
Become a brand for “quality” – trust is key.
Link to others as well – send traffic away.
Cross promote and cross platform (Twitter,
blogs, Facebook …).
But even then – the web is awash….
David Weinberger perceptively
argues that the solution to an
environment is more
Social network recommendations
surface content. Low-level ok.
 But critical importance of
understanding & wielding meta-data
 You’ve seen tag clouds – that’s just
one functionality.
Internet -
cf. blogs,
delicious, etc
I wanted pictures of an opening door.
Searched on Flickr for “door, ajar”
That depended on the image having been
captioned with those words.
I decided I wanted to narrow it to a glass
That depended on whether the
photographer thought it significant enough to
mention that the door was a glass one.
“Bird” = midlevel abstraction
“Male frigate bird” = lowerlevel
Do you note “blue” sky & “silhouette”?
More abstract like “flight”? “freedom”?
Synonyms?: “flying” / “in the air” / “aloft”
or “liberty”?
 Google score: sky (50 points), bird (60 points),
soaring (120 points), or frigate bird (150
points). Well, it depends…
 Inter-coder reliability is the same issue…
 Audio,
Video, Flash….
Check out:
Google: Suggest and Adwords’ Keyword Traffic
Estimator Tool and Trends
Microsoft: AdCenter Keyword Forecast Tool
WordTracker: Basic Keyword Suggestion Tool
Keyword Discovery: Basic Search Term
Suggestion Tool
Not so good: <title>Home</title>
Better: <title>Webmaster Central home page</title>
<title>Webmaster Central home page | Search
engine tips and tools for webmasters</title>
To be avoided (your site may be seen as spam):
<title>Webmaster Central seo optimization
search engine search engine google websearch
google searchresults improve search results seo
optimize search searching serps</title>
Google also understands:
 <meta name=“description”
content=“the keywords / phrases /
text that is inserted here can also
show up sometimes as part of the
snippet shown in search results”>
refers to the practice of loading a
webpage with keywords in an attempt
to manipulate a site's ranking in
Google's search results. Filling pages
with keywords results in a negative
user experience, and can harm your
site's ranking. Focus on creating
useful, information-rich content that
uses keywords appropriately and in
Google News requires URLs have at >2 digits.
Ensure web addresses include key words.
> 5 keywords in a Page <Title> dilutes each.
Use creative headlines as anchor text for
links to the article (aka 'linkbait') - but make
actual headlines literal.
Incorporate keywords from the meta-data
in <title> into the headline and deck. Not as
easy as it sounds, but perhaps the new copy
skill for the digital age.
What would a searcher do?
What words & language wd they use?
Do you have the terms as meta-data?
In headlines and intros?
In tags and keywords?
It’s more than technical SEO…
Taxonomy: how Dewey used to work –
categories and silos.
Based on physical model of
meaning: you classify books for a shelf.
You overcome exclusivity by having physical
index cards that cater to overlaps in
Note: Cards are smaller than referents.
Resource Descriptor Framework &
Uniform Resource Identifiers for
 Digital Object Identifiers (different
to URLs) (what, vs where).
Folksonomies – untrained, nonlibrarians, giving horizontal tags.
“A folksonomy is a user-generated
classification, emerging through bottomup consensus” (Emanuele Quintarelli)
Power connected to aggregating…
Without a social distributed context, tags
are just flat keywords.
Basic idea is simply to get people to
share content annotated with tags.
You can develop a
vocabulary, performing metadatadriven queries (also using multiple tags at a
time), monitoring changes, discovering
popular metadata. (Jon Udell)
Blue sky
on past
A narrow folksonomy (as the one of
Flickr) is the result of a number of
individuals tagging (using one or more tags)
different items for later personal
Broad Folksonomy: is the result
of many people tagging the same
item for shared use.
The goal is a metadata
Emanuele Quintarelli.
An electronic catalogue “entry” can be even
bigger than the object – beyond an index
card’s limitations of space (cf Amazon).
But free text search means: every word in an
can serve as metadata – in the sense of being
searchable, but then there’s no
All taxonomy categorisation and all
folksonomy keywords helps searchability.
But even so, computers are stoopid…
 BT:
broad term microorganisms
 NT: narrow term – ecoli
 RT: related term: gut
 UF: use e-coli and ecolli for
“ecoli” (latter is preferred)
A story on a sports match might
include value-based tags like “bully”,
“macho”, or “future olympiad” .
Or analytical words like “fandom” or
“hero” or “racism in sport” or “Low GI
diet” – terms that are not in the
story as such, but which identify
some of the meanings in and of it.
De Saussure saw that the train from Paris to
Geneva was only meaningful because of the
relationship, not physical coaches.
Meaningfulness is a function of relationships:
between people, places and things.
The existing web moving from silos of content,
to googels of undifferentiated data.
We need context relations for meaningfulness.
For example, you do a story about unhappy
workers on a winefarm in Stellenbosch.
The story refers to Clay Wine Estates,
locates it outside Stellenbosch, and the
owner as Paul Clayton.
It does not give phrases like “Western
Cape” or “South Africa” – that ‘s assumed.
It does not mention the names of
wines produced at the estate.
“Stellar” and “Star” wine brands come
from CWE; labour = a synonym for
Stellenbosch is in the Western Cape, SA.
By mapping these unwritten
connections, computer intelligence can
link your story to them
A search for “labour, Western Cape,
Stellar, Star”, will then find your story…
Automates some triples linkages: M&G
did people, cities, countries, companies.
(Meta data then inserted into text)
But automation still requires you to think
… what levels of abstraction are
significant, what languages, meta-content
judgements of the story, are meaningful
significances …
Don’t let a CMS kill yr folksonomies, etc.
“Enhance a user’s connection to a
given place using location-based
technologies such as GPS-enabled
devices and interactive maps.”
Test: tag this icon:
Get shared tags as soon as you
can when there’s a common
event or discussion on the go. For
instance: HA2008, mobileactive08
 That allows the coverage be
aggregated to best effect.
Try this
Guy Berger
Thinning hair…
Internet - documents to database
Social networks = part-answer
Metadata: all over… (title,
heads, text, alt=)
Meaning is in relationships…
AI (eg. link “Brisbane” to “Queensland” &
“Australia” without manually doing it).
But you still have to use your noggin!

Facing down facebook