Steven Arnold | The search continues
GCN Interview with Steven Arnold of the Google Government Report
"I really believe FirstGov does much more focused indexing [than Google]." Steven Arnold
Steven Arnold got an early start on search engines. In 1971, his
employer, Halliburton Co., assigned him to digitize the
company's technical reports in order to make them searchable.
He has worked in the field ever since. In the past decade, he has
moved over to consultancy, starting his own practice, Arnold IT. In
2000, he helped generate the technical plan for the first iteration
of the General Services Administration's FirstGov government
search engine. (His son, Erik Arnold, currently works on FirstGov.)
More recently, he launched the Google Government Report
(www.ggreport.com), a newsletter and electronic information service
offering tips on how to be better recognized by Google. We caught
up with Arnold to get his views on what is happening with both
enterprise and Web search.
GCN: What is the state of search these days?
Steven Arnold: This has been a time when people are
realizing that enterprise search doesn't work.
Folks with enterprise search systems are really on the lookout
for technologies that make search more useful for the users.
So what will carry us into 2007 is a collection of technologies
we think of as text mining, where software algorithms look at
documents and find the names of people, places and things and
attempt to relate them to one another.
GCN: What is wrong with search technologies today?
Arnold: Take a real-life example: You and your significant
other go to England and she says on the way home, 'I loved
that jumper that I saw at Harrods.'
So, if you're like me you don't have any clue about
what she's talking about. So now I go to a search system,
like at Neiman Marcus Online or Overstock.com, and type in the word
'jumper.' What I get back is not that sweater that she
saw in Harrods. I get stuff back unrelated to that idea. And that
is a very common problem.
GCN: What are the text mining companies doing that the
search companies can't?
Arnold: From a technical point of view, companies like
Attensity Corp. and nStein Technologies are not focused on search.
They are focused on figuring out the nuances, relationships and the
important concepts in a document. Their systems generate index
terms that an enterprise search system can suck in.
GCN: So how does this technology work?
Arnold: Every one of the new companies that I have looked
at'and I have a text mining report where I have been tracking
this'are approaching the problem both mathematically and by
doing vocabulary and knowledge-based analysis. Their software
decomposes sentences into subjects, verbs and adjectives and
analyzes the results with the predictive algorithms.
What is unusual is that the computer chips are so darn powerful
now. These new companies are basically saying that they are going
to use these chips and throw everything we learned in computer
science at the search problem and it will work out just fine.
So what you have now are hybrid [search/text mining] systems and
by golly they are very interesting. When you use one of those
[services] against scientific and technical information, your
accuracy can be 85 or 90 percent. If you run it against general
business news, you can hit 80 to 85 percent accuracy. That's
almost as good as a human search.
The system can run automatically, and when something [unusual]
comes up, you have a person who has been through school look at the
error report and fix the mistakes. When someone spells Al Qaida
differently, and an exception comes up, an analyst can look at that
and say, 'This is an OK spelling.'
GCN: So we had no idea this sea change was happening in the
industry.
Arnold: Yeah, it is going on under everybody's nose
because so much focus is given to Google.
Google actually has some nifty technology like this, but it is
so anchored in the consumer space. These smaller companies are like
cigarette boats racing around the destroyer.
The age of innovation for this is not over. I've been at
conferences this year where people are saying, 'Yeah, well
it's over. Microsoft is giving it away, or Oracle is there,
Google is there. There is no room for innovation.' I just
don't agree.
GCN: Why should agencies care about Google?
Arnold: Let me give you an anecdote. I got invited to meet
with the people from a large insurance company in Denmark. I asked
how much of their traffic came from Google [searches]. And they
said they thought about 35 to 40 percent. I asked if there is a way
to check, and they said they could do it right there from a laptop.
[When they checked], they looked at me and said 'You know
what? Last month Google was 80 percent of our traffic.'
GCN: So you are seeing an increase in Google-derived traffic
within the last few months?
Arnold: Literally, within the last 12 to 16 weeks. The
anecdote underscores what we've seen in other work, that
Google is basically the search engine of choice for virtually
everyone in the world.
As people realize how much traffic comes to them from Google, it
becomes more important to understand what other people are doing to
make sure their Web site is indexed by Google and their sites come
up in the context of the proper keywords. You really now have to
pay attention to Google not because Google is the greatest company
on the planet, but because Yahoo and Microsoft just haven't
done that good of a job competing.
GCN: So what can agencies do to better present their pages to
Google?
Arnold: The first step is to create a sitemap that conforms
to Google's guidelines, because Google has already convinced
Microsoft and Yahoo to follow its formula. So that is job one.
Job two is to take a very hard look at the page names and the
URLs on your site. Most government Web pages I look at have very
long and complicated URLs, and Google's robots can process
those, but they prefer to process human-understandable URLs.
The third thing is the government needs to do a better job with
content. The government has great information. The Department of
Agriculture has outstanding information, but it is presented in
such a way that makes it really hard to index and search
effectively. If you want a good report, you have to download a huge
PDF file.
So I think the government has to make more of its content easy
to comprehend, and not put out these 5-megabyte globs of data.
GCN: Any thoughts on the battle between the two premier
U.S.-government-focused search engines, FirstGov and Google U.S.
Government search?
Arnold: There is no battle at all. I really believe FirstGov
does much more focused indexing.
When Google sends its robots to an agency Web site, it looks at
the links, indexes the first 100,000 characters per page and
follows the links two levels down. But FirstGov looks hard at these
sites and goes very deep into the site. Remember the FirstGov
[result] set will be much smaller and more focused than the Google
set, which will be very broad.
If you work for an information service, you certainly can start
a search with Google, but if you want to be thorough, you will have
to look at FirstGov. If you're a government worker, you might
want to start with FirstGov, but you definitely want to take a look
at what Google has indexed.
So think of FirstGov as drilling down into a topic and Google
going very broad across many topics. So the two services are
complementary.