|
Google Research
Other Google Resources
|
Google's mission: Organize the world's information
and make it universally
accessible and useful.
Google is an engineering company.
The Google web site is powered by some amazing technology, most of it developed in-house.
Yet people often ask us what we do here at Google Engineering.
"What are you working on? Isn't search a solved problem?"
Glad you asked.
We're working on lots of interesting stuff and one
of the main reasons is that search is far from a solved problem.
Let's say you used Google to search for the topic "Michelangelo's David".
The results page would show "Results 1-10 of about 6,960,000" web pages.
That's pretty helpful, but we could do so much more.
Google prides itself on its algorithm for choosing the most
relevant pages, but it's a work in progress;
we're constantly finding ways to improve its selections.
Plus, the top ten pages listed are all in English; surely there are some interesting web pages in Italian
that we could translate for you, and chances are
at least some of them deserve high ranking.
Over at http://images.google.com
you can find some helpful photos of the sculpture (plus some knock-offs),
but there are video clips, museum guidebooks,
historical articles, and many other sources of information about
David that the web doesn't reach.
And it's likely someone at the Galleria dell'Accademia has a 3-D scan
of the sculpture you'd enjoy browsing.
(From the search results, it's clear that Stanford has some 3-D data too.)
So yes, Google is very good at searching the web for the most relevant
pages for the query you type, but that's really only a minor subset of the true `search
problem', which remains far from solved.
And consider this: We currently search billions of web pages. That's a
lot of information, but even that's not the whole web. And even if it
were, it's still only the web; what about all the other information
out there? Google's mission is to make all the world's
information accessible, not just a subset of the web.
So you see, we have our work cut out for us. Feel like helping?
You don't need to be an expert on searching; in fact, most of the
people in our engineering group had little or no background in
search technology before they came to Google.
To implement a good
search algorithm on the scale of the web requires ideas from just
about every area of computer science, including information retrieval,
artificial intelligence, natural language processing, distributed
computing, large-scale system design, networking, security, data compression, user interface
design; the list goes on.
Or look at it this way: a typical query to
Google can touch thousands of machines before returning the answer.
With all those machines and all that communication, the problems can
be daunting and new ideas and new technologies can always be brought
to bear.
You also don't have to be willing to move to the San Francisco Bay Area.
We have engineering offices in Santa Monica; Kirkland, WA; New York;
and other cities. We also have engineering centers outside the U.S.,
including Switzerland, India, Japan, China and many others. Each of
these offices works on the same problems as in Mountain View. In fact,
many Google projects have membership that spans engineering offices.
The Story So Far
The search problem is tough, but we've made significant progress.
Besides constructing the world's most heavily
used web search engine and a lot of other
public web services
(www.google.com,
news.google.com,
images.google.com,
groups.google.com,
catalogs.google.com,
labs.google.com,
maps.google.com,
gmail.google.com,
scholar.google.com,
adwords.google.com),
not to mention
the Google Appliance
and the Google Toolbar,
Google Engineering has achieved a number of other milestones
you may not be as familiar with:
- We designed and built an advertising system that automatically performs hundreds of millions of automated auctions per day to determine
the placement and prices of advertisements appearing on Google search result
pages and
non-search
pages on dozens of other sites.
- We built a very large scale, distributed, fault-tolerant file system,
called GFS, to help manage and process huge data sets. A paper in SOSP 2003 describes the file system.
- We designed and developed a fully automatic news system (news.google.com), which has rapidly
become a standard tool for human journalists. Custom algorithms group
together articles about the same story from different news
organizations from around the world, providing diverse viewpoints
about the day's events. Heuristics judge the importance of each story
relative to other stories in the news around the world to generate our
headline summary pages automatically.
- We built a searchable
archive of millions of catalog pages by scanning and OCR'ing
printed catalogs.
This experimental service demonstrates the benefits of
a searchable interface for information not previously easy
to search or browse.
- We're exploring large-scale machine learning as a means of
improving search quality. Our spelling correction system is one
excellent example (spehl korector?
phonitick
spewling? who needs a dictniary?). People
searching for Britney Spears have
clearly found it useful on many occasions. In more recent work, we
have been working on algorithms and techniques to construct very large
scale Bayesian network models to help understand the relationships
between words.
- We are building a
large-scale public e-mail system
capable of
storing 2 GB of data per user, with a unique user interface that
centers around search, scaling to many millions of users.
All this is achieved by connecting together tens of thousands of
servers behind google.com
and providing them
with a lot of custom-built, cutting-edge, innovative software.
It should be clear by now that the search problem involves much more
than just searching, and that some of the most exciting work at Google
happens behind the scenes. We're also working on a number of
interesting projects at the moment that are too preliminary to discuss
here, and we're always looking for new and interesting ideas.
Who we are
Who did all this?
A dedicated and growing
team of smart, creative programmers and computer
scientists -- but we just call them engineers.
They come to Google with expertise in a large range of topics.
And before they joined Google,
some of them also built software, hardware, tools and other technology you
might have used:
Needless to say, googlers have also published on a fairly wide
range of topics. Not all the engineers have done such public work, of course,
and their backgrounds vary enormously.
Some started at Google right after college;
others came after spending time in academia or industry.
Some love thinking long and hard about difficult problems;
others just enjoy getting their hands
dirty building and deploying massive, real-world systems.
What they all share, though, is an enthusiasm for
the challenge of making the world a better place through
the intelligent application of information technology.
It's a blast.
Interested?
There is so much still to be done,
so many hard (but fun!) problems to solve,
so much information in the world,
we'd like some of you to join us
to help us in our task of making Google even better.
The engineering challenges we face at Google are exciting
and the perks of working here are wonderful,
but the real reason most of our engineers came here
is that Google Engineering is a great place to work.
It's got lots of really smart people,
amazing technology,
fun problems,
and a chance to make a real difference in the world.
|
