Google And IBM Team Up Search Technology

An Anonymous Reader wrote to mention a Reuters report on the teaming of IBM and Google over Google's Desktop search technology. From the article: "IBM is linking up its OmniFind corporate search system with Google's free desktop search for business to make it easier for users to locate information throughout an organisation that is often locked up in many separate systems ... Google wins IBM's endorsement among corporate technical managers for its desktop search product and IBM gives corporate information workers an already popular entry point into back-office databases through Google's search. Searchable data ranges from e-mail to computer files to blog postings to corporate repositories of data, images, audio or video, Prial said. Much of this is not available using public Web search tools. Typically, it is hard to reach inside a company except by trawling through many different programs."

Comments

Why no mention of the Google Search Appliance?

[xmas2003]

Submitter (and the article) talks about merging Google Desktop Search with IBM's Omnifind product. Since it sounds like the purpose of this is to look inside the firewall/enterprise, I'm really surprised that the Google Search Appliance wasn't mentioned. Wouldn't this be the logical product to internally spider a Corporate network and provide Search Engine Services? And since it was not mentioned, does this suggest some shortcomings with the GSA (which Google released a few years ago, but hasn't seemed to have done well) ... or is this announcement a way for Google and IBM to buddy-up to each other?

X10 Halloween Webcam is online - is it a "Trick" or a "Treat" ... ;-)

Re:Why no mention of the Google Search Appliance?

[erwin]

my personal opinion is that the Google appliance is WAY too expensive, which is why enterprises haven't run to get on board with it. When we looked at it, as a small state agency, there was no way we could justify the cost.

Rolling a semi-distributed solution out to everyone's desktop seems like it might (without know any details) be a cheaper way to do it. It's not like there aren't CPU cycles just going to waste there or anything....

my $0.02

Yikes..

[grub]

I rirst read that as "Google and IBM Team Up with Scientology".

Go Xenu!

But will they like what they find?

[King_TJ]

It occurs to me that setting up this powerful of a search tool throughout the typical business means at least initially, people are going to dredge up all sorts of unexpected and perhaps unwanted things. (Say, just for example, a middle manager is working on plans for giving out raises or pay cuts for the next year. He might have an Excel or Word document or two out on a shared network drive mentioning some of his rough drafts for the idea. Bad idea of course, because he should be saving that stuff in a more secure place - but you know how things go. Might have even saved a version of a file there completely by accident? Traditionally, it would go unnoticed in most cases. But what if another employee is playing with the search tools, looking for all documents related to salary - and finds out what all of his co-workers are going to get paid next year?)

It'll only take one or two such incidents before management will demand removal or a serious lock-down of people's ability to use the search tools.

Also, think of the shock some people might get when their boss starts searching the contents of everyone's corporate email for key phrases. Once again, sure - we all get the general idea that "business email is not private", but how many of us took what seemed to be an infinitely small risk and made a couple off-color jokes, jabs at management, or other such things in an outgoing email written to a buddy over lunch break or something? Did you remember to make sure all that stuff was removed from your "sent items" or "deleted items" folders?

Re:But will they like what they find?

[BigGerman]

You are absolutely right on this. I participated in developing of a competing product (see my sig, they are good people and deserve a plug) and I remember that our spidering tools were getting stuff that definitely noone wanted to see public, like list of passwords from a backup CD, metadata (author, etc) from Office docs, unencrypted credit cards from databases, etc.

The solution is to educate people that "shared drive" is indeed shared and provide a separate network share for backups, etc, and make sure that share is not configured to be searchable.

Article Title

[dotslashdot]

"Google to help IBM search for more business."

It could be a good thing

[zappepcs]

This could be a good thing. The GSA does well for registered websites, but I'm not sure what it will do for standard files. If searchable files on the local machines have to be tagged as searchable by the user, and not done by default, then anyone can publish information to the companies repository of information without having access to a website, or needing to get the document published, just tag the document with a couple of keywords or some such, and off it goes to the internal wiki... or something like that anyway. That could be a very good collaboration tool, despite my poor explanation.

By having to purposely publish a document, the default is that your data is not published, thus allowing you to email your family without fear of publishing the holiday party plans at your mom's house.

Likewise, all those involved with a project can publish to that project's wiki without serious effort, just right-click and select the keywords etc. or something like that anyway.

Web search for IBM with Google!

[bubulubugoth]

THe IBM site es HUGE... really, really huge, and trying to find something specific is very hard, usng the search tool provided by IBM.

For example, trying to get to the Java JRE 1.4 for PPC 32 bits, to use at my IBOOK, is much more easier with google than ibm own serch engine.

Also, my last painfully search at IBM, looking for development information for their JPOS implementation, was almost impossible with IBM serch engine, but using google, wasnt so much..

Really IBM should hire or use the google services...Maybe this is a beggining... And IBM partering with Google, one more thing to worry about for Microsoft.

IBM lately, with porting a lot of apps to Linux, is showing a remarked "low profile" hostility to M$..

They are a lot of benefits if you switch from MS to Linux at IBM products...

I wonder

[Fahrvergnuugen]

I wonder if this means we'll be able to buy a google search appliance that can index Lotus Notes databases in the future. I work for a company that has spent the last 15 years throwing information into lotus notes. We now have several thousand notes "databases" and no one can find anything. It's a complete mess.

Search as dialog

[G4from128k]

The core challenge is that search is not a function, but a dialog. Currently, Google lacks the kind of interactive refinement of searches that would really help people find what they want. If I search on Apple, Google doesn't know whether I mean Apple (the fruit), Apple Macintosh (the computer), Apple Macintosh (the fruit, again), Apple (Fiona, the singer), Apple (Big = New York City), etc. The first page of hits for a search on "Apple" should offer one each of these various meanings -- something like this search for Apple at Clusty which is good, but doesn't take the concept far enough.

Even if the search engine knows that I want Apple Macintosh (the computer), Google still doesn't know if I want pages that describe hardware, software, peripherals, the company's site, old 68k Macintoshes, new machines, stuff for sale, fan sites, technical support, etc. Yes, I can add more search terms, but adding terms has two great problems. First, added terms often end up throwing away the very pages that I seek if the page doesn't contain the words I use (e.g., a search on Apple Macintosh misses all the sites that just call the computer "Mac"). Second, I may not even know what terms to use or how to spell them, but like pornography I'll know good search hits when I see them.

What would help is some interactive process that directly asks or deduces whether the search hits are on target or not. The second page of hits would differ depending on how the searcher interacts with the first page -- giving the user more or fewer of pages similar to those they liked or didn't like respectively.

Search should be more like a game of "hot-n-cold" in which the searcher can easily tell the search engine which hits are "warmer" (more on target) or "cooler" (less on target).

Niave Question: What does IBM do?

[LionKimbro]

In 1988, if you asked me what IBM does, I would say, "They make computers."

Because, while my mom owned a Compaq, my friend had an IBM at his house.

Now, it's 2005, and I find myself asking: "Just what exactly does IBM do do?"

Joel tells me that they make clear plastic telephones. One of my bosses say they provide "business solutions." This post seems to imply that they have something to do with search. Their website is indecypherable to me.

So, please, if someone could explain to me in plain language: What is IBM, and what does it do?

I'm aware of what Wikipedia says, but I'm interested in hearing what juicy nuggets local Slashdot readers may have to offer.

Re:Niave Question: What does IBM do?

[jiushao]

Well, one thing they do that is particularly interesting here is making money off patented research. The example that might be very relevant is that they have a patent on so-called hub/authority-ranking.

Hub/authority-ranking was invented about the same time as PageRank by Jon Kleinberg, a very simple overview is that for each search the method will assign hub and authority scores to the found pages. A page gets hub score for linking to authoritive pages, and an authority score is given by getting linked from hubs. This is of course a recursive definition, the results are approximately calculated by some nice matrix trickery. Any pages that don't contain the search terms but get a lot of authority are also included in the set.

The nice thing about this method is that it deals well with smaller communities and concepts, where a search for Java with PageRank (and thus Google) will go on and on about the programming language the hub/authorities ranking will have a better chance to pick out a separate set of results for the island and present a few hubs and authorities for each at the top of the results. PageRank fails this kind of things horribly since Java the language has such an incredible pagerank. It also does a lot better for sources that might not refer to the terms searched for directly, Google does this to a very small extent by adding the link text to the set of terms on a page, but the hubs/authorities method gets way better results.

The downside is that the matrix calculation has to be done for each search with hub/authorities whereas Google precalculates PageRank for all pages every few months.

So, this sure seems like a somewhat relevant nugget of information here :)

Google Search Appliances start at $3000...

[nuclearmoose]

The mini has been very successful for many small-to-medium size business: Google Mini
And while the 1001 @ $30K is a bit more expensive, it has 5x the doc count, it has many more features... database crawl, feeds, multiple collections, secure doc crawl,etc.

Novell to be next?

[dana340]

On an earlier Google article, I discussed what Google might be planning next. Full thread here, my reply here. And this adds fuel to that fire. Google and IBM is a more than worthy foe to compete with Micro$oft. Let me share my vision.

Google buys out Novell, and takes SuSE Linux. Although they have their own specialized search appliances, SuSE would allow business administrators to have some level of control over an IBM/Google supplied server. Why? An IBM/Google based computing system. Imagine, your business uses applications, such as Google office, Google bookkeeping, scheduling, and so on and so forth in one office, or with all of your traveling sales reps. Web based, your system no longer would require VPN access for users to work away from the office.

What's more is, workstations can be made to be nothing more than a glorified web surfing machine for the basic business user. No ability to install applications on a local machine, and therefore no issues with spyware. It would help to combat viruses too. Simplifying IT for the many companies. And lowering costs as compared to a windows based infrastructure.

Google likely would lease equipment and services to businesses. Small businesses would only need an internet connection really. Lease Google equipment and pay monthly, no assets to pay taxes on, and monthly revenue for Google, and they would start with the Novell/ SuSE market share under their belt.

Please remember it's still only speculation, but I'm confident enough to get some shares in Novell. Interestingly enough, while talking about this to the other geeks up here, someone mentioned this, a worthy watch. (I saw this after I came up with my theory) Epic2014