Researchers at universities must need to sift through tons of data quickly, which inspires so much progress in search technologies. The folks at Northwestern University developed the contextual search technology being used by Intellext for its Watson 2.0.
One interesting point is that Watson 2.0 is not supposed to raise the privacy concerns that Google Deskbar Search and its Sidebar raise. Al Wasserberger, CEO of Intellext, insists that “Nothing gets sent to our servers.” People using Google’s Sidebar can only customize it if they agree to allow the program to send information to Google servers about what they are viewing on their computers.
Even without the privacy concerns, there is certainly room for improvement with Google’s Sidebar. A number of people who use it have complained of receiving less than relevant results. For instance, Greg Sterling, managing editor at the Kelsey Group, was optimistic about its chances to improve, but he clearly wasn’t satisfied with it now: “There’s a lot of miscellany in there that’s not particularly relevant…Over time it will get better. I would give it a six on a scale of 10.”
Anyone whose job depends on being able to get relevant information, especially news, fairly quickly would be interested in a product that can help them find what they need (or even anticipate those needs). One would think that such a program would find its natural market among reporters, analysts, and technology writers, among others. Hmmm, that doesn’t describe anyone we know, does it? Always looking for ways to get our job done better and faster, we decided to stop by Intellext’s website and try out this new marvel of search.
Intellext charges a monthly fee for using Watson 2.0, but invites users to take its product on a week-long test drive. The page from which you can download the free trial explains that the application is 6 MB large and lists system requirements:
- 13MB on disk
- Windows 2000, XP or 2003
- Microsoft Internet Explorer 6.0 or higher
- 256MB of RAM or greater
- 766MHz CPU or greater
Also, while a broadband connection is not necessary, Intellext highly recommends it.
The installation was pretty quick and painless. A wizard came up with the usual licensing notice, whose terms you had to agree to before continuing. It next pops up a registration screen, requesting various information (name, address, etc.), of which only your first and last name, email address, company or organization, and phone number are obligatory. I was not happy about being required to include my phone number.
After you have installed the program, you will have to restart your system to complete the process. If you happen to have a desktop search installed on your computer, you will see a pop-up box that looks rather like this:
Watson says that it can work with MSN Desktop Search, Google Desktop Search, and X1 Desktop Search, so this should not be a problem. We found that the program did use the desktop search, but much slower than simply doing the search yourself – a theme that will crop up again.
Watson 2.0 sits as a one to two-inch-wide bar going down the right side of your screen. It can be minimized, and even turned off, but there’s no getting around the fact that, when it’s on and running, it takes up a good bit of real estate.
The next screen I’d like to bring to your attention is this one:
This screen comes up when the program cannot find sources. This happens if you have nothing open, and/or Watson can’t find good enough keywords. But it doesn’t always come up; depending on how Watson feels, you might see a blank screen instead. In the many cases where it comes up blank, you can’t be quite sure if it is thinking about bringing back something or just sitting idly.
Naturally, you’ll want to do something to make sure you don’t see this very often, which brings us to our next screen:
This does not come up automatically. You need to go to Tools > Info Sources. A number of info sources are part of your default selection. Looking at the top left of this image, you can see a section that lets you add information sources. Double click this, and you get a screen in which you can enter a domain name, and a name for this new source (for example, if you enter http://www.reuters.com/ where it wants a domain name, you can simply call it Reuters in the line below). The program claims you can even specify entire top level domains, such as .gov or .uk. If you have the Web folder highlighted under Information Groups (top right), as we do in this image, you will see the new domain automatically added to the bottom right, labeled Information Sources in “Web.”
If you don’t want the source in that folder (Reuters really ought to be in News, after all), it’s pretty easy to fix. Click on the source, then hit the “Remove” button. Watson moves it back to the list on the left (you may have to hunt for it if you have a long list of sources there). Then open the folder you want to move it to, highlight the source, and hit the “Add” button.
You may also see the “Google” source in the screenshot above. That was not there by default. We had to add it, and it didn’t even work. I’ll get more into this later.
Ready and eager to start working and have our capable assistant close at hand to feed us what we needed, we made our first stop – where else? – the SEO Chat page. Our faithful servant seemed to fail us right away here:
This is actually one of a number of issues we spotted with Watson (more on that later). Sometimes Watson proved to be reasonably light on his feet, fetching us information within about 10 to 15 seconds. Other times, we can only guess his old war injury caught up with him, as it took a minute, five minutes, and longer to get some kind of information. Sometimes, we didn’t get any information at all. By way of comparison, Dogpile, which also draws information from multiple sources, takes less than 10 seconds to return results.
When Watson does work, as it did with this Wikipedia page, the results can be pretty fruitful:
Watson gives you a title, source, and URL, which really doesn’t tell you a lot. When you highlight one of these, as we have here, a yellow tab comes up. Here’s a closer look at the one that came up for us:
As you can see, Watson takes bits of sentences from the page to give you some idea of what is on it. This does not always give you a good idea of whether that source will tell you what you need. In this case, Watson grabbed some phrases out of an ad. It didn’t even grab very much, and the description is downright unhelpful. To be fair, however, Watson does better with certain pages:
If you don’t want Watson sitting there blankly taking up space while you’re working, you can minimize it. And yes, the program sits quietly in the background then doing its job. You may notice a light bulb icon added to your system tray. When Watson has results for you, that white light bulb flashes yellow.
I have one more screen to which I’d like to direct your attention:
Imagine seeing that pop up and get in the way of your work. Annoying, isn’t it? What makes it even more annoying is that Mike got it every time he opened Word after installing Watson. It made Word take three extra seconds to start, and with as much time as we spend in Word, that adds up. It’s worth noting, however, that I haven’t experienced that problem.
One thing I found annoying (Mike less so) is the amount of real estate the program takes up. You can minimize it of course, and it still works. Even so, if you are running at a low resolution while you’re using Watson, this could become a real issue.
Watson supports only four search engines: Yahoo, Altavista, Dogpile, and MSN. We tried adding Google, but discovered that it would only search the Google domain. Watson did not do Google searches. That’s a real problem for folks who favor Google over other search engines.
Watson claims that it works with Firefox. We found that it did not work with Firefox as advertised. It did open search results in Firefox, but it does not performs scans or searches of pages browsed in Firefox. Effectively, then, it only works with Internet Explorer.
The problem we had with how slowly Watson returned results for web pages is worth reiterating here. Remember that your average person will wait only eight seconds for a page on the Internet to load. Since we have already loaded a page by the time Watson can do its thing, it’s not an issue if it takes a little while to bring us results. However, we’re pretty fast at digesting what’s on the page in front of us; in three minutes or less we’re ready to move on to the next page. And it happened far too often that Watson wasn’t ready for us.
In short, Watson performed unevenly at best, and slowly to the point of frustration at worst. We think we’ll wait before we hire an electronic research assistant like this; sadly, Watson did not meet our needs.