Google Squared, part 4: Summary

Having had my fun, what are my thoughts?

I understand that what people want is to get from their question to the answer with the fewest possible steps - and preferably without having to formulate and implement a search strategy. (That is, they want the computer to understand what they want and provide the answer rather than the source of the answer.) This incarnation of Google Squared does not do that. (Neither does any other computer that I'm aware of. Maybe in science fiction somewhere, though.)

The user still has to know (1) to use Google Squared instead of the ubiquitous Google search box (moving away from the "one simple little search box that does just about everything" is not to be undertaken lightly - for many reasons that's the golden standard everyone else tries to emulate) and (2) which types of searches will and will not work well in a Google Square. Those are basic elements of search strategy.

What it does do is provide a different search starting point that works well for particular kinds of searches. If you want to compare particular items within a category, Google Squared returns results in a format that facilitates that activity. (But it still doesn't tell you which flat screen TV to buy, given your budget and viewing habits.)

And, depending on the purpose of the search, one still has to link through to the website to get the desired information and evaluate the quality of the website from which the information actually comes.

And a lot of the description information Google Squared retrieves seems to come from Wikipedia. Are we more willing to trust the veracity of Wikipedia because it turns up as a response in a Google Square as opposed to a regular Google search? (The use and reliability of Wikipedia deserves an entirely separate blog entry....)

What I come away with is the sense that artificial intelligence (for all the advances in the field) still has a long way to go. Computers generally understand search terms, but they don't understand question context. The lack of coherence across rows demonstrated that in several searches.

Google Squared may be an early step on the long, long, long journey towards computers even more intelligent than the ones in Star Trek. I won't be holding my breath, but I'll probably continue to try it out now and again. Mostly for fun.

Google Squared, part 3: Free Play

Free of obligations and ready to try my own searches, I started with my interests.

A search for "Juvenile mystery books" returned only 1 result. Nancy Drew. "Teen mystery books" returned 4, but taking out the author/title redundancy leaves only 3. "Young adult mystery books" nets me 7 hits: 4 genres, 2 authors, and 1 award. This does not seem to be working.

A search for "Young adult authors" nets better results. The first 20 results nets me 18 people and only 2 titles. The fact that I'm familiar with the top 20 authors/books may indicate that there is some relevance ranking bringing the more popular people to the top. But it's also a disappointment - I haven't discovered anyone new to try.

Can Google Squared be used for reader's advisory? The search "vampire books teens" suggests maybe yes. The search "Harry Potter read-alikes" suggests no. The jury is still out.

What can be deduced, I think, is that some knowledge of search strategy is still necessary.

Alright, what about my other fallback search interest?

"Muppets" returns pretty good results. I get a mix of characters and TV shows, but that's okay.

• Down around item 42 Nelson Eddy comes up as a result, and I can't figure out why. There's no mention of "Muppets" on the website (Wikipedia, again) the description comes from. I find this irritating. Why, why why....
• On the other hand, I'm simply amused by the images pulled for the items "Semi-formal" (43) and "Foo-foo" (45). The concepts and descriptions are perfectly legitimate for the search, but the images aren't Muppet-related at all.
• I try adding columns for "Created by" and "First Appearance" but return no/inconsistent/poor information. Apparently the system knows what it knows, and automatically returns those columns when creating the square.

Searching for "Muppet movies" returned pretty good results, but some irony as well. All of the major Muppet movies returned high in the square, except for The Muppet Movie.

As the list extended, more and more false positives showed up.

Also notable was the lack of a sense of context and consistency across rows. For example, Frank Oz was returned as an item. (Of course, he is a muppeteer, but not a Muppet movie.) However, the image for him was the cover art for "The Muppet Movie" - which was not returned as a result.

Possibly my favorite result came from Tim's search for "philosophers." I had no idea that Plato lived until "8 May 1999." And, apparently all philosophers (even ancient Greek ones) speak English.

Now I'm curious as to whether Google Squared can help with some of the frequent questions fielded at the youth reference desk.

As I expected given the failure of the National Book Award search, trying "Texas Bluebonnet list" and "Texas Bluebonnet Award" did not net useful information. Better off doing a regular Google search. As both bring up the TLA website as the first hit.

"Arctic animals" is actually really nice, as far as it goes. This might be a good first step for finding an animal. Of course, then you have to find a book on the animal as that's required. And searching outside of the catalog leads to finding arctic animals we don't even have books about. Ditto "Birds" and "African American inventors."

When the day comes when students are taught to evaluate the quality of a website rather than "find a book" we might be in pretty good shape, though. Certainly browsing an instantly created list of "African American inventors" is a lot easier than trying to figure out the relevant LC subject headings and searching them all. Plus it extends our 'collection' if quality websites can be used instead of books.

How is Google Squared useful? Comparisons: places where a spreadsheet is handy, but you don't want to have to populate it yourself. A place to start shopping, for example. Searches on "Flat screen TVs" and "Digital cameras" retrieved useful results. And you can add columns for features of interest.

Google Squared, part 2: Assigned Play

I began with a few suggested (by my boss) searches to get a feel for Google Squared.

"Superheroes AND powers" quickly demonstrated a few of the shortcomings that continue to challenge programmers of automated search engines (and researchers in artificial intelligence).

• The "item name" column includes a mix of not only superheros, but also super powers and concepts common to the world of superheros (magic, masks, and -possibly my favorite- supporting characters).
• I also laughed out loud when I added a column for "Genre" and received not only 'superhero' and 'action', but also 'rap' and 'thrash metal'.

By way of contrast, I performed a regular Google search for the same terms. The 1st hit linked to a Wikipedia article offering an extensive list of superhuman features and abilitites in fiction. It was sorted by ability rather than character, however.

My immediate reaction to "Orchids" was that the images were beautiful (never discount visual impact).

• I began to wonder, however, about the exhaustiveness of the results. Without an authoritative source to compare to, how would I know about the quality of the results?
• Which led me to wonder about the default sort order. It seems like items are returned in descending order of information completness. If that's the case you're losing what makes Google useful so much of the time: relevancy ranking of results.

Again I performed a regular Google search. The first hit was for the American Orchid Society website - authoritative information, most of it freely accessible to non-members.

The third hit was for http://www.orchids.com/, a commerical site, but including a pronunciation guide and list of orchid gardens open to the public.

Since I'm always interested in quick and easy ways to access information from popular book lists and awards, the square for National Book Award recipients was the most disappointing.

• While it was fun to see pictures of all these famous authors, I thought, "Oprah Winfrey, really?" I had to track that down. Turns out she won the 1999 "Distinguished Contribution to American Letters," awarded by the National Book Foundation.
• Odd that the Michael L Printz is listed as a winner.
• I tried to add a column for "Year Awarded" but Google Squared did not retrieve any information.

The first hit in my comparison Google search brought up the National Book Foundation, which presents the NBA. Browse to find out who won, for what, when. Or search by author. This is where I found the answer to my Oprah question. The third hit was for the Wikipedia article with a quick-refernce list (easier than the NBF website) of NBA winners in each category.

Google Squared, part 1: Providing Context

This week's topic: the future of online searching.

Google describes Google Squared :

Google Squared is an experimental tool that takes a category (like US presidents, roller coasters, or digital cameras) and attempts to create a starter "square" of information, automatically fetching and organizing facts from across the web. You can modify your square by removing rows and columns you don't like--or by adding new rows and columns and having Google Squared attempt to fetch the relevant facts. Verify and correct the facts in your square by exploring the original sources and investigating other possible values. If you're happy with your square you can save it and come back to it later. Google Squared does the grunt work for you, making research fast and easy.

"Fast and easy" sums it up, doesn't it? That's what people want. And with every improvement in technology, that standard gets higher (or, should that be shorter?).

"Save the time of the reader."
-Ranganathan's Fourth Law of Library Science

Fast and easy isn't necessarily a bad thing. I'd much rather research in an online database than go through the steps of searching an index, finding the issues with the relevant articles, and finally photocopying them. All at the library, of course. It's a much more convenient method of getting to the same information.

"Same information" is the catch, though, isn't it? As librarians, we're concerned with the accuracy of the information retrieved. Online databases housing electronic journal articles provide the exact same information as the print versions.

As we all know, "information" on the world wide web varies widely in accuracy, relevance, and timeliness.

So, the question is: do changes in search format correspond to improvements in search results' quality?