May 13, 2007
Google Categories
Last month, Google started to test a new way to present search results that groups the pages by category and displays the top two results from each category. For example, if you search for a movie, you could see reviews, news, blogs, encyclopedia pages, stores. Each category can be expanded.
Google already shows refinements for a small list of topics like: health, computer games, travel, cars and photo equipments. "Topics are specific search areas that Google is developing with the help of expert contributors. Contributors to topics annotate websites that they think are especially useful, relevant, or authoritative to a topic with pre-defined category labels." But not only experts can contribute. You can select the topics that interest you and use Google Marker to annotate sites as you browse the web.
Google also has a list of general labels and those labels were used to categorize search results in the experimental view from last month. To annotate pages with the general labels, you can create a custom search engine and add labels from the list. I made a custom search engine to test the quality of the labels and I noticed that in many cases the labels didn't describe well the content of the page.
Many people enter very broad queries (like "Vista") and search engines try to guide them to the right results using refinements. One way to refine a query is by adding another word that eliminates the ambiguity (for example, "Vista reviews"), but this is not always the best idea because someone could review Windows Vista without including the word "review(s)" in the page. Having a list of pages that offer reviews could save the day.
Hakia, a semantic search engine that will officially launch this year, uses categories for general queries, like iPod or Italy, but they appear to be dynamically generated and look more like an encyclopedia entry.
It seems that Google Co-op's secret mission was to rebuild at a larger scale Google Directory (ODP), which also organized the web by topic into categories. One thing is for sure: this year, Google's interface will undergo radical transformations.
Related:
New ways to visualize search results
SearchMash
*does the Buffalo Bill dance*
ReplyDeleteThis is a great idea. I'd especially like to see categories defined to separate commercial from noncommercial information. They might be on to a real search improvement here.
ReplyDeleteTry searching for [Vista ~reviews] the tilde command (http://www.google.com/help/refinesearch.html) looks up synonyms.
ReplyDeleteClusty is a little search engine based in beautiful Pittsburgh PA that has been doing something similar to this, but in my opinion, much better.
ReplyDeleteCheck out my Clusty search results for the term Google
I love the idea, but you just can't beat Google for relevant results so I never really use them.
I like the idea Clusty has, though I fear we'll never again see a true picture of the web via search. That is, since ranking = profit, and making money = the reason for human existence.
ReplyDeleteIf a categorical view of results changed ranking methods so it required more precision to cheat (er, market), it might allow a clever search company to catch the web undressed. It would be very nice to have the option of viewing results smartly presented, at the cost of suffering ads, or wild, naked, and free.
Clusty is a meta search engine and a meta search engine will never be able to have a good way of ranking results because it mixes results from search engines that have different algorithms.
ReplyDeleteBut the clusters are pretty good, even though Google's related searches are almost as good.
Compare
the clusters for SUV on Clusty
with
Google's related searches at the bottom of the page.
For the record, here's what I see.
=== Clusty ===
Ford
SUV Accessories
Sport Utility Vehicle
Reviews
Photos
Cars, Trucks And Suvs
4x4
Used Cars
Tires
Honda
=== Google ===
suv comparison
suv reviews
ford suv
suv toyota
toyota
ford
chevrolet
jeep
Note that Clusty restricts the search for SUV to one of these clusters, while Google performs a new search.