SiteLines - Ideas About Web Searching

Not-So-Smart Answers at AskJeeves

The latest additions to AskJeeves "Smart Answers" were reported by Gary Price in his August 22 Search Engine Watch blog. Although the latest additions expand on the existing Jeeves collection of sources that are pre-selected to handle queries for many common factual searches, they remain, like so many other "answer" capabilities of the major search engines, pretty mundane. Search Portfolio's research team tested many popular topics (e.g. marijuana, botox) to see if Smart Answers would deliver pre-selected content, and it didn't. Surely those sample searches are as common as one for burkina faso, which in AJ turns up through Smart Answers via the CIA World Factbook. Nothing new here: all the major search engines would turn up the CIA WF in the first 10 hits for the same query.

AJ's Smart Answers harkens back to the early days of AskJeeves, when it differentiated itself from other search engines by matching simple queries for common questions against a set of pre-determined web sites which could provide a variety of content that could answer the "question". Eventually, the site morphed into a meta-search engine, and then, with the integration of Teoma, became a more conventional search-engine-with-benefits. There is clearly little new hear, and not enough of real value to recommend this over other answer engines.

More Silly Search Engine Size Stories

Since Yahoo disclosed the jump of its index size to just over 19 billion (!) documents, I've been following a series of interesting posts at the Technologie du Langage blog from Jean Véronis, professor of Information and Technology at the University of Provence. In great detail (and in English), Véronis recounts, with good link references, the index-size story starting with Yahoo's announcement. He then systematically and persuasively refutes both the allegations of database size and the research methodology of a US study comparing database sizes of Google and Yahoo.

On the US study, Véronis concludes, "I find it amazing how quickly such a flawed study could be quoted with so much excitement all over the blogosphere and even make its way to the respectable New York Times." Those of us who are used to the republication as "news" of unverified company press releases are, sadly, not so surprised.

Although most of Véronis's posts at Technologie du Langage are in French, the blog is an outstanding (and rare) source of competent criticism of search engines, and deserves to be in the RSS feeds of serious web-watchers.

Want a glimpse into the search future? Google Labs and Yahoo Next

Want to see what's new on the (commercial) search horizon? Try the not-quite-ready-for-prime-time search offerings of the major engines, like Google Labs and Yahoo Next.

One tip: If you're like me and you want to try to get a glimpse of trends, try to look for similarities between the offerings of these two hot competitors. Right now I'm seeing pointers to a differentiation of two types of Internet -- one for "serious" research-oriented questions (although I'm curious to see how either of these behemoths defines "research") and another for shopping.

We already see some of that differentiation with Google Scholar, Google Print, Yahoo Mindset and Yahoo Subscriptions. Each of these alleged "testbeds" offer searchers access to more highly differentiated content than would be available in straight-up Google or Yahoo.

Why separate research and shopping, and what will this do for engine revenues? These new spinoffs offer a great untapped market for paid consumer-level content, as well as context-sensitive ads. The big engines are clearly betting that serious searchers are ready to pay, if only to satisfy their need for quality information that they can't find in a sea of search engine spam.

By separating research (i.e. pay-per-view, with a few free tidbits thrown in, like content from .edu sites) from shopping (i.e. everything else), the search engines can return to more traditional relevancy algorithms. Remember relevancy conditions like link analysis, proximity of terms, frequency of term occurance, and currency of information? In an shopping-free engine, those conditions can be re-introduced, leaving the optimizers to focus their beat-the-engine techniques on the shopping side of search.

MSN Search Adds (Some) Useful Advanced Search Features

Gary Price reports in a recent issue of Resourceshelf.com that MSN Search has added some new advanced search features.

MSN Search joins the majority of other search engines that now offer the FILETYPE feature (filetype:pdf). But a few others are not only novel, but potentially useful, specifically CONTAINS (contains:) returns documents that contains words; LinkDomain (linkdomain:) returns sites that link to ANY page in the specified domain; InAnchor (inanchor:) returns results that have the keywords in the anchor text (that's the text that is surrounded by the A HREF tag)

No doubt the other search engines will rush to catch up. I still don't much care for MSN Search as an alternative to my other favorites - but some of these features may encourage advanced searchers to try MSN in cases when other search engines can't be exploited sufficiently.

Searchengineshowdown.com adds new search engine and more reviews

Of all the lists of search engines and features, I've come to rely on librarian Greg Notess' Searchengineshowdown.com as my best-bet source of trusted information and reviews of the spidered search engines. Last week, Greg posted several new articles and reviews of search engines to the site, bringing it up to date. Exalead, a relatively new, if smaller, search engine, with a bag of cool features, has been added to the list of search engines, together with a preliminary review. Greg has also reviewed the new MSN Search and has updated readers on new features in Yahoo, Gigablast and Google.

Links from commercial search engines to PUBMED citations

The staff at the National Library of Medicine added a great little feature for searchers who click on PubMed links found in a search engine like Google or Yahoo. Now, if you click on a link to a PubMed citation from a search engine, you will see your search terms entered in the PubMed query box and a notation indicating what that search would retrieve in PubMed along with a link to the retrieval. This is a great way to encourage searchers to conduct a supplementary search in PubMed, and possibly even introduce them to a tool that they might not already be aware of.

View the press release (which has some good illustrations) at http://www.nlm.nih.gov/pubs/techbull/ma05/ma05_google.html

Of course, the lingering issue remains just how likely users are to find any PubMed citations. As I wrote earlier this year in Just Because It's Indexed Doesn't Mean You'll Find It, Pubmed's content often doesn't rank high enough in search engine results to be noticed by searchers using common terms.

Lycos search results now come from Ask Jeeves

In it's March 2 press release Lycos announced that Ask Jeeves technology will now deliver its search results. Of course, that doesn't make Lycos a better search engine, just another rebranded version of something else.

Results include the usual sponsored results at the top of the list, but readers should note that the "pure" search results, which derive mainly from Ask Jeeves-owned Teoma search engine, also include paid placement and paid inclusion. The paid results are Looksmart's editorial and paid results as well as those from Lycos' Legacy Inclusion program.

Y! - Yahoo's Attempt to Improve Searching (sort of)

Earlier this month, Yahoo! unveiled Y!Q, it's new beta service. Designed to improve concept searching by allowing users to grab selections of words -- from web pages, added keywords/concepts, "more like this" options, and more, the service enables users to quickly try out various options using these selections.

It's an interesting way to generate a variety of results lists in a large search engine, and will be attractive to the vast majority of end users who don't want to -- or can't -- formulate Boolean search statements.

Barbara Quint has a nice overview of the beta service in InfoToday Newsbreaks for February 14, but interested users will want to play with the tool in order to get a sense of its relative value.