search engines

The article below claims that Google has invested in the 'Semantic Web' with the acquisition of Applied Semantics. The only connection I can see is the word Semantic in the name of the company. What this does seem to show is that Google is building up its armory of weapons to deal with analyzing content to produce better targeted advertising and that its core relevance ranking software isn't enough. Google Invests in the Semantic Web - search engine news blog

Anil suggests using a RAM disk as a solution to Windows' shabby 'find'. Well yes and no - sure you want to have your search index in RAM, accessing RAM is 1000 times faster than disk access, but the main problem with Windows is the lack of a decent searchable index in the first place. The Longhorn version of windows will have full text search built properly into the filesystem. By the time that Microsoft release this, XML databases, or even better, graph based databases may make their efforts appear to be already out of date. A filesystem index is a perfect application for a native XML database. anil dash - archives

"Google argues that SafeSearch is designed to err on the side of caution." Report criticizes Google's porn filters | CNET News.com Er... so what? Imagine if it didn't err on the side of caution. Any system such as this is either optimized for relevance or retrieval, and Google's is optimized for relevance. The best solution to filtering is an editor/software hybrid. Editors cost money, so the only practical way of dealing with this is to have site owners become incentivized editors and submit to a growing list of sites which wish to be removed from a software determined blacklist.

Interesting article on content based advertising. Gil Elbaz of applied Semantics claims that "the very fact that search engine algorithms remain largely keyword-based means that they aren't particularly sophisticated in learning what a page is "about." I think this is optimistic. Google have the expertise to develop a concept based approach, possibly using intellectual property gained through their acquisition of Outride, but they certainly need to get their act together here or text ads advertising Hummers from suvssuck.com aren't going to be that impressive. This is particularly important when you consider the following: "While clickthrough rates might indeed be lower, Google claims that their tests show that post-click behavior (conversions to sales) resulting from content-targeted ads is similar to that seen with search engine advertising."

One good thing to come out of consolidation in the search space is the general realization that Google is, first and foremost, an advertising company with an excellent brand. It is another advertising company, Overture, that has bought some of the main search engines, not the other way around. Search technology is a commodity and subtleties like PageRank are icing on a cake that others have the recipe for. The main issue with search from a technical perspective is scalability and, as FAST has shown, Google is not the only one to have figured this out.

Info on Longhorn, Microsoft's planned file system: Longhorn will finally allow you to search a single Windows machine with similar ease to the way you can already simultaneously search hundreds of thousands of other computers, via a web search. But, as Danny Ayers points out on RSS-dev, Longhorn is based on a relational file system, something which is perhaps obsolete for the purpose. An XML or even better, a graph based model would be more suitable. "Perhaps a little unimaginative of them to use a relational store (a graph model would be a better match for the networked computer environment IMHO), but I suppose they've already got the code. I guess at least it will mean that "Find File" will take days rather than months..."

This is huge- ok, perhaps overstating, but remember, search ads are a 1.5 billion dollar business and the following extends Google in the general ad space alongside companies like DoubleClick. The first signs of what the Google/Blogger combination will yield are shown with Google's latest Adwords initiative. Google has just launched adverts which are based upon the content of a web page, for an example see the old Industry Standard website. This is a major new direction for Google since it extends their main revenue source from adverts on their own and affiliated search sites to any website. It also marks a trend away from the banner ad - in fact web advertising will increasingly consist of targeted text advertising and rich media adverts. The key to the targeted ads is that they change according to the content of the page (Google continually spiders the pages and serves different ads...

So I was wrong last week, but correct originally. Overture have indeed bought FAST. Despite the drop in their share price, this is a good move by Overture. Google has not yet raised cash from an IPO and so can only offer stock. Overture is mopping up the loose ends in the search landscape by offering cash and is covering itself against longer term weaknesses. Overture to buy search services | CNET News.com

With the majority of their revenue coming from two partners who have destination sites and therefore can beat Overture down on revenue split (or 'cost of traffic' as Overture call it in SEC filings), Overture needs a destination site, but Google is way ahead of Altavista in terms of users. This consolidation is bad news for FAST, Microsoft could possibly buy them for MSN, but if inclined to buy at all they may look at AskJeeves. So what does the search space look like now: Google Syndicated and destination site search, paid listings (adwords) and relevance ranked search. Yahoo Syndicated (Inktomi) and destination site search, paid listings (they are probably thinking of building their own) and relevance ranked search. MSN Destination site search Overture Syndicated and destination site search (Altavista), paid listings and relevance ranked search. Possible acquisition targets: AskJeeves (Microsoft); FAST (Microsoft); Espotting (Overture or Google) Is this Overture's...

The two most important things on the web are publishing and searching. Three and a half years ago when I stepped off the plane at SFO with my carpetbag, I had two meetings lined up, one with Autonomy and one with Evan Williams from Pyra, a search behemoth and a tiny publishing (actually collaboration software) startup. Pyra was more exciting. Evan and Meg Hourihan had developed a product that was designed for people who needed to update websites often and easily. To do this they had to produce a beautifully elegant piece of software. Blogger along with Manila seemed to point to something profound, what happens when the web becomes two-way. Weblog publishing tools allow you to create embedded meaning within documents and they embed that meaning where it is relevant - in the nugget of information that is published as opposed to a web page. RSS syndication and the...

Searching for people is one of the most difficult problems for search engines. "Surnames cause the biggest problems in genealogy... In one document, my ancestor's surname appeared as Mury, Murry and Murray. Samuel's surname is Murry on his headstone. On the stone of his son Levi, it's Murray." The most widely used standard for searching for name variants is Soundex, however this does not take account of non Anglo-Saxon sounding names and does not work with a thesaurus of known variants. NameX, which was developed for use on Originsearch was designed specifically for looking at the name variants used historically, as well as phonetic variants. For example, it picks up 56 variants of Mury. It would be interesting to take the Namex concept further and allow you to choose name variants based on the ethnicity or language of the search. Floridian: How well do you know your name?

Overture's Achilles heel is not owning a destination site. As CNET reports, Yahoo's poaching of an exec from Overture is significant. The barrier to entry is too low for Yahoo to not build its own paid search, avoid a 35% commission fee and sever links with Overture. Yahoo is putting the pieces into place to replace partnerships with Google and Overture and provide their services in house. For Google this means that Yahoo will compete with them, but Google owns THE destination site for search and provides its own paid listings. Overture could become the Inktomi of paid search, beholden to those who own the destination sites, the majority of its revenues come from two partners, one of them being Yahoo. It proved the business model for search, its revenues are impressive and its $1.3 billion valuation quite an achievement, but the business model missed one key ingredient, it doesn't...