Expert Systems for Google

Would-be Google AdSense rival Expert System launches its Cogito Semantic Advertiser tool, which discerns the meaning and context of words to provide more relevant ads. By leveraging semantic technologies, Expert System joins a cadre of search providers that includes Microsoft-owned Powerset, Hakia, Yedda and Zoomix.

Google has a new challenger to its AdSense platform, courtesy of semantic software maker Expert System.

Expert System, which makes software that interprets information in text, has launched Cogito Semantic Advertiser, a tool that, when paired with Expert's semantic search engine, processes the meaning of text to ensure that an ad's placement is relevant to its assigned Web pages.
Expert System is the latest vendor trying to nibble at Google's search advertising market share by offering an alternative to Google AdSense, which uses keyword frequency to place ads. Another upstart, Proximic, debuted earlier in 2008 to attack Google on this front, but Proximic leverages a contextual ad-based approach without semantics.

So, what problem is Expert System trying to solve? J. Brooke Aker, CEO of Italy-based Expert System's U.S. subsidiary, showed me in a demo how Google's AdSense program will occasionally place misleading or inappropriate ads for certain articles.

For example, next to a New York Times science story on a jaguar, the Google AdSense algorithm picked out ads for Jaguar automobiles. In another article on an airline disaster, AdSense showed ads about low airfare rates for vacations. The algorithm didn't intend to provide results in poor taste, of course, and neither did Google's human engineers.

The problem is that AdSense relies on keyword frequency but doesn't drill down into the semantics—the meaning in the words. Cogito Semantic Advertiser attempts to go further by using semantic intelligence to analyze the text on each page and ensure that ads are placed appropriately to increase click-through rates.

Semantic search, also used by now Microsoft-owned Powerset, Hakia, Zoomix and Yedda, looks at how words in a sentence relate to one another and tries to understand the context of keywords. Terms with several meanings require semantic analysis of the other words around them for context.

Asher told me Cogito Semantic Advertiser understands content based on four key methodologies: studying the morphology of words; looking at parts of speech; sentence logic, or the reduction of sentences to subject, verb and object; and disambiguation, which in the case of the jaguar story paired with the Jaguar car ad would determine whether or not the text referred to a car or an animal.

Cogito Semantic Advertiser also provides details on actual use of specific Web sites, so companies can tailor their messages and coordinate the timing of ad placements.

For example, if users are reading articles about fuel efficiency during a specific time frame, Cogito Semantic Advertiser can provide these details so a hybrid vehicle manufacturer can tailor its advertising budget accordingly.

Asher told me Expert System plans to sell Cogito Semantic Advertiser, priced between $100,000 and $200,000 depending on implementation size requirements, to big media houses and publications.

I like what Expert and others of its ilk are trying to do in terms of the technological approach. Indeed, finding AdSense ads on vacations next to articles on Hurricane Katrina has got to be unsettling.

However, the big question is whether or not big media houses and others that would buy Cogito Semantic Advertiser are sick enough of Google yet to go with an unproven company.

Like it or not, Google is to search keywords what Microsoft is to desktop operating systems. If you go for Google, you have to know that you're at the company's mercy. Are advertisers and publishers ready for an alternative? That remains to be seen.

9 Questions on Googleeeee

You got your East Coast news. You got your West Coast news. Something about the cratering US financial system going on out there on the isle of Manhattan, sources tell me. Meanwhile, here on the Left Coast, it's round 74 of Google (GOOG) and Yahoo (YHOO). GooglyHoo is giving lots of people a case of the hives, an itching reaction in search of a rash.

The latest scratcher is the World Association of Newspapers. Yesterday, it denounced the proposed Google/Yahoo search "cooperation" deal as anti-competitive. For good measure, its statement released lots of frustration about newspaper companies' diminished standing in this new world order in creation. In part, WAN points out that of the $48 billion in online advertising revenue that Google has collected since 2001, less than one-third of that has been shared with online publishers. Those big numbers are of course the ones that hurt, more than the cost-per-click impacts of GooglyHoo.

So nine quick questions on the boiling Google/Yahoo cauldron:

1. Who gave the pile-on signal? Now, according to Bloomberg, the EU is joining the fray, asking for a few Yahoo and Google documents (no, not Google Docs). That makes the Department of Justice, eleven states and the EU. No word yet from Bruce Sherman.

2. Why is the inquiry only about Google's search dominance? Yes, it controls 70%+ of the paid search market, but its goals are clearly global ad dominance. It has made forays into print newspaper, print magazine and broadcast advertising. It bought YouTube, becoming a major video ad player. It bought DoubleClick, planning a major move into the display market. So on the sell side, it will be able to offer integrated packages of advertising -- a little search, a little display, a little pre-roll -- to ad buyers.

While today, much of advertising buying is segmented by type, I've got no doubt that there's a Starship Enterprise console out there in the ad buyer's future, with audience targetable, using various types of advertising through a single interface. Without legal roadblocks, today, you'd have to bet that the console would be branded "Google." Shouldn't DOJ ask P & G (PG), GM (GM) and Walmart (WMT) (all companies that have criticized the Google/Yahoo proposed combo) about Google's wider ad role?

3. Didn't Joe Nocera nail it in his Saturday New York Times column, describing the experience of one company, Sourcetools, as it first won big and then lost big in the Google ad world? The AdSense/AdWords stuff makes so many heads hurt; telling the story (journalism!) through one company's experience is a great analgesic.

4. Does it help or hurt newspaper companies? That kind of depends on whether they are bigger buyers of AdWords or bigger displayers of AdSense. There's little doubt that the further monopolization of paid search will lead to higher pricing -- there's not sufficient alternative inventory to buy of significant scale. So if you are buying AdWords, they should cost more.

But if you're a big AdSense partner, like the New York Times, your share of the take should increase as well. We don't know the particulars of each affiliate deal, but would hope that newspaper companies could get a fair packaged deal from Google. And yes, having Yahoo out there as somewhat of a paid search competitor, has made the chances of getting a better deal better.

5. Will it make a big difference to Newspaper Consortium members? Apparently not much. The Consortium members, almost half of US papers by circulation, take Yahoo search as part of their wider participation. For their participation, they get a contracted minimum payment, and sources say that earned ad payments haven't reached the minimums yet, generally. So, if GooglyHoo does increase pricing of ads, earned revenue should increase, but wouldn't result in actual new dollars falling into newspaper company pockets, at least for awhile.

6. The US Newspaper Association of America is a WAN member, but has been so far quiet on GooglyHoo. Is that because some US publishers think they'll be winners out of a GooglyHoo tie-up, some think they'll be losers, and some just don't know?

7. Isn't this more about the transfer of ad wealth than pricing? If you look at the big numbers, the US news industry was down $3 billion in the first half of 2008 compared to 2007. You can make the big-picture case that that most of that money is going to Google, Yahoo, MSN, AOL and a few other non-newspaper-owned places on the web. From my reading, it looks like newspaper companies used to pull in about 20% of the national ad pie in the US, pre-web, a percentage that of course is dropping annually. More significantly, newspaper company share of internet advertising is no more than 15%, and probably closer to 10%, if we were to untangle bundled ads. Yes, cpc pricing is at the base of this, but it's this larger transfer of wealth that's behind the current fuss.

8. How much of WAN's angst has been caused by Google's lukewarm response to ACAP? UK and European publishers have pushed ahead with their Automated Control Access Protocol, a program to better control and protect editorial content as it moves through the web ecosystem. In theory, publishers can express "terms and conditions" for use of their content by web bots. As with all such systems, they only work if they are near-universally used. And if the biggest search company in the world won't play, ACAP can't really catch on. Google keeps talking with ACAP, but won't commit.

9. How much of a competitor is Yahoo anyways? Jonathan Weber, who publishes the excellent New West site, makes a compelling case on Times Online today that Yahoo has long faded an as effective competitor.

-searchspiderz

Monetising Newspaper Content

Google has partnered with about 100 newspapers across the US to digitise them and make scanned copies available online, and to eventually cover newspapers worldwide.

NEWSPAPER sites around the world don’t make much money. With the exception of a few niche sites like the Wall Street Journal (WSJ.com), the overwhelming majority of newspaper sites give away the content for free.

This includes all the newspaper sites in Malaysia. Some considered charging in the early days but with the trend of “content wants to be free” pervading over the Internet, none has tried to charge a fee for news content.

Some newspapers do charge for archives but that doesn’t make much money. Few people bother to use the archives. Usually, the ones who pay for it are those who are looking for specific articles for research purposes.

Since this is not exactly mass consumer stuff, many newspapers find that it’s not really worthwhile to digitize their much older material. That typically requires either having people actually retype all the copy or using some scanning service to convert the printed articles into text. The problem with that is the converted text is never accurate and an editor will still have to clean it up to ensure accuracy.

Another innovation: Catering waiters working in front of a logo of Google at the Frankfurt Book Fair in this Oct 21, 2005 file photo. Its business model, as always, is to provide the service to consumers for free and for it to make money through its AdSense advertising service. — AFP

It’s a tedious process and the cost of doing so could never be recouped due to the relatively low demand for such archived materials. So, it often doesn’t get done, especially if it’s a newspaper that has a long history. Imagine doing this for a paper that dates back over 100 years or more.

Along comes Google with its “free” strategy supported by ads. Google’s business model all along has been to give all its products and offerings for free. And it is doing so again with this latest effort.

The search giant has partnered with about 100 newspapers across the US to digitise them and make scanned copies available online. Eventually it will cover other papers around the world.

“This effort will enable us to help you find an even greater range of material from newspapers large and small,” wrote Google product manager Punit Soni on the company blog. “This effort is just the beginning. As we work with more and more publishers, we’ll move closer towards our goal of making those billions of pages of newsprint from around the world searchable, discoverable and accessible online.”

Others from Google are lauding this as a real big deal. “This is huge,” said Google vice-president of search Marissa Mayer at the TechCrunch50 conference in San Francisco earlier this week. “We’re branching into a new form of content.” In her demonstration, Mayer showed off pages from the Rome News Tribune and called up a story covering an American moonwalk.

The way it’s done means users will see entire pages of the original newspaper as they were printed at the time. This kind of approach, of scanning whole pages, would actually make old content more interesting. It shows the relevant story and other headlines, adverts and promotions of the day giving a sense of what it was like at the time.

“The compelling part of the product for me is to get a sense of context and the importance of what else happened that day,” Mayer said.

You could easily imagine people looking up newspapers to find out how life was like and what happened on a particular day in time. Perhaps on their birthday. Many people would like to know about the significant events on their birthday.

The technology involved is not rocket science. In fact, it’s very similar to the scanning technology that is used for Google Books and it expands on an initial two-year effort by the firm to work with two leading American newspapers, the New York Times and Washington Post, to index old papers in Google News Archives.

Google, of course, is not doing this for charity. Its business model, as always, is to provide the service to consumers for free and for it to make money through its AdSense advertising service. Revenue earned will be shared with publishers.

Initially, the archives will be available through Google News, its popular news links aggregator site, but the company plans to give newspapers a way to make their archives available on their own sites. Will this be a new source of significant revenues for newspaper websites? I doubt it’d be a big income earner. But it’s one additional revenue stream.

And if a newspaper’s archives are being converted by Google for the newspapers for free, that’s not such a bad thing for the newspaper companies even if they don’t make a lot of money from it.

Oon Yeoh loves the idea of news archives. Check out his archives at www.oonyeoh.com.