September 27

Contents

Editor's Introduction

This week, I report on a new service by MetaCarta for the on-line publishing industry. Plus, 25 press releases.

— Matteo Luccio

MetaCarta Launches Local Alerts

MetaCarta specializes in searching for geographic content in unstructured text documents and displaying the results on a map. So far it has focused on large government and commercial clients, such as intelligence agencies and oil companies. Now, it is moving closer to consumer applications, by assisting customers that are serving the mass market. This week, the company launched Local Alerts, a service for the digital publishing industry that notifies users when free-form text content mentions specific places they deem important. Users register for the service, provide their e-mail address, and specify an extent of interest, using a map that they can zoom and pan; they are then notified by e-mail daily of any items in the covered content that mention places within that extent.

MetaCarta uses its core technology to geoparse and geotag the stories in real time and puts them in a database, which also stores the information on all of the registrants and the max-min lat-long of their area of interest. The company then maps out all the stories and alerts users to any stories that contain geographic references that fall within the range they have specified.

The first publication to sign up for the alert service is MySanAntonio.com, which is the on-line edition of the San Antonio Express-News newspaper, a Hearst Publication. To test the service, go to the site, then scroll down until you see an orange "CONNECT" logo.

For digital publishers, the main advantage of the service is that it draws readers back to their sites, "substantially increasing user frequency and page views metrics," MetaCarta says, and "enabling publishers to better capture local advertising dollars and maximize advertising rates." For users interested in news, listings, promotions, etc. pertaining to a particular neighborhood, city, or area, this service obviates the need to search for it in a publication covering a much larger area. I presume that the value of this service to users will increase as the number of participating publishers increases: then you will be able to sign up to receive alerts when your area of interest makes the news in regional, national, or international publications.

Clearly, this service is not of much relevance to some kinds of digital publications. Take GIS Monitor, for example. Even though this newsletter is all about geography, I rarely mention place names in my articles (as opposed to the "News Briefs") and, when I do, it is mostly to indicate where a company is based or a conference is taking place. Therefore, the location is just one more (alternative/additional) piece of information you can use in our site's search engine to find an article.

I discussed MetaCarta's announcement with its VP of Content Services, Rick Hutton. "We made our name selling to a wide variety of government agencies," he told me, "and the last couple of years we've established a rather large footprint in the oil and gas business. For the past year, we've been working on bringing products and technologies to the digital media market, to see how on-line publishers could use it. We realized that what they needed was applications and solutions. The MetaCarta Local Alert Service is an on-demand service, or software as a service model. It is enabling the Web site MySanAntonio.com to allow their users to register to receive local alerts about neighborhood-level news. Depending on the schedule of the customer (in this case, MySanAntonio.com), the alert service can run once or twice a day."

1. How do local alerts help on-line publishers?

   It allows them to target users and, hopefully, pull them back to the site — it is what is commonly referred to as a stickyness factor in an advertising-supported Web site — and then serve advertisements to them based on the geography that they are interested in. Since we also produce the metadata for the geography, it is very simple for publishers to pass that geographic metadata to their ad server and begin to sell localized ads.

   Newspapers, in particular, have an interesting challenge: they are extraordinarily profitable in most cases but their revenue numbers are sinking and are eventually going to cross below the profitability line. Their business is inarguably and irreversibly moving on-line. They need to figure out how to provide things on-line that they were able to do in the offline world, such as local editions. In the on-line world you don't have to do it on a global basis for all the people that live in a ZIP code or a town. You can do it on a one-to-one basis and our technology allows them to do that.

2. Does you system search the content on your customers' sites or do they feed their content to your site?

   We don't crawl their sites, because that is inefficient — you pick up too much noise. Typically, any site we would deal with has some type of content management system. When they publish stories to the site they simply also publish them out to a Web service to us. We pick them up, process them, and stick them in a database. Then, on a schedule that they determine, we run the process of matching up the stories with the registered users.

3. Why not just geo-tag each story using its dateline?

   It just doesn't work in the news business. First of all, it is a single geographic reference and, second of all, it is often-times inaccurate: for example, a reporter in Boston talks about an IBM executive from New York that was kidnapped in Mexico. Also, many stories are rich in geographic references, including ones that are rather implicit, and our technology picks them up better than any other technology on the market today.

4. As editors become more aware of this technology, will they start to geographically tag articles?

   We have talked to many, many on-line news publishers and many of them are asking the people in the field — the ones who are actually writing the stories — to provide them geographic metadata, but it just doesn't happen. There are not many journalists out there who are writing stories for big, on-line entities that did not come up through the standard ranks of off-line journalism. Over time — and this will probably take a decade or so — you will start to get journalists that came up through the ranks of the digital media and are much more attuned to working in an electronic environment and in putting in metadata.

5. Why did you start with MySanAntonio.com?

   MySanAntonio.com was simply the first customer that we were able to sign up for a pilot. When we were bringing out technology to the digital media realm, we were evangelizing it to a large degree. On-line publishers said "Who's using it?" and we said "Only people in the enterprise and Shell, Chevron, Hess, etc." We finally figured out that the newspapers were the ones that had the most need for something that could solve personalization and geographic targeting, which is something for which there is a big thirst in the advertising realm on line. MySanAntonio.com was just the one that had the vision to go first.

6. Can a user set multiple extents?

   Unfortunately, not at this time, but that is going to be a version 2-type of feature that we are going to introduce.

7. What is your rate of false positives?

   I don't have the number of false positives and the reason is that we have a rather high f-measure — a combination of precision and recall — as to how we are actually able to recognize place names and then the precision with which we are able to match them up with exact places. So, if a document says Fenway Park, our ability to recognize Fenway Park, geo-parse that out, and match it up to the Fenway Park that is at a specific latitude and longitude in Boston, Massachusetts, is very, very good. Our f-measure is above the 90th percentile. When we create geographic metadata, we generate something called the geo-confidence score: that is the degree of confidence with which the system believes it hit the target 100 percent. The publisher can use that as a sliding scale. If he is getting too many false positives, he can set the confidence score a little bit higher. He'll have less recall and miss a couple of things now and then, but he'll get fewer negatives.

8. That's similar to the way spam filters work. Also like spam filters, you can go into a folder that contains rejected references and find false positives…

   Yes, that is a very good analogy.

9. What about using live, GPS-based tracking?

   While we can't today figure out where the user is and do it that dynamically, that application could be created in the future by leveraging our technology.

10. There are plenty of systems that do the tracking part. Then you'd just have to feed that data stream into your system. It would be like updating the extent of interest every few minutes.

    Exactly.

11. How does this new service fit into your general strategy?

    Local alerts is the first of a couple of applications that we are building on top of our core technology and providing as a hosted service. Our return on investment argument is extraordinarily strong, because there is low upfront cost and people can pay us on a transaction basis or even on a revenue share basis.

12. Are you starting to move into the consumer arena?

    We are moving in to service customers that are serving the mass market. We are still a b-to-b play, but now we are moving into a much larger horizontal opportunity marketplace, where we are serving places like MySanAntonio.com that have something like one and a half million users per month. It is a different business model, but we are still serving the business and they are serving their customers.

» Back to our GIS Monitor SEP 2007 Issue