Exploring Local
Mike Dobson of TeleMapics on Local Search and All Things Geospatial

Google’s Broken PlaceFinder

April 6th, 2010 by MDob

(Note from Mike – I’ve been busy with work on a contract and not so good at writing new entries for Exploring Local. This one turned out to be quite long, but there are a bunch of pictures to speed it up. So, grab a cup of Joe and take a gander. Or, read some now and some more the next time you think I should have written another blog. I’ll try to ratchet my schedule up a bit – my apologies.)

Yesterday I was browsing around the Web to see what happened at the Where2.0 show in Santa Clara and came across a presentation by Michael Jones, formerly the CTO of Google Maps and Google Earth, now Google’s Chief Technology Advocate. Mike spoke on the following topic “The New Meaning of Mapping.” His presentation can be found here for those of you interested (for those of us who follow Mike, it’s not one of his best presentations, but he does reveal some interesting use statistics).

Part of the presentation was focused on the use of the map as a “place browser”. I chuckled, when I saw this, since it seems that one cannot use Google Search or Google Map Search for place browsing, at least, not if you actually want to know where places are located. However, if you are looking for a good afternoon of misinformation, it may meet your needs.

I got started on this topic because of Mike Blumenthal. In his blog on Understanding Google Maps & Local Search, Mike has been probing the mysterious disappearance of towns from Google Maps. Mike sent me a note on additional mysterious disappearances and I thought I would look at a couple to try to understand the problem. It’s a doozy.

Missing Town Number 1

When I searched for Imperial Beach, CA, (a San Diego beach town) using the Google Search Bar (not the Google Map Search Bar), the inset map shown in the SERP (Search Engine Results Page) is centered on Lemon Grove, CA, which is 20 miles to the northeast of Imperial Beach.

Map in Google Search showing the incorrect location of Imperial Beach, CA

I thought there might be a possibility that Google Search would find Imperial Beach under the name “City of Imperial Beach, CA, which is one of the names for this city in the GNIS (Geographic Names Information System).

When I search for City of Imperial Beach, CA in the Google Search Bar, the SERP shows a map at the bottom of the page, but it is the same map of Lemon Grove that I found when I was searching for “Imperial Beach, CA”, so the name form seems not to make a difference in Google Search.

Faced with the same map reference from the two searches, I clicked the inset map on the City of Imperial Beach SERP, and was taken to a larger map centered on Lemon Grove, but the left panel of the Google Maps page indicated that the Red map pin was actually Imperial Beach, CA, although the photos of the area were from Lemon Grove.

Imperial Beach in the Info Panel, Lemon Grove on the map?

I requested a route from Imperial Beach to Lemon Grove (knowing that the two locations were distant), but the result was a bust as there was only one maneuver suggested – staying put. It would appear that Google Search and Maps are convinced that Lemon Grove is really Imperial Beach, or at least Google’s place name lookup is convinced of this false geography.

Well, truth be told, I already knew where Imperial Beach was located and after having moved the map viewport to the Imperial Beach area, I placed my cursor on Imperial Beach Boulevard and clicked on the “what’s here” option. The results were hilarious.

Apparently, the data point I selected is not recognized by Google as a location within the city of Imperial Beach, since the Google Info Panel described the location only as (within) “California”.

Could it be that the polygon file for Imperial Beach is incorrect? It would appear that the point I marked was not recognized as occurring within the polygon for Imperial Beach (or Lemon Grove) based on the information returned. (Yes, I know that I was sampling a raster map, but the transformed coordinates need to be analyzed as a point in polygon in order to determine the appropriate political geography and other attributes available to describe the geography visible in the map viewport).

Curiously, Google Maps does not seem to recognize Imperial Beach as a location in San Diego County either. When I queried, “What’s here” on different locations along Imperial Beach BLVD, the only geographic qualifier returned by Google Maps was that I was in California. However, at each location, the Info Panel on the left side of the page changed to show a different set of “nearby” attractions.

For the first location on Imperial Beach BLVD I, the info panel indicated that nearby locations included the Aquarium of the Pacific in Long Beach, CA, then Grauman’s Chinese Theater* in Hollywood and LegoLand, about 40 miles north of Imperial Beach. I moved the cursor a little to the left and was informed that “What’s here” included the Westin Bonaventure Hotel in Los Angeles and Hollywood and Highland (an intersection in Hollywood). I moved the marker a third time (see below) and the new nearby attractions were the Santa Cruz, CA Boardwalk, Two Embarcadero Center in San Francisco and the Los Angeles International Airport. The accompanying photos of nearby attractions in the Info Panel include Grauman’s Chinese Theater (Hollywood), The San Francisco Palace of the Legion of Honor, the San Francisco Museum of Modern Art and the Getty Center-Los Angeles.

The map point is in Imperial Beach, but the points of interest are hundreds of miles away

What a fantastic Place Browser Google has invented. No wonder Mike Jones was speaking at Where2.0 on the “New Meaning of Mapping.” Google does seem to have a unique approach to geography – and an inability to fix what ails the company’s mapping and search engines.

Today when I tried to replicate these results, the page indicated that all locations along Imperial Beach BLVD were located in….Lemon Grove. Obviously, Google is trying to fix this problem, but they do not seem to have a clue about how to accomplish the goal.

When I moved my cursor to the location of the Imperial Beach Plaza Shopping Center, Google maps now told me that I was pointing to Imperial Beach Plaza Shopping Center in … Lemon Grove. Is this the worst or what?

When I viewed the area of Imperial Beach on Google Maps and used Google’s Street View, the images were labeled Imperial Beach, CA, even though the Info Panel provided the erroneous information that this was Lemon Grove, CA.

Street View is tagged with the appropriate Geo Location, even if the Info Panel thinks this is Lemon Grove

When I searched for Imperial Beach using the Google Map Search Bar, Google provided a disambiguation list that included locations labeled as occurring in Imperial Beach, CA. Clearly, the Google Maps database includes information about Imperial Beach – just nothing specific about the location of its geography, although the disambiguation lets us know that the database included within the bounding rectangle of the map window may contain information on the locations listed.

The disambiguation list includes all sorts of places within Imperial Beach

At this point, I had a headache from trying to figure out where Google screwed up. So, egged on by Mike Blumenthal’s email, I entered Woodstock, VA in the Google Search Bar, not the Google Maps Search Bar.

The map that appeared was particularly unhelpful, since it did not include the name Woodstock anywhere on it.

Just where is Woodstock on the map?

I clicked the map and was taken to Google Maps and a page that showed a map of Woodstock – great – oh wait – it was a map of… Woodstock, Georgia.

Oh, I searched for Woodstock, V-A, but this is Woodstock, G-A

If you click Woodstock, Shenandoah, VA on the disambiguation list that appears on the Woodstock, GA map page you are taken to a map of Woodstock…Georgia. How cool is that?

Now, if you search for Woodstock, VA 22664 using the Google Search Bar, the inset map in the SERP appears to show the correct location,

Now that looks like the Woodstock in Shenandoah County, Virginia

but when you click this, map it takes you to Google Maps page showing a location near scenic Nassawdox, Virginia on the Delmarva Peninsula

Where is this? - it certainly is not the area clicked in the last map.

For those of you who are not familiar with the Delmarva, the location is shown below. It is about as far as you can get from Woodstock, Shenandoah, Virginia and still be in the right state.

So, now Woodstock in Shenandoah County, Virginia has moved to the Delmarva Peninsula?

Wait, there’s more

Well, I don’t have the energy to chase all of the place name-location errors in the Google Mapbase, but there appears to be plenty of them and they have been hanging around for several weeks, if not longer.

I find myself wondering if Google has imported a commercial database of neighborhoods that has confounded the boundaries in whatever Google uses for a place-polygon file. I am presuming that the bounding border polygons are associated with place names in some manner, but do not have any real information on the file structure Google uses as a container for their data on geography. Finally, I saw some non-linearity in the results. I do not know if this was contextual learning of the part of the Google Search engine or possible differences in their map databases between server farms.

It seems to me that that Google’s apparent geographic illiteracy is a screw-up of major magnitude. Advertisers are not getting their money’s worth because their ads are showing up in the wrong geography. People cannot find the places to which they want to navigate, because Google cannot tell them where the location is on the surface of the earth. Communities that rely on Google Maps to help people find and navigate to their city facilities cannot depend on the maps they serve to show the actual location of the city.

Of course, Google does not apologize or even refer to the problem and these gaffes do not seem to influence their success. Is that a great business model or what”? My good friend Marv White, speaking of the Internet in the 1990s, said “It’s not real yet.” Guess Marv was prescient, as it seems to be true, still.

In any event, we need to rub some salt in Google’s wounds. NAVTEQ has both places correctly located in their database (as does MapQuest). Somewhat curiously, Bing Maps seems to have a problem with “Imperial Beach, CA’ (see below), although they did find Woodstock in the Shenandoah Valley of Virginia.

Hmm, Bing could not find Imperial Beach either.  Must be some form of boycott.

Maybe Google should hire some geographers? Nah, all you need is Algorithms. Well, maybe you need algorithms and somebody who understand how they should be associated. Perhaps that’s no longer in vogue at Google?

Finally, it appears to me that having Mike Dobson write about how good your mapping system is going to be apparently is the kiss of death. It appears that both Google and I believed what I wrote about them. Maybe I should have more faith, since I think Google knows that something is wrong, but apparently has no idea how to fix it. Maybe those 500,000,000 User Generated Content corrections for their maps mentioned by Mike Jones include the solution to this vexing problem. Alternatively, maybe somewhere among those 500,000,000 user generated changes you could find the source of the problem.

Click for contact Information on TeleMapics

Bookmark and Share

Posted in Authority and mapping, Bing maps, crowdsourced map data, Data Sources, geographical gazetteers, Google, Mapping, Mike Dobson, Navteq, User Generated Content, where20

6 Responses

  1. Phil

    Its worse that what you say above.

    Just checked Google “My Maps”, takes longer to load, so they may
    have a problem, but don’t know if its local or wide spread?

    Almost everything there that I had saved, is now gone.

    Might want to check your “My Maps” if you have any?

  2. Another Phil

    Seems to be general – have tried from 3 browsers, on two platforms in two locations, and get the same issues today. Basic HTML is OK though.

    Thanks Phil. Are you seeing this problem with locations in the UK (presuming the UK is your location)?
    Mike

  3. Andrew Smith

    I have a similar problem with Middletown Md.We are an adwords advertiser and also a listing on local maps,but the actual location comes up like 60 miles away, no answer from Google..

    Ouch! I hope Google is able to resolve this for you soon It is one of those “hard to imagine” problems.

    Mike

  4. Mark

    I’ve noticed this too ever since Google went off the reservation. Many place names in NYS are out of whack, off by 20+ miles. It seems that unincorporated areas have been hit hardest in the East.

    Keep up the New World Order of Mapping, Google…Tele Atlas and Navteq are thrilled!

    Thanks, Mark.

    I would guess that NAVTEQ and TeleAtlas are thrilled. Unfortunately, NAVTEQ, on its own mapping application at http://www.navteq.com, still has not managed to correct the re-alignment of Highway I-195. While Teleatlas, at the TomTom Route Planner (www.routes.tomtom.com), does route correctly on the re-alignment, it still shows the section of the higway that is being torn down. I suspect that they fixed the error in their routing engine, but have not yet fixed it in their pre-rendered map database. However, both companies do seem to know where all those towns lost by Google are located – so good for them.

    This controversy isn’t doing a lot of good for Google’s corporate mission, which is “To organize the world’s information and make it universally accessible and useful.” Maybe they should change their name to “Garble”.

    Mike

  5. Duane Marble

    Well, Google found Florence, Oregon correctly. However one of the
    photographs it displayed showed a river with multiple bridges. We
    have one rather nice bridge. Ah! Florence, Italy!
    After many years of arguing for full transparency with respect to
    spatial data, I am dismayed that the neocartography movement has
    swung back to the assumption that digital spatial data = truth.

    Thanks, Duane.
    Mike

  6. Duane Marble

    I see that I neglected in my previous posting to remind all of us of
    Edgar M. Horwood’s Second Law of Data Processing and Information
    Systems: Bad data drives out good.
    Also, his Fourth Law would also seem to apply: The respectability
    of existing data grows with elapsed time and distance from the data
    source to the investigator.

    Duane – Thanks, again.
    Mike