New Meetinghouse Locator on LDS.org - Beta Release

Discuss ideas and suggestions around the Church website.
Post Reply
RossEvans
Senior Member
Posts: 1345
Joined: Wed Jun 11, 2008 9:52 pm
Location: Austin TX
Contact:

#31

Post by RossEvans »

borenmt wrote:Concerning the meetinghouse locations, we rely on latitude and longitude because matching to street address will result in too many "not found" situations, especially internationally. Even in the US, a match rate of 80% or so is common, leaving 20% not found. For those that are found, the algorithms for placing an address at the correct point along a stretch of road is always approximate, and can result in errors of many feet or even miles in rural areas. Algorithms can also not reliably place the meetinghouse on the right side of the street, not yet anyway. So... we rely on local employees, clerks, and members to help us gather accurate lat/long coordinates.!

I am a little surprised at the 80% match rate for meetinghouses. Are the 20% problems concentrated in rural locations? For cities and suburbs, I would expect the rate to be much higher, approaching 100%, especially since the church presumably has clean street-address data for its meetinghouses. (Perhaps the geocoder is tuned to be too reliant on postal address validation, and meetinghouses typically do no receive mail service, even though their street addresses are perfectly valid on the local street map?)

If there are this many problems mapping a few thousand meetinghouses with good addresses, imagine the problems developers will face trying to map a few million members with the relatively dirty street-address data harvested from MLS.

As for the solution of using clerks to send in corrected coordinates, I think that is generally doable. But this beta site does not come close to mobilizing the army of clerks. Only a tiny fraction are even aware of it. I think you need to broadcast a message through the MLS system, explicitly tasking the clerks to validate their locations and giving precise instructions on how to capture correct coordinates if need be and communicate them back to you.
User avatar
Mikerowaved
Community Moderators
Posts: 4734
Joined: Sun Dec 23, 2007 12:56 am
Location: Layton, UT

#32

Post by Mikerowaved »

RussellHltn wrote:Do you have any details on how this was done? Was it the FM group?

But there's still an issue with GPS coordinates - if the spot measured happens to be closer to another street, then the directions are going to route the user to that street - even if there's no driveway there. In the case of the chapel on 1515 SW Cherry Park Rd, TROUTDALE, OREGON; the GPS coordinates may be "correct" in that it's on church property, but just too far back from the main entrance and it's created erroneous directions.
Russel brings up a valid point. Perhaps custom geocoders should be encouraged to favor pinpointing the main driveway leading to the building, rather than targeting the center of the roof. I remember one case that I used strictly lat/lon to find a remote location (Amateur Radio Field Day site). Having them target the entrance of the dirt road leading off the main highway was far more helpful than if they had geocoded the actual site itself.
So we can better help you, please edit your Profile to include your general location.
User avatar
borenmt
New Member
Posts: 31
Joined: Mon Sep 17, 2007 10:58 am
Location: Utah (Wasatch Front)
Contact:

#33

Post by borenmt »

As long as new streets are being paved and new homes (and chapels) are being built, address matching will be incomplete. My neighborhood didn't show up on maps until a year ago. So although matching is tougher in rural areas where data may be incomplete, it's also a problem in suburban areas where there is growth.

On another topic referenced above, I think I remember fuzzily someone telling me that the goal was to put the marker on the steeple or part of the chapel nearest the road... but I could be wrong. If the biggest data problem we have is that we are on the lot, but not on the right part of the building, I will be very happy :-)
User avatar
garysturn
Senior Member
Posts: 606
Joined: Thu Feb 15, 2007 11:10 am
Location: Draper, Utah, USA
Contact:

Zip Codes

#34

Post by garysturn »

I have noticed that if you leave off an address and just enter a City, State and Zip Code you can get strange results. Zip Code pin points are not consistant with what you get with just a City and State without a Zip Code. Try Fillmore Utah then Fillmore Utah 84631 and see the difference. You get a pin point in the middle of the desert when the zip code is included.
Gary Turner
If you haven't already, please take a moment to review our new
Code of Conduct
RossEvans
Senior Member
Posts: 1345
Joined: Wed Jun 11, 2008 9:52 pm
Location: Austin TX
Contact:

#35

Post by RossEvans »

Mikerowaved wrote:Perhaps custom geocoders should be encouraged to favor pinpointing the main driveway leading to the building, rather than targeting the center of the roof.

That should be configurable in the geocoder, by the application implementer at runtime.

I don't know of any geocoders that actually target rooftops,despite the misleading moniker of "rooftop-level" geopcoding. This usually is based on parcels from county tax records, linked back to street addresses. The geocoder then reports a centroid for the parcel. For small single-family lots, this is likely on the house. For sprawling apartment complexes or five-acre spreads, it is somewhere else.

And the coverage of parcel-based geocoding is quite incomplete, even in the United States.

Having recently experimented with commercial parcel-level geocoding for our ward -- which produced only a handful of parcel-based coordinates for apartments mixed in with the more typical coordinates interpolated from street-address ranges -- I concluded that it was more confusing than not. It is easier to explain dots on a map that are uniformly offset a few feet back from the road.
russellhltn
Community Administrator
Posts: 34418
Joined: Sat Jan 20, 2007 2:53 pm
Location: U.S.

#36

Post by russellhltn »

If I'm not mistaken, the geocoders in use simply plot things based on the idea that "if 1500 is at that intersection and 2000 is there, then 1750 must be half way in-between on the even side of the street."

I suppose there might be a database that will take an address and do look up on that parcel, but it would be a big honkin database. (But then computers have become so big and fast, I could be wrong.)

I'll have to look at Troutdale more closely, but the coordinate might be for the FM Group's shed - great to tell the new guy where to find the tools, but not so good when "directions" point you to a driveway that doesn't exist.
RossEvans
Senior Member
Posts: 1345
Joined: Wed Jun 11, 2008 9:52 pm
Location: Austin TX
Contact:

#37

Post by RossEvans »

RussellHltn wrote:If I'm not mistaken, the geocoders in use simply plot things based on the idea that "if 1500 is at that intersection and 2000 is there, then 1750 must be half way in-between on the even side of the street."

I suppose there might be a database that will take an address and do look up on that parcel, but it would be a big honkin database. (But then computers have become so big and fast, I could be wrong.).

The interpolation method you are describing is the way most geocoding processes work. However, there are leading-edge geocoders that also incorporate parcel-level databases, which have far from complete coverage but are being expanded.

The geocoders used in the church's application may be the former, or the addresses you are testing may have no parcel-level coverage. (I do note that the address of my father's house in El Paso TX, which most interpolated geocoding sytems miss by a noticeable distance, is mapped precisely in the beta app. But my own address in Austin is obviously interpolated.)

In any event, if the capability exists here, that would be an internal property of the Google Maps and Microsoft Virtual Earth systems employed by the church, which possibly are also controlled by the API calls made by this application.
RossEvans
Senior Member
Posts: 1345
Joined: Wed Jun 11, 2008 9:52 pm
Location: Austin TX
Contact:

#38

Post by RossEvans »

I just looked at Russell's example in Troutdale OR. The only problem is with the street directions. If the icon were mapped closer to the curb of the street -- the street with the main address -- this problem would not occur.

Obviously, when generating the directions, the back-end has to jump from the lat/lon coordinates to a street, and compute directions from there.

I don't know how those coordinates got into the database, but if they were produced by parcel-level geocoding, which then stored a parcel centroid, that could explain how this happened.

This is similar to the problems I experienced when experimenting with parcel-level geocoding of my own ward's addresses. It can produce more confusion than useful information, as Mikerowaved commented above. I think the ideal use of parcel-level geocoding is if it can be confugured to resolve addresses precisely along the main street -- a significant improvement over interpolated addresses. But I think all addresses are best plotted at a modest, uniform setback from the street.

I don't know if this is possible using the APIs to the two services (Google Maps and MS Virtual Earth) employed in the church's app.
rmrichesjr
Community Moderators
Posts: 3827
Joined: Thu Jan 25, 2007 11:32 am
Location: Dundee, Oregon, USA

#39

Post by rmrichesjr »

borenmt wrote:...
So... we rely on local employees, clerks, and members to help us gather accurate lat/long coordinates. You're right that many need to be improved; please let us know via the feedback link when you find a meetinghouse out of place or missing. You can even click on the right spot with the Place Marker tool to send us the corrected coordinates. So far during the beta period, many, many people have submitted corrections which should soon be incorporated into the database, once they are verified. I hope you submitted the problems you found yourself through the Feedback mechanism.

Thanks again everybody for your comments!
Are you also looking at reports of incorrect meetinghouse location (or directions) on the old meetinghouse locator? I think it was at least two years ago, perhaps three years ago, that I reported that the location of the Newberg, Oregon meetinghouse was incorrect. I was disappointed to see the same location error in the beta. I did report it again.
RussellHltn wrote:...
I'll have to look at Troutdale more closely, but the coordinate might be for the FM Group's shed - great to tell the new guy where to find the tools, but not so good when "directions" point you to a driveway that doesn't exist.
You might be on to something there. The (incorrect) location for the Newberg, Oregon meetinghouse is pretty close to where the storage shed is sitting on the far back corner of the lot.
russellhltn
Community Administrator
Posts: 34418
Joined: Sat Jan 20, 2007 2:53 pm
Location: U.S.

#40

Post by russellhltn »

Checking the aril photo for Troutdale, the location is a corner of the building - it's what I'd expect if someone used a hand-held GPS since they usually don't work too well inside. ;) However the corner marked is a back corner away from the street.
Post Reply

Return to “Main Church Website”