A few of the workings behind the scenes at Google Local Search have been recently exposed through a patent granted to Google today and a pending patent application for the search giant published in January.
They might make you think a little differently about how Google Local Search works, and one presents an interesting question about the difference between local search results at Google Maps and Google’s Web search.
Categories and Local Map Listings
The granted patent was originally filed in December of 2004, and it describes the categories that some local searches are placed within by the search engine.
In my last post, How Google Might Top Search Results with Additional Information, I described a patent from Google that told us how the search engine might attempt to categorize the pages that show up in search results (presumably a top number of the results, such as the top 10, or top 100, or top 1,000), and show us an extract from a page taken from one of the results that matched the category that seemed to be the predominant one for those search results.
Presumably, showing that extract as a summary for the search results would help a searcher know if the search results for their query might be a close fit for the intent behind their search.
The categories used in that process might have been chosen by looking at the pages listed in the search results and their URLs, their page titles, snippets for the pages, and labels that may have been attached to the sites. If you’ve submitted a business listing to Google Business Local, you know that when you submit your business, you can add categories for it as well as other information.
The newly granted patent doesn’t detail very deeply how the local search categories they show are determined. Still, the method used to find categories for local search results may share some similarities, such as looking at URLs, page titles, and snippets for pages that show up related to businesses listed.
When you perform a search at Google local, in addition to a map and a list of businesses or organizations from a specific geographic location, you may also see categories listed at the top of your search result. For instance, if I enter “tires Philadelphia” (without the quotation marks), I see the categories “tires” and “auto repair” about the pages listed from that search.
We are told where some information about categories for local map listings may come from:
Category suggestion engine 420 may suggest one or more categories that relate to the search. In operation, category suggestion engine 420 may identify categories associated with the top N (e.g., 1000) documents in the list of search results.
The categories may be obtained from many different category providers, such as yellow pages and web directories, or derived using an automatic text classification system. A category associated with a document may be pre-stored with the document in a database associated with server 220. In this case, category suggestion engine 420 may identify the category by looking it up in the database. A document may have one or more associated categories.
Category suggestions relating to a search
Invented by Daniel Egnor and Elizabeth Hamon Reid
Assigned to Google
The United States Patent 7,523,099
Granted April 21, 2009
Filed December 30, 2004
Abstract
A system determines categories for business listings identified in a list of search results and assigns scores to the categories. The system presents one or more high-scoring categories as one or more category suggestions relating to the search results list.
If you listed a business in Google Business Local, how much time did you spend thinking about the categories that you assigned to your business? What other categories show the above search results when you see your site listed in Google Local Search results?
The What and Where of Local Business Search
The patent application published by Google in January takes on a slightly different problem.
When you type a query into a text box for a local search, you’re typing in the “what” and the “where” of your search.
So, a search for “jack in the box Stanford ca” is a search for “jack in the box” as the what, and “Stanford ca” as the where. Or is it? The search engine might see the “in” within the business name “Jack in the box” and interpret that as a separator between a what and a where, and decide to look for “jack” as the what, and “the box Stanford ca” as the where. The patent application tells us that problems like this can be addressed by looking at both potential breakdowns and seeing which seems to be the more obvious one based upon the number and quality of search results that each might return.
This patent filing isn’t mentioned because the search engine might have switched from having separate search boxes for a “what” and a “where” for a local search to a single search box for both.
Google will sometimes show information in their Web search results from other sources, such as their news results, images, videos, or local search results.
One guess is that in having the same type of searching format, a single text box, for entering queries for local search and Google’s organic search, Google might be provided with a better idea of when they might show local search results in their organic search listings when the search engine frequently sees a similar query in local search.
The patent application is:
Interpreting Local Search Queries
Invented by James Norris, Gregory John Donaker, and Nina Weiyu Kang
Assigned to Google
US Patent Application 20090019028
Published January 15, 2009
Filed July 9, 2007
Abstract
A search query may be interpreted as many possible interpretations, and each interpretation may be explored before the search results are sent to a user. In one embodiment, a device may split the search query into partitions. Each of the partitions may be submitted, as a search query, to search repositories. Confidence scores based on the results returned from the repositories may be used to determine a measure of confidence of the repository in the search query interpretation.
More interesting, the patent application tells us that sometimes people use local search but don’t include the “where” in their query. Instead, they type in only the “what” or the “where” isn’t defined very well. In that case, the search engine might look at other information to provide context for the search, such as looking at:
- A Google Local Search map currently being viewed,
- Geographic location information obtained from the user’s IP address,
- Geographic location information or other information obtained from a profile previously registered by the user with a local search engine,
- Information based on the search history of the user,
- Information based on the viewing history of the user,
- Information based on the language of the search query, and/or;
- Information based on the hostname through which the user accessed the local search engine
Oddly, if I search for “pizza” in Google’s Web search, I’m shown a map near my location with a box (ten listings) of local businesses that sell pizza. Suppose I type “pizza” into the search box at Google Maps, I’m shown a map with the United States centered in the middle, and results shown in the US, in Canada, in Mexico, in Central America, and even at the tip of South America. Google Web results seem to be inferring the “where” of my search, but Google Maps doesn’t.
Why would Google guess the “where” for a Web search and not a search at Google Maps?
Taking into account the ‘where’ for any search certainly makes the intentions of the searcher that much more complicated: whether the person doing the search specifies it or not, the ‘where’ dynamic adds yet another variable to the search equation.
I think Google is doing its best to interpret the searcher’s intentions at a more abstract or birds-eye level and learn from the local searches so they better understand the ‘where’ part of a person’s search down the road.
I am a little surprised how Google are suggesting they picking out the what and the where from queries; seems like a bit to much computation to me. From observation it seems obvious that Google analyse each word in the query. So I would have thought that they would analyse, “cheap plums in hull” and assign a weight to Hull, e.g. we think Hull is 50% a place since it is a city in the UK and also a noun to describe a part of a boat and fruit casings.
I’m basing this on using Google Custom Search. If you haven’t looked at it, it provides some interesting insights as to how Google works. Late last year they allowed you to define your own synonyms. So you can do things like:
<synonymentry word=”bill slawski”>
<synonym>guillaume slawski</synonym>
<synonym>guglielmo slawski</synonym>
<synonym>vilmos slawski</synonym>
</synonymentry>
@People Finder
I also agree that Google is doing the best job at trying to understand a searcher’s intentions. The implementation of the Local 10-Pack Update is evidence of that, where a user can view the Local 10-Pack of results even with a national search. This is based upon IP address.
Also, I thought that part of the algorithm in Local Search engines relied on zip. I.E. When you specify the “where”, it displays results according to the distance from the center of that particular zip code (as identified by the USPS)? Maybe I’m wrong?
Hi People Finder,
I agree. I do think that is why we now see a single search box on Google’s local search – so that comparisons between queries on Google Web search and Google local search are easier to make.
Hi David,
It looks like the language of query can play a role in determining the “where” of a local search. The list from the patent application that I included in my post of additional information that they might use for additional context points to some other information to determine a “where.” The language of the query can be part of that, but there may be other things that they may be looking at as well, including IP address and past search histories, etc. The process does involve a fair amount of computation.
The synonym information from custom search is interesting. Google may also be looking at other sources of information to help them with determining the “wheres” of searches, including dictionaries of place names.
Your example of “cheap plums in hull” might present some problems to the search engine, but the patent application does describe the use of some words such as “in” or “near” as separators that might help them distinquish a “what” from a “where.” That’s why they purposefully included the “jack in the box stanford ca” example, since it’s an exception to the use of the word “in” as a separator between a what and a where.
Hi Agent SEO,
Good questions. I’ll tackle the “zip” part of your question. Google local search doesn’t necessarily rely upon postal codes and presenting search results from a center point within a specific zip or postal region for a number of potential reasons. Some of them include:
1. There are geographic regions where geographical location information isn’t available because of trade restrictions or inaccurate and contrasting information from third party providers – see my post on Google Local Search in China: Export Restrictions, Filtering Sensitive Keywords, and Limited Data. In that instance, postal codes and even latitude and longitude information isn’t helpful.
2. The location sensitivity of a query may be different based upon the query used, whether the location is rural or suburban or urban, and a few other factors. A search for pizza in Manhattan might cover a small area with possibly many zip codes while a search for automobiles in a rural location might cover a much larger area with one or a few zip codes. So, a local search may cover one zip code area or a broader range of longitude and latitude within the boundaries of a map that Google shows, and the size of that map may depend upon a number of factors.
3. The Location Prominence of different results may determine which business is ranked highest within a map boundary determined by “location sensitivity.” While the distance from a center point may or may not play a role in which results are listed first, the patent filing describing location prominence includes a number of other factors that Google might look at.
There’s a section in my local search glossary goes into more depth on some of the factors involving location prominence. As I noted there:
Google local search is just a step according to search engine technology . Search engines are now trying to serve results depending upon users
Nice to catch up, Bill, and nice to see that when catching up….you had recently written on my favorite topic–local.
The algortythmic methods Google uses, particularly in Google Maps still seem to have various gliches. One example is the search phrase for “advertising agencies”, a phrase in google maps….that turns up some very odd entities; hotels, museums, universities. Its been doing that for some time. Currently there are far more actual advertising agencies populating 10 pacs for various cities in the US than was the case, a year or 2 ago.
Meanwhile a search for “advertising agency” in Google Maps tends to universally show entities that are marketing or advertising firms. It appears that organic google does a better job of determining the category on the plural variation of the term.
All of which seems to confirm that the methodologies utilized, and the sources chosen are somewhat different in Google Maps and Organic Google.
There are other quirks in Google Maps that point to those differences.
Google’s ability to ascertain from where a search is made is contingent on a wide variety of potential inputs. The simplest of course might have occurred if one personalized one’s account with Google and gave them a city/state address, a zip code, a metropolitan region, a state name, etc. If one doesn’t have personalized results…the output can be very very funky within Maps, or alternatively within the Maps insert for an organic search.
For instance, I have friends working in various cities whose home office is located in California. If those individuals haven’t personalized results with Google…a search for pizza will always turn up a 10pac for Southern California Pizza’s in the Los Angeles area. It doesn’t matter whether these people work out of NYC, Philadelphia, St. Louis, etc. In fact that company has offices worldwide.
So if one doesn’t personalize their results on their office computers, I’d be hesitant to order a pizza from the maps results. I suspect it would take days for delivery and be quite cold…if the request was being made by a member of that firm in Bonn, Germany or NYC.
Dave
Hi Miami,
It is one step, but it does have it’s own unique features. Local is one aspect of search, but it often can be associated with an offline activity that someone is going to perform. That means that some other functions from the search engine might also be triggered as a result of local search, such as a request for driving directions, or the serving of coupons.
Hi Dave,
Good to see you.
Local search does seem to be a work in progress, and I suspect that it will be for a while. Hopefully it will evolve as new approaches are tested, and as the search engines learn from the different methods that they try. From what I’ve seen, plural and singular versions of terms or phrases can often bring back some very different results in organic search as well.
It makes sense that a logged in user is going to receive better results than one whom the search engine knows less about. I don’t know if that will be something that changes much in the near future. That would be some cold pizza.
Bill: This isn’t really on topic….but it speaks to issues in Google Maps. G Maps uses different methods for identifying categories than does Organic G. G Maps is newer and definitely a work in progress. That is okay…Google has always been an evolving work in progress.
Nevertheless the algo’s within G Maps are subject to problems and issues. You’ve identified some of the quirks over time. They continue.
While not relating to the topic of these patents…here is an interesting thread from inside Google groups for business owners identifying some of the quirkyness and problems with existing applications of Google Maps algos:
There seem to be some bugs and quirks working out the exact “where” question:
http://www.google.com/support/forum/p/maps/thread?tid=6c515a66e1584e5e&hl=en
Hi Dave,
Search involving Google Maps means indexing information about businesses rather than websites, which means some unique issues that don’t plague the indexing of much of the content of web pages. Some examples:
1. Businesses don’t need websites to be listed in Google Local.
2. Some businesses have many websites, all of which could be considered the authoritative site for that business.
3. Some sites may offer better business location information about a business than the owner’s site, which may cause confusion about which site is the authoritative site for a business.
4. Information about businesses often come from third party information vendors, and their reasons for collecting information about businesses may place more importance on the collection of some data instead of other data. A Telecom collecting business information may care less about correct street address information than correct phone number information, and provide incorrect or incomplete address information.
5. Some third party information providers haven’t had much incentive historically to update address information when it changes, especially telecoms that publish printed annual printed directories, and when telephone numbers don’t change.
6. Many businesses share an address, whether through sharing an office or location or owner, or using a mail forwarding service or registered agents address.
7. Many businesses have multiple addresses
8. Many businesses, especially home based businesses, would prefer not to have their locations listed.
9. When businesses move locations, they sometimes don’t change all of the address information that appears on their own web sites.
10. When businesses move, they often don’t make changes to online directories where their addresses are listed when they can, and have little control over other directories and sites where their old address may be listed.
11. Different businesses often have the same names in different counties, states, and countries, where DBA names are recorded on a county level, corporate and LLC names are recorded on a state level. How many different “tony’s pizza” parlors are there in the US? Near New York City?
12. Business merge, are acquired, dissolve, close down, relocate, change names and information about those changes aren’t reported or recorded in a way that makes it easy to track.
13. There’s no set standardized format for address information for locations listed on the Web, and the formatting used by different sites may vary considerably from one site to another, making the extraction of that information difficult. The adoption and use of something like hcards might help.
14. Some sites don’t do a very good job of showing their addresses on their websites, such as using image text or flash to display their locations.
15. Errors in address information also happens on web sites.
16. Some organizations, such as nonprofits, schools, government agencies, and parks don’t spend money on directories which list address information, and there might be less information about their locations on the Web than for businesses which do.
I suspect that I’m only scratching the surface of reasons why it’s difficult to get locations and addresses correct for Google Maps.
The merged business listings problem sounds like a bad one. I wonder how many businesses make an effort to check and see how they might be listed in a place like Google Maps, including the ones that don’t even have web sites.
Bill:
That is a very inclusive list. Bravo. It probably speaks quite well to the complexities involved in getting Google Maps or Yahoo Local to reflect accurate information; the difficulties in ironing out issues as pointed out with the above link.
As a business operator/seo…I can tell you that very very few business operators check the accuracy of their records. I can only imagine that the volume of complaints in the Google forum for business owners has to be a tiny percentage of actual misinformation.
Similarly, those that see or record the impact of wrong information in Maps and particularly the map inserts into organic search on Google speak to incredible impact on their business(es). Maps inserts are extremely powerful.
As to that list you created above–:D– It made me smile. Possibly a year ago I was trying to help an institution whose information kept showing wrongly in Maps. We finally identified some aspects of item #9 on your list. Even after identifying this as a possible source for wrong information, the institution’s webmasters took a couple of months to fix the problem. (knock’s head).
Thanks for the list above. Its a tricky problem…and one that is causing some significant problems.
Thanks, Dave
I’ve wondered what percentage of problems reported in the Google Local forum could be resolved in businesses looked more closely at the information that they had control over on their own as well, whether address information on their own sites, or address information in directories and other places that they could login to and update.
I’ve seen more than a few of those things that I listed firsthand, and there’s a lot of incentive for many businesses to resolve those kinds of problems as quickly as possible. As you’ve noted, those map inserts can be really powerful, and help bring customers to businesses.
I agree with you completely Bill. Businesses need to get more involved in making sure their information is up to date and providing things such correct address, contact and even pictures of the location would help bring in customers. I always look for local listings on Google and it makes me mad when I see a place down the street that I can’t find there and Google is not at fault at all.
Hi pays to live green,
Thanks. I’ve seen more than a few listings lately where I can’t find the business listed. It’s impossible to tell if the information about the business is incorrect, or if the business has closed down, or what happened exactly. Ultimately, the problem is Google’s because if people can’t use the service to find correct information, then they will stop using it at all.
Local search is just riddled with problems. Whilst the whole idea is fantastic in its concept, this area of Google is being gamed unbelievably.
For professional web consultants such as myself who make sure clients are listed in Local Search, it becomes very difficult to explain how one business who is their competitor can have multiple listings on the Local Search. It’s not as if this is a new thing, it’s been going on since Local Search became available, but as yet, we don’t seem to be seeing any changes to the Google algorithm to stamp out these black hat practices.
No doubt the day of reckoning will come with these spammers, but until that day, it is difficult to convince clients that they shouldn’t engage in similar practices.
I’d love to give some examples here of the Australian Local Search gaming, but it would only promote the manner in which some of these businesses are generating these listings. Those of you in the know will be well aware of it anyway.
Cheers,
Hi Ric,
Unfortunately, there are some significant problems with local search, and it is being abused.
I guess that we shouldn’t be surprised that there are problems considering the massive scope of the undertaking itself – trying to provide a searchable directory of businesses from information offered by data providers and web sites who all use different standards in how they display information about the businesses mentioned in their databases or on their pages.
There are sites that take advantage of the uncertainty surrounding the confusion involving correct location information for businesses. It isn’t easy explaining why some sites seem to get away with things like having multiple listings. Hopefully Google will find ways to address problems like this in the future.
The single most important action you can take to optimize your local search impressions is providing your own, keyword loaded categories for your listing. I’ve had great success with this to date.
Thanks for providing this space for Local Business Directory. Google is already changing the face of how people look for a business with this service. It is amazing to me that the SEO and Marketing communities as whole, don’t know much about LBD. I see it as the new “Yellow Pages”.
Is there any definitive list of the categories (like Yellow Pages categories) that Google uses? LBD seems to to be sensitive to/dependent on predetermined categories.
Hi Tom,
You’re welcome. I do believe that many SEOs and internet marketers know about local search.
No, there isn’t a definitive list of categories – While there were predetermined categories available from Google, and other categories that they inherited from the data providers that they used (causing a great deal of confusion), Google also lets you type in your own categories these days when you submit and verify your business.
Does anyone know if Google exposes the ‘categories’ field via an API call? For example on maps.google.com I search for ‘Bally’s health club’ this location has the category ‘Health Club’. Is the a screen scrape the only way to access the category name for this location? Does anyone know how many different categories Google uses?
Hi David,
Not sure that I can help you with this one. It might be something that you may want to pose at the Google Maps API Group.
Google seems to now allow business owners to submit their own categories, so there may be an unlimited number of categories.