A Mobile Portal Service to Provide Location Dependent Information

Katsumi Takahashi
Nippon Telegraph and Telephone Corporation (NTT)
2.What is "Mobile Info Search"
2.1 Location-Oriented Meta Search
2.2 Location-Oriented Robot Based Search
3.Experimental Results


How to collect and index resources associated with places or areas is an important issue for the location dependent information service. We are developing software to utilize the Internet resources relevant to some locations and we have been testing the portal site "Mobile Info Search" (http://www.kokono.net/ and its English ver.) to provide Japanese local information for mobile users. Core technique of Mobile Info Search is "Automatic Location Metadata Generation" that extracts the address strings from the original resource, converts the address strings to the coordinates by looking up the location information repositories, generates a metatag which represents the relevant location to the resoure, and index them for searching or mapping. The technique, the system and the experimental results are reported.


Information on the Internet is becoming useful for our daily life today. It includes restaurant guides, local maps, public transportation, and weather reports. Moreover, due to the progress of the mobile computing, we can access to the Internet even when we are not in offices or home. A methodology and a service of utilizing such local information for mobile computing users is required.

On the Web, we can find many resources associated with a particular location. But, automated discovery and geographical indexing for these resources are not easy. Some of these resources include address or some other strings that describe the location they associated with but they do not explicitly. There also are resources that can be accessed through a certain query with a certain notation of location. We are developing the techniques that mediates these resources for each location that the user is or wants to find information.

2.What is "Mobile Info Search"

Mobile Info Search is to provide local information from the Internet by collecting, structuring, organizing, and filtering in a practicable form. To utilize such local information, Mobile Info Search employs a mediator architecture; a software agent between users and the information sources. The architecture is shown in Fig. 1. Between users and information sources, Mobile Info Search mediates Database-type resources using Location-Oriented Meta Search and static files using Location-Oriented Robot-based Search ("kokono Search") using Automatic Location Metadata Generation. To use the service, only Internet connectable PDSs or PCs with Web browsers are required for users. Additionally if the user have a location-aware PHS or GPS unit, the user location is automatically obtained.

Mobile Info Search is open to the public on the Internet since 1997 as our test service. From the address http://www.kokono.net/ , anyone can enjoy both searches. When user access to the Mobile Info Search, the index page of the location is automatically displayed (Fig.2). User's current location is displayed as the address, the longitude-latitude, and the nearest station on the top of the page. About 20 WWW services, provided by the courtesy of 8 companies are available. They include following; kokono Search, Yellow Pages and Restaurants guide under the Shops information, several maps, train tables, hotel guides and reservations, weather reports, and TV listings. This index page can be generated automatically for any Japanese location requested.

Fig.1. An architecture of Mobile Info Search
Fig.2. Index of Mobile Info Search

2.1 Location-Oriented Meta Search

Location-Oriented Meta Search provides a mediation service for database-type resources. It provides a simple interface for local information services which have various search interfaces. Only to select the service such as the maps or the yellow pages, users can get information of the location easily from each server.

Both location information from the user and for the information sources are in various form. It can be address strings, longitude-latitude, postal-codes, or landmarks. The mediator converts the user location into the location information suitable for the target using the location information repository. The repository we constructed on Mobile Info Search is the set of location information that can convert each format to others. This mediation software is prepared manually for each services. The access method, the request form, and the usage of local information of the target is defined in the software.

Location-Oriented Meta Search performs in following way;
  1. to receive user location (longitude latitude) and the name of the target service
    ex. (x=139.36.27, y=35.25.24, target = AYellowPages)
  2. to converts the user location into the location suitable form for the target using Location Information Repository
    ex. (x=139.36.27, y=35.25.24) -> address = "Tokyo, Chuo-ku, Ginza 4"
  3. to determine the search scope of the location for the target
    ex. ``Tokyo'', or ``Tokyo, Chuo-ku'', or ``Tokyo, Chuo-ku, Ginza'', ... ?
  4. to create a query
    ex. http://www.AYellowPages....co.jp/seach.cgi? address = "Tokyo, Chuo-ku, Ginza 4"
  5. to search (yellow page sample)

2.2 Location-Oriented Robot Based Search (kokono Search)

kokono Search provides the spatial search that searches the document close to a location. Just like other search engines, it employs a software called "robot" or "spider" that collects documents from the Internet and creates local database for collected documents. While other search engines provide a keyword-based search, kokono Search do a location-based spatial search. It displays documents in the order of the distance between the location of the document and the user's location.
Fig.3. Automatic Location Metadata Generation

3.Experimental Results

We analyzed about half million (479,669) documents collected and organized by Automatic Location Metadata Generation and found about 40 % (210,369) of documents contained address strings (Table 1.). As some documents contains more than one addresses, we collected 1,122,380 total addresses. The collection rate for addresses is shown in Table. 2. We collected almost every cities, 30.3% of towns, and 6.8% of Chome level addresses of Japan. Collecting some deeper addresses from the Web pages is not a easy matter. Because some Chome or deeper level addresses are used only for describing the location for personal houses, they may not appear on the Internet.

Table 1. Numbers of collected documents
  Prefecture City Town Chome
(Block #)
Documents that have addresses (A) 210,369 151,272 68,381 32,755 210,369
Addresses collected total (B) 418,488 501,992 117,066 88,926 1,122,380
Addresses per documents (B/A) 2.0 3.2 1.6 2.6 5.3

Table 2. Collection rate of documents for Japanese addresses
  Prefecture City Town Chome
All addresses in Japan (A) 47 3,883 121,172 343,269
Addresses collected unique (B) 47 3,876 36,799 23,326
Collection rate (B/A) 100 % 99.8 % 30.3 % 6.8 %