Powered by Google

Stephen Cummings explains how Ontario's mycommunityinfo.ca uses search engine technology to integrate local information

Stephen Cummings (netConnect) -- netConnect, 1/15/2005

Mycommunityinfo.ca (MCI) provides access to community and government information for nearly one million Ontario residents in over 24 urban and rural municipalities. MCI grew out of a brainstorming session back in 1999 where we asked: How can we provide a cost-effective way to integrate the information offered from three levels of government? The solution we came up with was to not build a database but to utilize search engine technology to access public-sector web sites.

Startup funding was provided by the Ontario Management Board Secretariat and Human Resources Development Canada. Ongoing in-kind support is provided by the City of London's Technology Services Division for the search engine and by Middlesex County Library for project administration. The project manager (who, along with a part-time graphics designer, remain MCI's only staff) was hired in June 2001, by 2002 the Google Search Appliance was identified as the best technical solution, and we had our launch in May 2003.

Governance is by consensus. A senior advisory committee appoints a working group, which gets together four times annually but communicates nearly daily through email. The project is not incorporated. It has no board that separates it from the founding partners. Strategic and operational decisions are based on goodwill and consensus.

MCI supports two channels of service: person-to-person at service desks and by telephone and self-serve via the Internet. A key concept is that MCI supports rather than provides direct service. In all instances, MCI serves as a gateway, redirecting citizens to agencies and published sources.

Service partners in the Middlesex-London area include London Public Library, Middlesex County Library, Ontario Government Information Centre, Concierge Desk of London City Hall, and GAIN (Government Access Information Network) Centre. The GAIN Centre predates MCI and is a consortium of government and community agencies sponsored by the Middlesex County Library. It offers a broad spectrum of person-to-person services.

The challenge for MCI was to discover how to support cooperative service among these very different agencies. What was very clear from the beginning is that citizens don't care at all who provides relevant information—they just want it fast.

MCI hosted a series of in-service sessions for front-line staff featuring topics such as Google search strategy. The hidden agenda was to create interaction among the staff. Participants had huge differences in credentials and came from very different organizations. But as the day progressed, they discovered that they shared the same passion for helping. The outcomes aren't quantifiable, but staff now exchange new information resources such as brochures and URLs and talk to each other regularly. The issue of turf has diminished. The citizen approaching an information service desk or making a telephone call has no need to know that a substantial network of information specialists stands behind the person who says, "How may I help you?"

Until recently, to provide comprehensive community information support you'd have to build a database—or, in the case of libraries, perhaps include the data in your OPAC. Organizations acquire Internet-friendly software and then hire part-time and summer students to fill these databases with local info. Most agree that a database of community information is expensive and difficult to keep current.

The Internet is host to millions of web sites paid for and maintained by government and community organizations. They contain contact data, mission statements, explanations of service parameters, descriptions of events, and often extras such as maps and multimedia. Grass-roots organizations that cannot afford a separate web page are almost always picked up by umbrella organizations, which provide description and contact information. For these reasons we decided to implement an Internet search engine that would look at local public-sector web sites (see below).

In July 2003, we automated compiling of an access database of queries sent to MCI (both to the local municipal/community index on the project's Google Search Appliance and to Google itself for Federal/Provincial results). One year later the database had logged 286,732 distinct query strings.

What do users want? We massaged the database to yield the top 100 queries (identical query strings entered more than once). The most specific interest was employment. Queries ranged from short-term interests (summer jobs) to long-term interests (careers, apprenticeships).

The biggest general topic of interest was municipal information—consistent with the idea of citizen-centered government, which acknowledges that the citizen is primarily interested in information and not with the level of government that provides it. The top 100 interests covered many municipal concerns, such as bylaws, recycling, garbage, local transit, parking, and property taxes. The next largest topic was general community information, e.g., housing, shopping, schools, real estate, churches, and volunteer opportunities. The last category is recreation. Users had a continuing interest in golf, restaurants, recreational programs, and municipal and private entertainment facilities—which supports the often overlooked but potent idea of culture as a local economic driver.

The numbers indicate that citizens expect that information published to the Internet, particularly by municipal governments, should not be a static warehouse but a dynamic resource, constantly updated and refined. Timeliness is important, a finding addressed by the nightly crawls that are a central feature of the MCI Google Search Appliance. They ensure that the index is up-to-date and as current as the web sites indexed. The crawl of roughly 160 web sites, comprising some 110,000 URLs, takes six to eight hours with a 2Mb Internet connection. For legacy services that struggle to maintain databases of community information, the difference in quality can be striking. And the main public Google search engine may take up to eight weeks to pick up a content change on a low-ranked web site—after all, it is dealing with at least 4.2 billion URLs.

A major win for MCI users is that it does not rely on metadata. The appliance indexes the text of web sites crawled, not the metadata tags that have been assigned them. Databases are always built in the language, syntax, and semantics of the database builders and have targeted user bases. As a result, it is difficult to search them using natural language. Search engines use the native vocabulary and language of the web sites they index. This means that no advance knowledge of the structure of the information is required.

MCI expects that the effectiveness of participating sites will increase as the results of query string analysis are distributed. For instance, the top 100 includes 466 instances of "licence" and 250 instances of "license." Search engine technology is literal, while spelling and syntactic formulation of query strings by users is far less exact. By understanding what is being sought, authors can include variants of core terms.

The top ten Referrer Sites (sites from which an MCI query was launched) account for nearly 50 percent of all queries. Of these, 47 percent of all queries were launched or initiated from municipal web sites, the predominant site being the City of London. Because of this, MCI has adjusted its original goal of becoming a destination site separate from the organizations it serves. Instead MCI has developed a Searchbar for municipal and library partners to place on their own web sites. In addition to user convenience, this approach provides Google site search technology to the partners' web sites as a value-added service.

The frame of reference of the user is mainly community and municipal. Typically, a user of the City of London's web site uses the city web site's search function and is not aware that the service is being provided by MCI. The user likely expects to find information provided by a municipal source. The good news is that many citizens are using municipal web sites. The interesting news (for municipal publishers) is contained in the range of query strings sent to MCI by citizen users—useful intelligence about what citizens expect.

Following development in the late 1990s, especially in the UK and Australia, the Canadian federal and the Ontario provincial governments have published Life Events Bundles (LEBs). In Europe and Australia, LEBs were preceded by a decision to "join up" government agencies to facilitate services for citizens. In Canada, LEBs have been service finder tools. These bundles will be immediately recognizable to librarians as a variation on library pathfinders.

While the federal and provincial LEBs represent a giant step in the direction of citizen-friendly service, they fail at the local level. Federal and provincial LEBs cannot provide storefront information (the address of the local passport office for every community in Ontario) without falling prey to the ills of database scope and maintenance. We saw this as an opportunity to develop local LEBs, which are provided as a link from our homepage.

So far, local area librarians have published five LEBs: Spousal Abuse, Moving to Middlesex-London, Next Steps After High School, Retirement, and Lost Pets. Next Steps and Lost Pets are the most popular because of a low cost advertising campaign (Post-it™ Notes in counselors offices and animal shelters). We're considering LEBs on First Jobs, Having a Baby, and Living Safely in Middlesex-London (a one-window access to police, fire, emergency, and victim services).

A core group of two to four individuals with expertise in a topic produce an LEB. Area librarians serve as author. The author and the core group brainstorm the scope and inclusiveness of the bundle, with the author doing most of the groundwork. Drafts are shared with all agencies mentioned in the web page. Once all content issues have been agreed upon, MCI formats the document and publishes it to the web site.

Coming aboard as an MCI client is simple. Typically an upper-tier municipality contracts with MCI for search engine services for itself and all of the lower-tier municipalities in its jurisdiction. This means one contract for about ten municipal entities. The contracting municipality appoints a single contact person, generally the IT director. The contact then provides MCI with a list of the URLs for the municipal and public sector organizations that it wishes to include. MCI does the rest, and generally testing and implementation are complete within a week.

Clients contract annually, and the cost of participation is slightly less than the annual cost of licensing the Google Search Appliance, without the added cost of local network and Search Appliance administration. It's also a relatively risk-free introduction to the community benefits of the Google Search Appliance, useful in case the client wishes to go it alone. MCI also freely offers ongoing consultation and support for development of value-added services.

The mycommunityinfo.ca project is a flexible, scalable, and cost-effective approach to integrating general advisory and referral services of governments and public-sector organizations. Getting it off the ground requires close contact with organizations (for more, see "The Librarian's To Do List," below). Implementing search engine technology as a solution creates a peer environment among content providers, with the net result of increasing traffic to their web sites. Interaction among front-line service staff creates a robust capacity for person-to-person service within the community. Within mycommunityinfo.ca, no organizations are excluded, all are celebrated, and all share control.