Article Preview
Top1. Introduction
The Internet has an abundance of information and has become one of the most significant resources in our daily lives. These resources contain an enormous number of elements, the retrieval of which depends mostly on webpage search engines. However, most of this knowledge comprises either non-structured or semi-structured data (Mitra et al., 2003), and at present the ability of regular search engines is limited to the retrieval of basic keywords, rather than analysis of the subject matter and content of the webpage itself; these applications then, are still far from perfection. For the reasons mentioned above, much research on efficient message and data extraction has been focused on the effective management of data.
This research has resulted in many developments in information retrieval and data mining strategies. However, these strategies are mostly aimed at semantic data only. Based on a recent estimation, about twenty percent of web-users enquire for spatial context, such as searching for restaurants, theatres or academic institutions; in addition, eighty percent of those web-users type their queries in services with a location orientation (Kornai & Sundheim, 2003; Souza et al., 2005), such as searching for New York delicacies or hostels in Portland. However, they are hindered not only by the current state of development of search engines, but also by the limitations imposed by differences between written English and Chinese. Because it is not possible to put a space between written Chinese characters, this has caused perplexities in relation to word segmentation. For regular search engines therefore, this is quite a barrier to effectively indexing web-content written in Chinese. For example, if the user is browsing for “Portland 民宿” [read as ‘min su’], this means hostels in Portland; a regular search engine might only perform some search on either Portland or on 民宿 separately from the database. This inquiry therefore lacks thematic and spatial context, and this creates a big gap between the demand and supply of the query (Buyukkokten et al., 1999).
Although the Geographic Information System (GIS) has the capability to handle geographic data, access to spatial data is mostly limited to coordinates created by geometric space expressions. Even though Web GIS is presently available, it is quite difficult to combine GIS analysis and text analysis, since the former’s usability does not surpass traditional GIS. Generally, people express their knowledge of geographic locations by using spatial content such as place names, labels, addresses or even telephone numbers, instead of using geometric coordinates (Jones et al., 2001). The content of a webpage is a concrete example of this phenomenon, where people express location data in a spatial context. However, by using this implicit spatial data, the connection between web pages and any particular location can be established.
In regard to the above-mentioned requirement, neither text mining nor GIS are sufficient in themselves; instead, a new search engine needs to be built with the capability to manage thematic and spatial contexts, not separately but simultaneously. Geographical Information Retrieval (GIR) has become one means of satisfying such queries, and it is starting to receive some attention from academic and commercial communities (Byrd & Ravin, 1999; Jones et al., 2002). GIR combines text mining, information retrieval (IR) and geographic metadata with spatial cognition related to the research area. The main purpose of this is to retrieve spatial-related information from text documents more accurately.
Roughly seventy-five to eighty percent of human activity, especially traveling, is related to geographic location (Lee et al., 2007). Following the increasing trend to holiday outings in Taiwan, there has been a noticeable expansion in agricultural leisure farms and rapid growth in the number of hostels and homestay facilities. At a conservative estimate there are more than ten thousand homestays in Taiwan, both listed and unlisted and about two hundred new hostels are being opened every year. Consequently, the number of related web pages is also increasing tremendously. Nevertheless, some of these web pages are not mainly related to this subject, or they offer only text descriptions, rather than concrete spatial information display. Such information may confuse users during their searches, and this is where GIR can provide searches with better functionality.