| his thesis reports a term-weighted dissimilarity algorithm and its application to address matching in Location Based Services (LBS). It is composed of two parts: address matching and web application of location based services. We start with a brief introduction of location based services, the background of address matching and the main objectives and accomplishments of this study.;In Part 2 we are concerned with a mobile application of address extracting algorithms for location based services. We build an intelligent agent for online LBS via wireless Internet accessing. Such an agent, based on efficient and accurate address identification, can analyze the content of certain Web pages (ex. Yellow Pages) to search desired LBS information. We propose an ontology-based conceptual information retrieval approach combined with tree structure matching and the nearest neighbour methods to perform address extraction from texts and documents. A prototype, RouteInfo Mobile LBS for automobile drivers, is developed and tested successfully for LBS searching, mapping and route finding. Future application to target marketing through a combination with customer behavioural and transaction data will also be considered.;In Part 1 we discuss various string similarity measures, e.g., the Levenshtein Distance, the Damerau--Levenshtein Distance, the Longest Common Subsequence, the Searching Minimum Errors and Hamming Distance as well as vector space model. Upon evaluating their strength and weakness, we introduce a term-weighted dissimilarity for effective address matching. This is a combination of edit distance similarity and Term Frequency--Inverse Document Frequency weighting. We implement this algorithm into a software and show its effectiveness via a real application for address matching and correction based on Canada Post's address standard. |
