View article

Searching for locations in web data and associating a document with a corresponding place on the map becomes popular in user’s daily activities and it is the first step in web page processing. People often manually search for locations on a web page and then use map services to highlight them because geographic information is not always explicitly available.

In this work, we present a geo-tagging framework to extract all addresses from web pages. The solution includes an efficient web page processing approach, which combines a probabilistic language model with real-world knowledge of addresses on maps and extends geocoding services from short queries to large text documents and web pages. We discuss the main problems in dealing with web pages such as: web page noise, identification of relevant segments, and extraction of incomplete addresses. The experimental result shows …