�ݺ�ߣ

This document summarizes research on using tags and metadata from Flickr photos to build models of locations and derive the semantics of tags. It describes experiments with a dataset of 42,000 geotagged photos of San Francisco to cluster photos by location and score tags based on metrics like TF-IDF to generate "tag maps" that depict important places and landmarks for a given city. The research aims to improve image search, develop automated gazetteers of places and events, and associate time/place data to tags for richer understanding of content in large photo collections.Read less