Semantic Relation Extraction from Socially-Generated Tags: A Methodology for Metadata Generation

Miao Chen, Xiaozhong Liu, Jian Qin

Abstract


The massive social semantics resource presents both
opportunities and challenges for metadata to leverage its power for information
content representation. One such challenge is the lack of context information of
these tags when they are used in retrieval and automatic processing. This paper
reports a study that uses user-generated tags from Flickr as an example of social
semantics sources to explore a new approach to enriching subject metadata. The
proposed method involves using Flickr tags as the source, Google search results as
the context of co-occurring tags and their relations, and natural language
processing and machine learning as the processing techniques. The preliminary
experiment built a context sentence collection from Google search results, which was
then processed by natural language processing and machine learning algorithms. This
new approach achieved a reasonably good rate of accuracy in assigning relations to
groups of tags. The paper explored further the methodological implications of this
new approach in using social semantics to enrich subject metadata.

Full Text:

PDF