糖心TV

IAS seminar: Karen Stepanyan
Title: Semantic Annotation of Social Media Content by Exploiting
Embedded Geolocation Data
Abstract: Semantic annotation of unstructured documents is one of the
primary directions for interconnecting the Web of Documents with the Web
of Data. Despite the considerable progress in the areas of natural
language processing and data extraction, the problem of annotating
unstructured documents remains open. This paper proposes an approach
that improves the process of annotating user-generated social media
content by exploiting the embedded geolocation data. We conducted a
survey of more than 1.4 million weblogs for demonstrating the rationale
for using geolocation data for improving semantic enrichment. The main
contribution of the paper is a geolocation-aware semantic annotation
model that extends the existing solutions for spotting and
disambiguation. The main contribution of the paper is a
geolocation-aware semantic annotation model that extends the existing
solutions for spotting and disambiguation. The evaluation of the model
is based on a corpus of 3,165 weblog posts with embedded geolocation,
obtained from 1,775 distinct web feeds. The results demonstrate a high
level of accuracy in annotation (87.7%), suggesting the adoption of the
model for improving the existing solutions.