Improving Categorisation in Social Media using Hyperlinks to Structured Data Sources

Presented at: 8th Extended Semantic Web Conference (ESWC2011)

by Sheila Kinsella, Mengjiao Wang, John Breslin, Conor Hayes

Social media presents unique challenges for topic classification, including the brevity of posts, the informal nature of conversations, and the frequent reliance on external hyperlinks to give context to a conversation. In this paper we investigate the usefulness of these external hyperlinks for determining the topic of individual posts. We focus our analysis on objects which have related metadata available on the Web, either via APIs or as Linked Data. Our experiments show that the inclusion of metadata from hyperlinked objects in addition to the original post content significantly improved classifier performance on two disparate datasets. We found that including selected metadata from APIs and Linked Data gave better results than including text from HTML pages. We also make use of the semantics of the data to compare the usefulness of different types of external metadata for topic classification in a social media dataset.

Keywords: Linked Data, hyperlinks, microblogs, social media, text classification

Resource URI on the dog food server:

Explore this resource elsewhere: