Characterising Emergent Semantics in Twitter Lists

Presented at: 9th Extended Semantic Web Conference (ESWC2012)

by Andrés García-Silva, Jeon-Hyung Kang, Kristina Lerman, Oscar Corcho

Twitter lists constitute a form of organising Twitter users into sets, and can be created and maintained by any user in Twitter. In this paper we describe a characterisation approach of the emergent semantics in these lists, which consists in deriving semantic relations between lists and users by analyzing the co-occurrence of keywords in list names. We use the vector space model and Latent Dirichlet Allocation to obtain similar keywords according to co-occurrence patterns. These results are then compared to similarity measures relying on the WordNet synset hierarchy and to existing Linked Data sets. Results show that co-occurrence of keywords based on members of the lists produce more synonyms and more correlated results to that of WordNet similarity measures.

Keywords: Emergent semantics, Similarity, Twitter lists

Resource URI on the dog food server:

Explore this resource elsewhere: