Ground-truth rankings of semantic associations

In this zip file, the data directory contains 720 files, or 240 groups of files. Each group consists of 3 files. 'XXX_entities.txt' contains a query consisting of a set of query entities. 'XXX_xh.nt' and 'XXX_xl.nt' contain two semantic associations. Each semantic association is a subgraph of DBpedia (version 2015-10) connecting all the query entities, serialized in a N-Triples document (.nt).

The result.tsv file contains the results of comparisons made by human experts. This TSV file has 1,200 lines. Each line consists of the id of a pair of semantic associations, the id of a user, and the result of comparison of the pair made by the user. The numerical result is interpreted as follows: 1.0/0.5 indicating that xh is significantly/slightly more important than xl; -1.0/-0.5 indicating that xl is significantly/slightly more important than xh; 0.0 indicating that xh and xl are equally important.

How to cite: Gong Cheng, Fei Shao, Yuzhong Qu. An Empirical Evaluation of Techniques for Ranking Semantic Associations. IEEE Transactions on Knowledge and Data Engineering, 29(11):2388--2401, 2017.

Contact Gong Cheng with any questions or comments.

©2017 Websoft Research Group, Nanjing University, China