For each Frame-Role (i.e. Frame Element) pair we have extracted a list of senses, where a sense is a WordNet synset, that possible fillers of that role might take. The process is the following. We start from the FrameNet 1.5 annotated corpus. First, we link role fillers’ heads to Wikipedia and then, with the help of BabelNet and Yago, assign one or several WordNet synsets to each filler. After that, we generalize over the examples for each frame-role in order to select the most representative set of senses for a given role.

Further details on the resource with the link to download it and a tool to compute statistics are available here.


Irina Sergienya, Volha Bryl, Sara Tonelli. A Semantic Role Repository Linking FrameNet and WordNet: Resource description for the Monnet Challenge (conversion of existing linguistic resources into linked data). Co-located with Multilingual Linked Open Data for Enterprises (MLODE 2012) Workshop, Leipzig, Germany, September 23-25, 2012. Awarded a runner up prize.

Sara Tonelli, Volha Bryl, Claudio Giuliano, Luciano Sera?ni. Investigating the Semantics of Frame Elements. In Proceedings of the 18th International Conference on Knowledge Engineering and Knowledge Management (EKAW 2012), pp. 130-143. See the presentation on YouTube.


satonelli [at] fbk [dot] eu