introduction.tex (867B)
1 \begin{frame}{Introduction} 2 \alert{Sentential approaches}: extract sentences' relation independently (\(\sentenceSet\times\entitySet^2\to\relationSet\)) 3 4 \alert{Aggregate approaches}: maps a set of sentences to a set of facts (\(2^{\sentenceSet\times\entitySet^2}\to2^{\entitySet^2\times\relationSet}\)) 5 6 \medskip 7 8 \begin{block}{Goal} 9 Exploit dataset-level regularities to leverage additional information 10 \end{block}% 11 \begin{block}{Plan} 12 \begin{enumerate} 13 \item Model datasets as graphs 14 \item Related relation extraction work only uses \alert{linguistic} similarities 15 \item Proof that \alert{topological} information can be used 16 \item How topological features are usually extracted (GCN) 17 \item How to extract them differently (WL isomorphism test) 18 \item Experimental results 19 \item Perspective 20 \end{enumerate} 21 \end{block} 22 \end{frame}