This paper defines a standard for comparing relation extraction (RE) sys- tems based on a Distant Supervision (DS). We integrate the well-known New York Time corpus with the more recent version of Freebase. Then, we define a simpler RE system based on DS, which exploits SVMs, tree kernels and a simple one-vs-all strategy. The resulting model can be used as a baseline for system comparison. We also study several example filtering tech- niques for improving the quality of the DS output.
Creating a standard for evaluating Distant Supervision for Relation Extraction
Moschitti, Alessandro
2014-01-01
Abstract
This paper defines a standard for comparing relation extraction (RE) sys- tems based on a Distant Supervision (DS). We integrate the well-known New York Time corpus with the more recent version of Freebase. Then, we define a simpler RE system based on DS, which exploits SVMs, tree kernels and a simple one-vs-all strategy. The resulting model can be used as a baseline for system comparison. We also study several example filtering tech- niques for improving the quality of the DS output.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione