Classification and Clustering of Arguments with Contextualized Word Embeddings

About

We experiment with two recent contextualized word embedding methods (ELMo and BERT) in the context of open-domain argument search. For the first time, we show how to leverage the power of contextualized word embeddings to classify and cluster topic-dependent arguments, achieving impressive results on both tasks and across multiple datasets. For argument classification, we improve the state-of-the-art for the UKP Sentential Argument Mining Corpus by 20.8 percentage points and for the IBM Debater - Evidence Sentences dataset by 7.4 percentage points. For the understudied task of argument clustering, we propose a pre-training step which improves by 7.8 percentage points over strong baselines on a novel dataset, and by 12.3 percentage points for the Argument Facet Similarity (AFS) Corpus.

Nils Reimers, Benjamin Schiller, Tilman Beck, Johannes Daxenberger, Christian Stab, Iryna Gurevych · 2019

Related benchmarks

Task	Dataset	Result
Qualitative Evaluation of Stance Distribution and Argument Organization	Allsides Election	Informativeness1.84	8
Argument Organization	Allsides Immigration	Coherence1.45	5
Argument Organization	Allsides Education	Coherence1.7	5
Argument Organization	Allsides National Security	Coherence1.44	5
Argument Organization	Perspectrum Abolish nuclear weapons	Coherence1.84	5
Argument Organization	Perspectrum All nations have a right to nuclear weapons	Coherence1.71	5
Argument Organization	Perspectrum Social networking sites are good for our society	Coherence1.6	5
Argument Organization	Perspectrum There is a need for developing tactical nuclear weapons	Coherence1.68	5
Argument Organization	Allsides Politics	Coherence1.15	5
Argument Organization	Allsides Gun Control	Coherence1.93	5