Argumentation Mining on Essays at Multi Scales

Abstract

Argumentation mining on essays is a new challenging task in natural language processing, which aims to identify the types and locations of argumentation components. Recent research mainly models the task as a sequence tagging problem and deal with all the argumentation components at word level. However, this task is not scale-independent. Some types of argumentation components which serve as core opinions on essays or paragraphs, are at essay level or paragraph level. Sequence tagging method conducts reasoning by local context words, and fails to effectively mine these components. To this end, we propose a multi-scale argumentation mining model, where we respectively mine different types of argumentation components at corresponding levels. Besides, an effective coarse-to-fine argumentation fusion mechanism is proposed to further improve the performance. We conduct a serial of experiments on the Persuasive Essay dataset (PE 2.0). Experimental results indicate that our model outperforms existing models on mining all types of argumentation components.

Fusion weighted features and BiLSTM-attention model for argument mining of EFL writing

“…They can appear anywhere in a paragraph, either proposed at the beginning, summarized in the end, or given in the middle. They are at the paragraph level ( Wang et al, 2020 ). Similarly, the different elements in Toulmin models correspond to token-lever.…”

“…Taking the overall performance into account, Wang et al (2020) proposed a multi-scale mining model to mine the argument elements (major claim, claim, and premise) at the discourse level, paragraph level, and word level. They also designed an effective coarse-to-fine argument fusion mechanism to further improve the precision rate.…”

Cross-genre argument mining: Can language models automatically fill in missing discourse markers?

“…The standard BiLSTM with a CRF output layer emerged as the state-of-the-art architecture for token-level sequence tagging, including argument mining [9,17,59]. Current state-ofthe-art on ADU identification and classification employs BERT [15] or Longformer [3] as base encoders (in some cases, with a CRF layer on top), typically accompanied with specific architectures to tackle a target corpus or task-specific challenges [16,24,41,69,71]. We follow these recent trends by employing a BERT-based sequence labeling model.…”

End-to-end Argument Mining with Cross-corpora Multi-task Learning

“…Mayer et al (2020) presented a BERT model with a component pair classifier for predicting relations. Wang et al (2020) also used BERT in encoders. In contrast to the above studies, since a given input argument can have many sentences, we use Longformer (Beltagy et al, 2020) to encode text, as it can handle a longer input sequence than BERT.…”