Discourse Self-Attention for Discourse Element Identification in Argumentative Student Essays

Song, Wei; Song, Ziyao; Fu, Rao; Liu, Lizhen; Cheng, Miaomiao; Liu, Ting

doi:10.18653/v1/2020.emnlp-main.225

Cited by 11 publications

(8 citation statements)

References 31 publications

(43 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following Song et al (2020) , we also used the position information as the feature input. The sentence, part-of-speech tagging and position information was then input to BiLSTM for feature extraction as a whole to better understand the contextual information.…”

Section: The Model and Resultsmentioning

confidence: 99%

“…They also designed an effective coarse-to-fine argument fusion mechanism to further improve the precision rate. Song et al (2020) took a step further and identified four argument components (main claim, claim, premise, and other) in two datasets (a Chinese dataset and an English dataset) with structural sentence positional encodings to explicitly represent sentence positions and inter-sentence attentions to capture sentence interactions and enhance sentence representation. To the best of our knowledge, most existing models are rule-based or feature-based ( Persing and Ng, 2016 ; Stab and Gurevych, 2017 ), which require considerable manual efforts and are not flexible or robust in cross-domain scenarios.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

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

2023

View full text Add to dashboard Cite

Argument mining (AM), an emerging field in natural language processing (NLP), aims to automatically extract arguments and the relationships between them in texts. In this study, we propose a new method for argument mining of argumentative essays. The method generates dynamic word vectors with BERT (Bidirectional Encoder Representations from Transformers), encodes argumentative essays, and obtains word-level and essay-level features with BiLSTM (Bi-directional Long Short-Term Memory) and attention training, respectively. By integrating these two levels of features we obtain the full-text features so that the content in the essay is annotated according to Toulmin’s argument model. The proposed method was tested on a corpus of 180 argumentative essays, and the precision of automatic annotation reached 69%. The experimental results show that our model outperforms existing models in argument mining. The model can provide technical support for the automatic scoring system, particularly on the evaluation of the content of argumentative essays.

show abstract

Section: The Model and Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

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

2023

View full text Add to dashboard Cite

show abstract

“…We also propose to use sentence position (spos) embedding as an input feature because it has been proved to be useful in other studies (e.g., Song et al, 2020). The sentence position encoding is…”

Section: Multi-task Learning With Structural Signalmentioning

confidence: 99%

“…Note that ICNALE-AS2R was annotated at the sentence level while PEC was annotated at the segment level. Therefore, we convert the PEC annotation to the sentence level, following the strategy described by Song et al (2020). If a sentence contains only one AC, we use the whole sentence as an AC; if a sentence contains two or more ACs, we split it into multiple sentences while including the preceding tokens into each AC.…”

Section: Multi-corpora Trainingmentioning

confidence: 99%

Multi-task and Multi-corpora Training Strategies to Enhance Argumentative Sentence Linking Performance

Putra¹,

Teufel²,

Tokunaga³

2021

Proceedings of the 8th Workshop on Argument Mining

View full text Add to dashboard Cite

Argumentative structure prediction aims to establish links between textual units and label the relationship between them, forming a structured representation for a given input text. The former task, linking, has been identified by earlier works as particularly challenging, as it requires finding the most appropriate structure out of a very large search space of possible link combinations. In this paper, we improve a state-of-the-art linking model by using multi-task and multi-corpora training strategies. Our auxiliary tasks help the model to learn the role of each sentence in the argumentative structure. Combining multi-corpora training with a selective sampling strategy increases the training data size while ensuring that the model still learns the desired target distribution well. Experiments on essays written by English-as-a-foreign-language learners show that both strategies significantly improve the model's performance; for instance, we observe a 15.8% increase in the F1-macro for individual link predictions.

show abstract

“…The existing neural methods obtain a generic representation of the text through a hierarchical model using convolutional neural networks (CNN) for word-level representation and long short-term memory (LSTM) for sentence-level representation [ 4 ], which is not specific to different features. To enhance the representation of the essay, some studies have attempted to incorporate features such as prompt [ 3 , 13 ], organization [ 14 ], coherence [ 2 ], and discourse structure [ 15 , 16 , 17 ] into the neural model. These features are critical for the AES task because they help the model understand the essay while also making the essay scoring more interpretable.…”

Section: Introductionmentioning

confidence: 99%

Improving Automated Essay Scoring by Prompt Prediction and Matching

Sun

Song

et al. 2022

Entropy

View full text Add to dashboard Cite

Automated essay scoring aims to evaluate the quality of an essay automatically. It is one of the main educational application in the field of natural language processing. Recently, Pre-training techniques have been used to improve performance on downstream tasks, and many studies have attempted to use pre-training and then fine-tuning mechanisms in an essay scoring system. However, obtaining better features such as prompts by the pre-trained encoder is critical but not fully studied. In this paper, we create a prompt feature fusion method that is better suited for fine-tuning. Besides, we use multi-task learning by designing two auxiliary tasks, prompt prediction and prompt matching, to obtain better features. The experimental results show that both auxiliary tasks can improve model performance, and the combination of the two auxiliary tasks with the NEZHA pre-trained encoder produces the best results, with Quadratic Weighted Kappa improving 2.5% and Pearson’s Correlation Coefficient improving 2% on average across all results on the HSK dataset.

show abstract

Discourse Self-Attention for Discourse Element Identification in Argumentative Student Essays

Cited by 11 publications

References 31 publications

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

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

Multi-task and Multi-corpora Training Strategies to Enhance Argumentative Sentence Linking Performance

Improving Automated Essay Scoring by Prompt Prediction and Matching

Contact Info

Product

Resources

About