TY - GEN

T1 - Explainable Event Detection with Event Trigger Identification as Rationale Extraction

AU - Hettiarachchi, Hansi

AU - Ranasinghe, Tharindu

PY - 2023/9/4

Y1 - 2023/9/4

N2 - Most event detection methods act at the sentence-level and focus on identifying sentences related to a particular event. However, identifying certain parts of a sentence that act as event triggers is also important and more challenging, especially when dealing with limited training data. Previous event detection attempts have considered these two tasks separately and have developed different methods. We hypothesise that similar to humans, successful sentence-level event detection models rely on event triggers to predict sentence-level labels. By exploring feature attribution methods that assign relevance scores to the inputs to explain model predictions, we study the behaviour of state-of-the-art sentence-level event detection models and show that explanations (i.e. rationales) extracted from these models can indeed be used to detect event triggers. We, therefore, (i) introduce a novel weakly-supervised method for event trigger detection; and (ii) propose to use event triggers as an explainable measure in sentence-level event detection. To the best of our knowledge, this is the first explainable machine learning approach to event trigger identification.

AB - Most event detection methods act at the sentence-level and focus on identifying sentences related to a particular event. However, identifying certain parts of a sentence that act as event triggers is also important and more challenging, especially when dealing with limited training data. Previous event detection attempts have considered these two tasks separately and have developed different methods. We hypothesise that similar to humans, successful sentence-level event detection models rely on event triggers to predict sentence-level labels. By exploring feature attribution methods that assign relevance scores to the inputs to explain model predictions, we study the behaviour of state-of-the-art sentence-level event detection models and show that explanations (i.e. rationales) extracted from these models can indeed be used to detect event triggers. We, therefore, (i) introduce a novel weakly-supervised method for event trigger detection; and (ii) propose to use event triggers as an explainable measure in sentence-level event detection. To the best of our knowledge, this is the first explainable machine learning approach to event trigger identification.

M3 - Conference contribution/Paper

SP - 507

EP - 518

BT - Proceedings of the 14th International Conference on Recent Advances in Natural Language Processing

A2 - Angelova, Galia

A2 - Kunilovskaya, Maria

A2 - Mitkov, Ruslan

PB - INCOMA Ltd

CY - Shoumen, Bulgaria

T2 - 14th Conference on Recent Advances in Natural Language Processing

Y2 - 4 September 2023 through 6 September 2023

ER -