Traditional event extraction models are typically pattern-matching methods, mainly based on syntax trees or regular expressions. A pitfall of this approach is poor portability and flexibility from domain to domain.

Traditional event extraction models

A CNN-based event extraction model captures the essential information in a sentence, and it uses a dynamic multi-pool layer to store more critical information based on event triggers and arguments

CNN-based event extraction model

In the encoding stage, it uses a RNN to summarize the context information. Furthermore, it predicts both trigger and argument roles in the prediction stage.

RNN-based event extraction model

The attention mechanism determines what global information is to be used to model local context. Additionally, the attention mechanism makes the model pay attention to the important feature information of the sentence while ignoring other unimportant feature information

Attention-based event extraction model

GCN-based models work to present an efficient method for directly linking words to their informative context for event detection in sentences

Graph Convolutional Network (GCN) based event extraction model

A transformer-based event extraction model promotes event extraction by using a combination of an extraction model and a generation method based on pre-trained language models

Transformer-based event extraction model

Event extraction models are taken from either a traditional approach or a deep learning approach. Deep learning event extraction models can be broken down into CNN-based, RNN-based, Attention-based, Graph Convolutional Network (GCN) based, and Transformer based models.

Carleton College

Event extraction is the task of identifying mentions of events in texts. In this case, an event mention is any expression denoting an event or state that can be assigned to a particular point, or interval, in time. 

Event extraction is generally modeled via supervised learning, detecting events via sequence models with IOB tagging, and assigning event classes and attributes with multi-class classifiers.

Event Extraction

Li, Q., Li, J., Sheng, J., Cui, S., Wu, J., Hei, Y., Peng, H., Guo, S., Wang, L., Beheshti, A., & Yu, P. (2021). A Compact Survey on Event Extraction: Approaches and Applications. https://arxiv.org/abs/2107.02126.

A Compact Survey on Event Extraction: Approaches and Applications

With both the events and the temporal expressions in a text having been detected, the next logical task is to use this information to fit the events into a complete timeline. Such a timeline would be useful for applications such as question answering and summarization.

The temporal relation between events is classified into one of the standard set of Allen relations depicted in the figure below. This task is done through training on the TimeBank corpus with features like words/embeddings, parse paths, tense and aspect.

Temporal Ordering of Events

Given input text, an event extraction model will first identify an event and then classify that event into a particular schema. From there, the model will identify arguments related to the event and then classify the roles of those arguments

Event extraction can be broken down into four subtasks:

1. Trigger identification
2. Trigger classification
3. Argument identification
4. Argument Role Classification

Overview of Event Extraction

Event extraction models

Open domain event extraction is the task of detecting events from text and clustering similar events through extracted event keywords in the absence of a predefined event pattern

Open domain event extraction

Closed domain event extraction task is to find words that belong to a specific pre-defined event schema

Closed domain event extraction

Document-level event extraction aims to extract events across an entire article rather than a single sentence example. Challenges include argument scattering, which is when an event is scattered across several sentences and multi-event documents.

Document-level event extraction

Precision (P), Recall (R), and F1 are used to measure the performance are used evaluate the four subtasks of event extraction

Event extraction evaluation metrics

- Event Factuality Identification
- Event Relation Extraction, which includes Event coreference resolution, Event casual relation extraction, and Event temporal relation extraction
- Script Event Prediction

Learn Before

Related

Learn After