A pipeline-based paradigm treats all sub-tasks as independent classification problems and makes event extraction a multi-stage classification problem. A classifier is used at each subtask level,  which takes advantage of dependencies between subtasks

Pipeline-based event extraction

A joint-based paradigm identifies the triggers and arguments according to candidate triggers and entities. From there, trigger classification and argument role classification are realized simultaneously, which helps minimize error propagation 

Joint-based event extraction

Event extraction can be done  through two different paradigms, which describe the timeline of how each of the subtasks are completed

1. Pipeline-based
2. Joint-based

Carleton College

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

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

An object or group of objects in a semantic category. Entity mainly includes people, organizations, places, times, things

Entity (event extraction)

The phrase or sentences that describe the event that contains a trigger and the corresponding arguments

Ex) "Baghdad, a cameraman died when an American tank fired on the Palestine Hotel" contains an event

Event mention (event extraction)

Describes the nature of  the event and refers to the category to which the event corresponds, usually represented by the type of the event trigger.

From the previous example, "Baghdad, a cameraman died when an American tank fired on the Palestine Hotel", the event type would be *Die*

Event type (event extraction)

Refers to the core unit in event extraction (a verb or a noun in the event mention)

From the previous example, "Baghdad, a cameraman died when an American tank fired on the Palestine Hotel", the event trigger would be "died"

Event trigger (event extraction)

Event argument is the main attribute of events. It includes entities, nonentity participants, and time, and so on. 

From the previous example, "Baghdad, a cameraman died when an American tank fired on the Palestine Hotel", the event arguments are “Baghdad”, “cameraman”, and  “American tank”

Event argument (event extraction)

Describes the relationship between the event trigger and the event argument

From the previous example, "Baghdad, a cameraman died when an American tank fired on the Palestine Hotel", the corresponding argument role to the event argument "Baghdad" would be *place*

Argument role (event extraction)

Considered the core unit in event extraction; task is to find the word that clearly express an event’s occurrence in the text

Trigger identification (event extraction)

Task of determining one or several events types that the event trigger belongs to

Trigger Classification (event extraction)

Task of identifying all the arguments contained in an event type from the text

Argument Identification (event extraction)

Act of classifying the argument role of each identified argument in an event

Argument role identification (event extraction)

There are several ways of implementing event extraction:

- Classification-based
- Machine Reading Comprehension (MRC)-based
- Sequence labelling-based
 - Sequence-to-structure generation-based

Learn Before

Related

Learn After