Event Schemas Based on FrameNet

This repository contains event schema resources derived from FrameNet-style frame definitions. It provides event types, event scenarios, semantic roles, hierarchy views, and relation mappings between events.

The resource is useful for work on event extraction, schema induction, event argument modeling, ontology construction, semantic role analysis, and FrameNet-based event representation.

Contents

Path	Description
eventSchemas/	XML definitions for individual event schemas.
event_schema_list.txt	Flat list of event schemas and their argument roles.
event_scenario_list.txt	Flat list of higher-level event scenarios and their argument roles.
eventRelations.xml	XML file describing relations between event schemas, including role mappings.
EventFrameHierachy.html	Static HTML view of the event frame hierarchy.
EventScenarioHierarchical.html	Static HTML view of the event scenario hierarchy.

Resource Summary

• 655 event schema XML files in eventSchemas/
• 50 event scenario entries in event_scenario_list.txt
• 10 event relation types in eventRelations.xml
• 2,070 event-to-event relations
• 12,393 role mappings between related events

Event Schema Format

Each file in eventSchemas/ defines one event schema. For example:

eventSchemas/Attack.xml
eventSchemas/Commerce_buy.xml
eventSchemas/Death.xml
eventSchemas/Employment_start.xml

The XML files contain:

• the event/frame name and ID
• a textual definition
• argument roles
• role definitions
• role core type, such as Core, Peripheral, or Extra-Thematic
• semantic type information when available
• lexical units and annotations when available

For example, the Attack schema includes roles such as:

• Assailant
• Victim
• Weapon
• Place
• Time
• Purpose
• Result

Lists

event_schema_list.txt provides a compact tabular view:

Event Type          Event Argument Roles
Attack              (Assailant, Victim, Weapon, Place, Time, ...)
Commerce_buy        (Buyer, Seller, Goods, Money, ...)
Death               (Protagonist, Cause, Time, Place, ...)

event_scenario_list.txt contains higher-level event scenarios, for example:

Commerce_scenario
Crime_scenario
Employment_scenario
Medical_interaction_scenario
Motion_scenario
Transfer_scenario

These files are convenient starting points when browsing the resource or selecting event types for downstream tasks.

Event Relations

eventRelations.xml defines structured relations between events. The relation types include:

• Causative_of
• Inchoative_of
• Inheritance
• Metaphor
• Perspective_on
• Precedes
• ReFraming_Mapping
• See_also
• SubEvent
• Using

Each relation can include mappings between argument roles in the related events. For example, a relation may connect a child event to a parent event and specify how the child event's roles correspond to the parent event's roles.

How to Use

Clone the repository:

git clone https://github.com/EmanuelaBoros/EventSchemasBasedOnFrameNet.git
cd EventSchemasBasedOnFrameNet

Browse the resource directly:

ls eventSchemas/
less event_schema_list.txt
less event_scenario_list.txt
less eventRelations.xml

Open the hierarchy files in a browser:

EventFrameHierachy.html
EventScenarioHierarchical.html

No installation is required unless you want to parse the XML files programmatically.

Programmatic Parsing

The XML files can be read with standard XML parsers. Because the schema files use XML namespaces, namespace-aware parsing is recommended.

Minimal Python example:

from pathlib import Path
import xml.etree.ElementTree as ET

ns = {"fn": "http://framenet.icsi.berkeley.edu"}

schema_path = Path("eventSchemas/Attack.xml")
tree = ET.parse(schema_path)
root = tree.getroot()

event_name = root.attrib["name"]
roles = [
    role.attrib["name"]
    for role in root.findall("fn:Arg", ns)
]

print(event_name)
print(roles)

To parse relation mappings:

import xml.etree.ElementTree as ET

ns = {"ev": "http://Eventnet.icsi.berkeley.edu"}

tree = ET.parse("eventRelations.xml")
root = tree.getroot()

for relation_type in root.findall("ev:eventRelationType", ns):
    relation_name = relation_type.attrib["name"]
    for relation in relation_type.findall("ev:eventRelation", ns):
        sub_event = relation.attrib["subEventName"]
        super_event = relation.attrib["superEventName"]
        mappings = [
            (m.attrib["subRoleName"], m.attrib["superRoleName"])
            for m in relation.findall("ev:RoleMapping", ns)
        ]
        print(relation_name, sub_event, "->", super_event, mappings[:3])