Another look at causality: Discovering scenario-specific contingency relationships with no supervision

Mehwish Riaz, Roxana Girju

Research output: Contribution to conferencePaper

Abstract

Contingency discourse relations play an important role in natural language understanding. In this paper we propose an unsupervised learning model to automatically identify contingency relationships between scenario-specific events in web news articles (on the Iraq war and on hurricane Katrina). The model generates ranked contingency relationships by identifying appropriate candidate event pairs for each scenario of a particular domain. Scenario-specific events, contributing towards the same objectives in a domain, are likely to be dependent on each other, and thus form good candidates for contingency relationships. In order to evaluate the ranked contingency relationships, we rely on the manipulation theory of causation and a comparison of precision-recall performance curves. We also perform various tests which bring insights into how people perceive causality. For example, our findings show that the larger the distance between two events, the more likely it becomes for the annotators to identify them as non-causal.

Original languageEnglish (US)
DOIs
StatePublished - Sep 2010
Event4th IEEE International Conference on Semantic Computing, ICSC 2010 - Pittsburgh, PA, United States
Duration: Sep 22 2010Sep 24 2010

Other

Other4th IEEE International Conference on Semantic Computing, ICSC 2010
CountryUnited States
CityPittsburgh, PA
Period9/22/109/24/10

Fingerprint

Causality
Scenarios
Unsupervised learning
Hurricanes
Likely
Causation
Unsupervised Learning
Natural Language
Manipulation
Curve
Relationships
Dependent
Evaluate
Model

Keywords

  • Causality
  • Contingency
  • Scenario
  • Topics

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Theoretical Computer Science

Cite this

Riaz, M., & Girju, R. (2010). Another look at causality: Discovering scenario-specific contingency relationships with no supervision. Paper presented at 4th IEEE International Conference on Semantic Computing, ICSC 2010, Pittsburgh, PA, United States. https://doi.org/10.1109/ICSC.2010.19

Another look at causality : Discovering scenario-specific contingency relationships with no supervision. / Riaz, Mehwish; Girju, Roxana.

2010. Paper presented at 4th IEEE International Conference on Semantic Computing, ICSC 2010, Pittsburgh, PA, United States.

Research output: Contribution to conferencePaper

Riaz, M & Girju, R 2010, 'Another look at causality: Discovering scenario-specific contingency relationships with no supervision', Paper presented at 4th IEEE International Conference on Semantic Computing, ICSC 2010, Pittsburgh, PA, United States, 9/22/10 - 9/24/10. https://doi.org/10.1109/ICSC.2010.19
Riaz M, Girju R. Another look at causality: Discovering scenario-specific contingency relationships with no supervision. 2010. Paper presented at 4th IEEE International Conference on Semantic Computing, ICSC 2010, Pittsburgh, PA, United States. https://doi.org/10.1109/ICSC.2010.19
Riaz, Mehwish ; Girju, Roxana. / Another look at causality : Discovering scenario-specific contingency relationships with no supervision. Paper presented at 4th IEEE International Conference on Semantic Computing, ICSC 2010, Pittsburgh, PA, United States.
@conference{1540f4c5e2864412b8b3adbc95ca0443,
title = "Another look at causality: Discovering scenario-specific contingency relationships with no supervision",
abstract = "Contingency discourse relations play an important role in natural language understanding. In this paper we propose an unsupervised learning model to automatically identify contingency relationships between scenario-specific events in web news articles (on the Iraq war and on hurricane Katrina). The model generates ranked contingency relationships by identifying appropriate candidate event pairs for each scenario of a particular domain. Scenario-specific events, contributing towards the same objectives in a domain, are likely to be dependent on each other, and thus form good candidates for contingency relationships. In order to evaluate the ranked contingency relationships, we rely on the manipulation theory of causation and a comparison of precision-recall performance curves. We also perform various tests which bring insights into how people perceive causality. For example, our findings show that the larger the distance between two events, the more likely it becomes for the annotators to identify them as non-causal.",
keywords = "Causality, Contingency, Scenario, Topics",
author = "Mehwish Riaz and Roxana Girju",
year = "2010",
month = "9",
doi = "10.1109/ICSC.2010.19",
language = "English (US)",
note = "4th IEEE International Conference on Semantic Computing, ICSC 2010 ; Conference date: 22-09-2010 Through 24-09-2010",

}

TY - CONF

T1 - Another look at causality

T2 - Discovering scenario-specific contingency relationships with no supervision

AU - Riaz, Mehwish

AU - Girju, Roxana

PY - 2010/9

Y1 - 2010/9

N2 - Contingency discourse relations play an important role in natural language understanding. In this paper we propose an unsupervised learning model to automatically identify contingency relationships between scenario-specific events in web news articles (on the Iraq war and on hurricane Katrina). The model generates ranked contingency relationships by identifying appropriate candidate event pairs for each scenario of a particular domain. Scenario-specific events, contributing towards the same objectives in a domain, are likely to be dependent on each other, and thus form good candidates for contingency relationships. In order to evaluate the ranked contingency relationships, we rely on the manipulation theory of causation and a comparison of precision-recall performance curves. We also perform various tests which bring insights into how people perceive causality. For example, our findings show that the larger the distance between two events, the more likely it becomes for the annotators to identify them as non-causal.

AB - Contingency discourse relations play an important role in natural language understanding. In this paper we propose an unsupervised learning model to automatically identify contingency relationships between scenario-specific events in web news articles (on the Iraq war and on hurricane Katrina). The model generates ranked contingency relationships by identifying appropriate candidate event pairs for each scenario of a particular domain. Scenario-specific events, contributing towards the same objectives in a domain, are likely to be dependent on each other, and thus form good candidates for contingency relationships. In order to evaluate the ranked contingency relationships, we rely on the manipulation theory of causation and a comparison of precision-recall performance curves. We also perform various tests which bring insights into how people perceive causality. For example, our findings show that the larger the distance between two events, the more likely it becomes for the annotators to identify them as non-causal.

KW - Causality

KW - Contingency

KW - Scenario

KW - Topics

UR - http://www.scopus.com/inward/record.url?scp=79952047541&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79952047541&partnerID=8YFLogxK

U2 - 10.1109/ICSC.2010.19

DO - 10.1109/ICSC.2010.19

M3 - Paper

AN - SCOPUS:79952047541

ER -