Data Mining of Causal Relations from Text: Analysing Maritime Accident Investigation Reports

TL;DR

This paper explores text mining techniques to automatically extract causal relations from maritime accident reports, aiming to improve safety analysis by identifying contributory causes through pattern classification and connectives methods.

Contribution

It introduces and compares two novel methods for extracting causal relations from maritime accident reports, demonstrating their effectiveness against manual extraction.

Findings

Pattern classification achieved 65% F-measure

Connectives method achieved 58% F-measure

Text mining can effectively extract causal relations from reports

Abstract

Text mining is a process of extracting information of interest from text. Such a method includes techniques from various areas such as Information Retrieval (IR), Natural Language Processing (NLP), and Information Extraction (IE). In this study, text mining methods are applied to extract causal relations from maritime accident investigation reports collected from the Marine Accident Investigation Branch (MAIB). These causal relations provide information on various mechanisms behind accidents, including human and organizational factors relating to the accident. The objective of this study is to facilitate the analysis of the maritime accident investigation reports, by means of extracting contributory causes with more feasibility. A careful investigation of contributory causes from the reports provide opportunity to improve safety in future. Two methods have been employed in this study to extract the causal relations. They are 1) Pattern classification method and 2) Connectives method. The earlier one uses naive Bayes and Support Vector Machines (SVM) as classifiers. The latter simply searches for the words connecting cause and effect in sentences. The causal patterns extracted using these two methods are compared to the manual (human expert) extraction. The pattern classification method showed a fair and sensible performance with F-measure(average) = 65% when compared to connectives method with F-measure(average) = 58%. This study is an evidence, that text mining methods could be employed in extracting causal relations from marine accident investigation reports.