CRET: A Tool for Automatic Extraction of Causal Relations

Abstract

There is an interest in extracting knowledge and retrieving information

automatically from the current availability of a large collection of electronic

resources and from the academic literature available on the Web. In this work

a tool called Causal Relation Extraction Tool (CRET) has been developed

to extract causal relations from texts. The tool is a Relation Parser to extract

relation patterns from medical documents. The causal patterns are detected

through a fuzzy matching process between the causal patterns database and

partial detected string patterns in the electronic medical documents. The

extracted knowledge is stored as an index for the documents and the researchers

can consult the indexed databases. The main contribution of this work is a

method for cause, effect and condition extraction using a fuzzy relation. The

causal extraction method is based on extracted noun and adjectival phrases

associated with causal verb (patterns). Quantitative matrices measurements

like, Precision, recall, and F-score for the classifiers and the causal pattern

extraction were used and computed to evaluate our result. The results indicate

that CRET generates 77% of the keywords and the casual relations which

have manually been associated by human expert.