کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
382365 | 660760 | 2014 | 11 صفحه PDF | دانلود رایگان |
• A relation extractor is developed to extract all relation candidates from sentence.
• Features explored from lexical semantic resources may improve performance.
• Feature selection may enhance performance and reduce computational complexity.
Extracting causal relation underlying natural language is an important issue in knowledge discovery. Most previous studies of casual relation extraction focus on simple cases like causal relations between two noun phrases indicated by fixed verbs or prepositions. For more complicated causal relations, such as causal relations between clauses, the previously developed algorithm may not work. To solve this problem, this paper develops a system that is able to extract causal relations in multi-level language expressions such as, words, phrases and clauses without fixed relators. The information extraction system is composed of a multi-level relation extractor and an ensemble-based relation classifier. It may extract more subtypes of causal relations than previous work because extracting domain is expanded in terms of both syntactic expressions and semantic meanings. In addition, the proposed method outperforms previously developed methods because extended features based on lexical semantic resources are explored. Experiments show that our system achieves an accuracy of 88.69%88.69% and F-score of 0.6637 in a dataset with 300 sentences.
Journal: Expert Systems with Applications - Volume 41, Issue 16, 15 November 2014, Pages 7171–7181