Start Date
12-18-2013
Description
The ever-increasing number of publications in the behavioral sciences has yielded a large set of behavioral theories, constructs and their relationships. In this study, we propose a novel method to discover this data set through construct-level citations. In behavioral science, the reuse, adaptation and modification of an existing behavioral construct often requires an explicit citation. However, finding the construct-level citations are not trivial, since citations are at the paper level of analysis. Based on state-of-the-art information extraction techniques, we propose an automatic construct-level citation extraction method, which consists of five steps: article crawling, article annotation, citation extraction, construct extraction and referring relation extraction. Serving as a proof-of-concept, our method is applied to a data set consisting of publication of two Information Systems journals, and is evaluated against human decisions. The initial results represent a promising opportunity of the application of our proposed method to a large set of behavioral publications.
Recommended Citation
Li, Jingjing and Larsen, Kai, "Tracking Behavioral Construct Use through Citations: A Relation Extraction Approach" (2013). ICIS 2013 Proceedings. 18.
https://aisel.aisnet.org/icis2013/proceedings/KnowledgeManagement/18
Tracking Behavioral Construct Use through Citations: A Relation Extraction Approach
The ever-increasing number of publications in the behavioral sciences has yielded a large set of behavioral theories, constructs and their relationships. In this study, we propose a novel method to discover this data set through construct-level citations. In behavioral science, the reuse, adaptation and modification of an existing behavioral construct often requires an explicit citation. However, finding the construct-level citations are not trivial, since citations are at the paper level of analysis. Based on state-of-the-art information extraction techniques, we propose an automatic construct-level citation extraction method, which consists of five steps: article crawling, article annotation, citation extraction, construct extraction and referring relation extraction. Serving as a proof-of-concept, our method is applied to a data set consisting of publication of two Information Systems journals, and is evaluated against human decisions. The initial results represent a promising opportunity of the application of our proposed method to a large set of behavioral publications.