Page 148 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42

HUA Bolin / Types and description rules of knowledge elements about methods in academic papers 147

(2) Knowledge element representation and modeling
Wang Yu et al. (2013) proposed a journal paper knowledge warehouse construction method
using 6-tuple (serial No., navigation, source, type, feature word, content) knowledge element.
The method is based on a journal paper knowledge element base. Their contribution is that they
designed a method to extract knowledge from the knowledge element base to the knowledge
warehouse. Based on the knowledge node from Brooks and the latent linkage method from
Swanson, Jiang Yongchang et al. (2007) proposed a knowledge network construction method
and implementation model. The method mainly uses ontology semantic linkage. Zhong Yanqiu
et al.(2012) extracted general feature elements and their relations from scenarios and proposed a
scenario meta model. Based on this research they proposed acknowledge element-based scenario
concept model on specific fields. The scenarios which decision makers face are instantiations of
the model.
(3) Knowledge element extraction and implementation
Wen Youkui et al. (2005) discussed scientific paper knowledge innovation and creation,
knowledge value-adding management, knowledge integration and application. They also analyzed
the representation of innovation points and conducted experiments on the mining of innovation
points. The result showed that innovation point based knowledge element mining was an effective
method of text knowledge mining. Leng Fuhai et al. (2013) proposed a composite semantic
information extraction method. The method consists of semantic labeling, rule extraction and
regular expression. They extracted research methods, performance indicators from the scientific
literature. The method not only keeps the original semantic content of the scientific literature,
but also represents the key innovation content of the scientific literature. Zhou Ning et al. (2006)
proposed a knowledge element representation and extraction model. The model segments a
document into several paragraphs and parses basic knowledge elements from the paragraphs.
Knowledge elements are represented using structure, length and content constraints. Knowledge
element extraction is conducted through three processes, i.e. structure analysis, length analysis and
content analysis. Zhu Liping et al.(2015) analyzed three elements in the introduction section, i.e.
background, problem analysis and contribution. They summarized the common sentence structure
and features in introduction section and extracted the three elements using these rules. They ran
automatic classification tasks on biological and medical full-text literature and categorized sentences
to one of the three elements.
(4) Automatic rule extraction
Researchers in Universitaet Dortmund use the unsupervised neural network in the extraction
of rules from factual data. Then they transform these rules to PROLOG rules (Ultsch, 2006).
Xie Mengjun et al. (2002) proposed an automatic rule extraction method. Sun Chen et al. (2000)
argued that although neural networks are widely adopted in various fields, it is difficult for us
to understand the knowledge in the trained model. Thus we can extract rules from the neural
network and use the rules to represent the knowledge hidden inside. Hou Guangshen et al.(2000)