Page 12 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 44
P. 12
ZHANG Xiaolin / Disruptive changes and the post-library era: Toward supply-side structure reform of knowledge services 011
class universities to cultivate innovative and social cultural talents. In the process, it has become
essential abilities for studies and research to analyze knowledge and its layout, trend, gap,
unusualness, conflict and transition, and to explore possibilities, roadmaps, policies, etc. Users
need knowledge resources which are computable and analyzable, and tools and methods for
analysis more than services. By developing users’ abilities in knowledge mining and analysis, users
can integrate knowledge discovery with their own reflections.
2.2 Redefining knowledge discovery
1) Smart knowledge retrieval. In the past we normally considered knowledge discovery the same
with retrieval and access. However, in the age of big data and machine learning, smart retrieval has
been the new normal (Luther, 2016), such as Google Scholar, Bing Scholar, integrated platform of
EBSCO and Ex Libris, the smart search engine of the National Science and Technology Library
(2017), Semantic Scholar (2017), Yewno (2017), Meta (2017), etc. Their main features are to
provide linked retrieval to other knowledge objects related to retrieval objects to support further
exploration.
2) Knowledge mining and analysis. Based on computable digital contents, further knowledge
discovery can be realized by bibliometrics, text analysis, knowledge object extraction, content
summary and machine learning. For example, Mapping science structure based on co-citation
analysis and gravity matrix clustering method can identify subject domains and track the evolution
of themes (X. M. Wang, Han, Li, Chen, & Zhang, 2017). By means of object extraction and
knowledge ontology, key points of American clean energy policies and the dynamic changes were
analyzed to depict the evolution (J. H. Liu, 2017). By means of text structural analysis, knowledge
mining and structural clustering, research fingerprint of methods, process, parameters and results
were discovered to make mining and comparison of solutions available (Qian, Zhang, & Q. Wang,
2017). By means of text mining, roadmap texts were analyzed to construct multiple roadmaps (Xie
& Zhang, 2017). These are only a few examples. Scientometrics (Qiu, Zhao, & Dong, et al., 2017),
mapping knowledge domains (Z. Y. Liu, Chen, & Hou, et al., 2008) and even data mining of social
media have become popular.
3) Knowledge discovery. Knowledge mining and analysis aforementioned explores knowledge
already known. Discovery of unknown knowledge based on big data and machine learning has
recently become hot topic in study and practice. For example, the National Aeronautics and
Space Administration (NASA) (2017) discovered two exoplanets using artificial intelligence
from Google. The BioAI platform introduces predictive visual analytics to summarize the
literature, detect trends and uncover hidden connections (Business Wire, 2016). These methods
have been applied earlier in science, such as high-throughput screening and materials informatics
(Materials Informatics, 2017), but now these methods begin to be applied in scientific and
technical literature. In humanities and social sciences, articles in Wikipedia were mined to find