Scholarly Knowledge (KS) is the knowledge that has been gained through disciplined, systematic inquiry. It is different from experiential, craft or religious knowledge because it is based on research and observation and is governed by scientific principles of inference. Although scholars may differ in their research approaches, they must adhere to transparent standards of evidence and methodology. KS is distinct from popular knowledge that can be obtained through a wide variety of sources, including the internet, popular and traditional books, television programs, documentaries, and media reports.
In the field of scholarly communication, a Knowledge Graph (KG) is a representation of relationships between entities in the scholarly domain. It enables various applications such as information retrieval, collaboration, citation analysis and research impact assessment. KGs are mainly constructed through a combination of structured and unstructured data from a wide variety of trusted sources. Among the most popular structured datasets for KS construction are bibliographic records from online databases, research articles, conference abstracts, and disciplinary repositories.
This study is concerned with the use of KS to enhance the understanding and utility of research in a global, multilingual academic environment. Through in-depth interviews, we examined the perceptions of scholars across the global academic community regarding what characteristics define quality KS and how these factors are applied during knowledge production. Using the conceptual framework of Harvey and Green, our results show that the five dimensions they identify (academic standards, standard of competence, service standards, organizational standards, and recognition standards) are relevant to many aspects of scholarly knowledge production.
Our participants identified the importance of transparency, ethical and fair practices, and integrity as defining qualities of quality KS. They also emphasized the importance of collaborating with colleagues across disciplinary and national boundaries. These themes are consistent with the notion that scholarship should be viewed as a public good. However, our participants also noted that they are not always able to practice these values in the real world. For example, a researcher in a developing country stated that she relyes on colleagues from the US to send her journal articles because she does not have access to them locally.
A number of scholarly literatures focus on the development and evaluation of KS. They aim to achieve high accuracy, scalability and expressiveness by using a variety of methods. The majority of these methodologies utilize natural language processing and deep learning techniques to process raw text. They employ methods such as word2vec and neural networks to capture semantic meaning and grammatical structure.
Another approach to achieving accurate and expressive KS is by leveraging the contextual information present in existing structured data resources. For this, the KS is enriched using external knowledge bases such as DBpedia and Wikipedia to provide a rich set of entity-relation relations. These enrichments allow KG to better capture asymmetric and anti-symmetric relations, such as paper_is_part_of/paper_was_written_by. Moreover, multiplicative embedding models are employed to find missing links between entities in the scholarly domain. This allows KG to achieve more accurate results for complex relations such as a paper_is_part_of/paper_was_written_by and author_name_email_address/author_name.