In the new information environment, the way of knowledge organization based on documents cannot meet the users knowledge needs.In order to achieve fine grained semantic knowledge organization and refined knowledge service, the fine grained knowledge units contained in the documents become one of the hot topics in multi disciplinary research. Revealing the connotation and structure of the knowledge unit becomes an important basic theoretical problem to be studied urgently. Based on the existing concepts, classifications and descriptions, we further clarified the connotation of knowledge unit. From the perspective of knowledge production and creation, the knowledge unit is divided into three types which include common sense knowledge unit, citation based knowledge unit and innovative knowledge unit. And then, from the point of theoretic and empirical analysis, we proved the hypothesis of “a knowledge unit is the logical combination of N (N≥1, N is a positive integer) semantic triples”. In addition, in order to promote the management and use of knowledge unit, we construct the semantic description model of knowledge unit referring to the idea of nanopublication.
Knowledge unit refers to the smallest unit of content that expresses a relatively complete set of specific knowledge semantically. Common sense knowledge unit is a generally acknowledged basic knowledge content in public domain and professional subject area. Its citation often does not need to explain or label specific sources and is used to prove the innovative knowledge unit. Citation based knowledge unit refers to the text content unit that can represent the citation of references. Citation based units are usually used to demonstrate or support innovative knowledge unit. Corresponding to common sense unit and citation based unit, the innovation knowledge unit is innovative professional knowledge, which is a new discovery or new cognition about the specific problem. The innovation knowledge unit can be decomposed into at least one semantic triples of subject, predicate and object. There are logical relations such as coordinate, conditional, explicative, causal, progressive and comparative relations among the N (N≥1, N is a positive integer) semantic triples of one knowledge unit. Meanwhile, there are also semantic relations such as synonym relation and implication relation.
The description model of knowledge unit consists of content of knowledge unit, provenance and supporting information of content, and publication information of knowledge unit. The content of knowledge unit is a scientific discovery or core scientific conclusion in the document. And it contains three description elements: subject, predicate and object. The publication information is the metadata about the knowledge unit itself. The description elements of publication information mainly include ID, name, creator, creation organization, creation time, regulatory agency and version number of knowledge unit. Supporting information is the background or context information of the content of knowledge unit, which is used to ensure its scientificity and accuracy. Descriptive elements of supporting information mainly include the research method, research data, research object, research tool and theoretical basis for generating innovative knowledge units, as well as the preconditions for the establishment of knowledge unit. Provenance information is the source characteristic information of knowledge unit content, which is used to ensure its traceability, trustworthiness and innovativeness, and to ensure the intellectual property of its creator. The description elements of provenance information mainly include source literature, source journals, source authors, type, language and publishing time.
Due to limited space, this paper selects only one Chinese paper and one English paper in this subject area for case study. In the follow up, a large sample of examples will be selected to explore the characteristics and structure of knowledge unit to further verify the research hypothesis proposed in this paper. In addition, we only select one example of knowledge unit to verify its descriptive model. 3 figs. 6 tabs. 52 refs.