Page 163 - Journal of Library Science in China, Vol.47, 2021

162 Journal of Library Science in China, Vol.13, 2021

and presentation of individual archives to the organization and association of memory units.
(1) Positioning. This process means avoiding overemphasis on political, racial, religious, and
other subcultural biases, allowing the various named entities in archival data to be presented
to researchers and users without reservation, while establishing “entity-text” or “entity-photo”
source relationships. In the traditional perspective of history and archival research, various entities
within archives often play the silent and filtered role. Historians and archivists select entities from
archives for compilation and research on the basis of their personal experience, values, morality
and methodology formed by long-term training. In this process, it is not uncommon to artificially
add semantic relationships between entities and predetermine the narrative conclusions. The
greatest value of the DH perspective is that it highlights the spatial sense and objectivity of the
researcher himself and the archives he studied, and can anchor the “memory entity” contained in
the archival data by using corresponding technologies, which avoids the problems of subjective
choice to a certain extent.
(2) Context recognition. In archival data, the memory entity that has been located and processed
often plays the grammatical role of subject, object, predicate and so on, embedded in a specific
archival discourse or data set. In fact, the memory entities contained in the archival data do not
exist alone, but are closely related to other memory entities in the same paragraph, text or file,
while the correlation coefficient varies slightly due to the change of research perspectives and
dimensions. Therefore, it is necessary to put specific entities in the context environment for
recognition under the premise of concealing their syntactic characteristics to a certain extent, so as
to deepen the understanding of entities and their semantic environment, and improve their basic
attributes as memory nodes. In this way, by using the DH methods to elaborate archival data,
[26]
archives will be understood more as a space of activity than as a record of certainty and absence .
(3) Mining. After positioning and context recognition, the memory entity can form an archival
memory unit based on a specific context scene, reflecting the memory fragments formed by the
entities in series under a specific organizational mode. The volume and dimension of this memory
unit are not fixed. They may exist in the same sentence, or be a collection of multiple paragraphs
and chapters. The same point is that the memory unit generally contains memory points, memory
objects, memory scenes and other entity elements that constitute memory, forming a weak
association and pre-clustering between various memory elements. At this time it is necessary
to mine and store the memory points together with the context, and complete the “discovering”
process of archive data with the archival content as the core on the basis of concealing biases, and
retain the right to speak in the “self” (entity referent in mainstream cultural memory) and “other”
(entity referent in non-mainstream cultural memory) in archival data.

2.3.2 Reorganizing: multi-dimensional organization of archival data
Knowledge organization systems such as ontology and semantic knowledge graph can deeply
penetrate into books, archives and other literature materials, linking knowledge units with same