=Paper=
{{Paper
|id=Vol-3805/ICBO-2022_paper_5047
|storemode=property
|title=A Data Structure for the Implementation of Referent Tracking Systems
|pdfUrl=https://ceur-ws.org/Vol-3805/ICBO-2022_paper_5047.pdf
|volume=Vol-3805
|authors=Lauren Wishnie,Alexander Diehl
|dblpUrl=https://dblp.org/rec/conf/icbo/WishnieD22
}}
==A Data Structure for the Implementation of Referent Tracking Systems==
https://ceur-ws.org/Vol-3805/ICBO-2022_paper_5047.pdf
A data structure for the implementation of referent tracking systems
Lauren M. Wishnie1 and Alexander D. Diehl1
1
University at Buffalo, 77 Goodell Street, Buffalo, NY, USA
Abstract
Our goal is to implement a Referent Tracking paradigm in tandem with
a fully axiomatized Alzheimer’s Disease Neuroimaging Initiative (ADNI)
knowledge base.
Keywords 1
Referent tracking, Alzheimer’s disease, ontology, knowledge base
Referent Tracking (RT) seeks to address a design flaw often seen in databases wherein the majority of
assertions lack an explicit reference to what the assertions are about. Referents are the entities that assertions
are about. RT originated as a response to this problem in the context of Electronic Health Records (EHRs)
but the problem exists in research-derived data as well. One of the primary tenets of the RT paradigm is the
assignment of unique identifiers (here referred to as RUIs) to all referents and the assertions made about
them, so that changes in the data mirror either changes in reality itself, resolves an error, or comes from
changes in the users’ understanding of reality. In RT, these changes are tracked using predicates called
tuples. There are a total of eight tuple types, each of which holds a specific type of information related to
the assertion. Importantly, the various tuples allow us to make explicit changes in relations between
particulars and particulars as well as particulars and universals. Thus, RT frameworks are designed to work
in tandem with realism-based ontologies. Our goal was to create a relatively simple framework for
generating tuples as part of tracking changes in the Alzheimer’s Disease Neuroimaging Initiative (ADNI)
data. The RT implementation will be supported by the ADNI Ontology, which is currently in development
and will be annotated with the ADNI data, forming an ADNI knowledge base. This RT data structure was
built in Python. Each tuple type is generated by a function which is specifically designed to adhere to that
type’s structure. The data structure accepts axiomatized assertions, which express relationships between
different types of entities, including real-world referents and ontology terms. The program then places the
assertion into the appropriate slot in the tuple template. Next, the tuples are compiled into lists, which are
then converted into dataframes using the .DataFrame() function in the Pandas library. The creation of each
tuple is timestamped using the datetime library. The UUID library is used to auto-generate new RUIs for
every new tuple and any other relevant portions of reality. Implementations of the RT paradigm vary
depending on what kind of data are in the database and how changes are currently done. In the case of this
work, the modes of input for the data structure have been developed to track changes in the ADNI data.
The ADNI data are longitudinal, with several measurements per participant across multiple years. A major
goal of this work is to create an RT system that allows the tracking of events in the evolution of a
participant’s medical history. Possibilities for future work include using RT databases as a basis for the
visualization of patient timelines and other data elements.
ICBO 2022, September 25-28, 2022, Ann Arbor, MI, USA
EMAIL: lmwishni@buffalo.edu (A. 1); addiehl@buffalo.edu (A. 2); ORCID: 0000-0002-7245-3450 (A. 1); 0000-0001-9990-8331 (A. 2)
2022 Copyright for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
©️
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings