Reconciling taxonomies of electoral constituencies and recognized tribes of indigenous Taiwan

Abstract

Over the years, information science professionals have been studying biases in Knowledge Organization Systems (KOS), for example, bibliographic classifications. The robustness of classifications has been examined in diverse measures, ranging from the representation of race, gender, ethnic minorities, to indigenous peoples. In this study, we aim at (a) uncovering implicit assumptions about minorities in everyday taxonomies; (b) comparing and reconciling these different taxonomies. Specifically, we study the use case of Taiwanese Indigenous Peoples' tribe classifications and the indigenous constituencies of the legislature electoral representation. We compare four finer‐grained taxonomies for indigenous people with the coarse‐grained indigenous peoples' electoral constituencies that only recognize two regions (Lowland, Highland). The four taxonomies are: the recognized tribes in the past, the recognized tribes in the present, other possible tribes, and re‐scaled groups based on population. We employ a logic‐based taxonomy alignment approach using Region Connection Calculus (RCC‐5) relations to align these taxonomies. Our results show different options when modeling and interpreting the use case of Indigenous Taiwan constituencies, and also demonstrate that multiple perspectives can be preserved and co‐exist in our merged taxonomic representations.

A systematic review of methods for aligning, mapping, merging taxonomies in information sciences

“…We classified these two studies as logic-based approaches because these relations can be formally written into logic axioms. In studies relating to the use of Euler/X approach (Cheng and Ludaescher, 2020; Cheng et al ., 2017, 2020; Franz et al ., 2015, 2016; Johnston, 2016; Thau and Ludaescher, 2007), qualitative reasoning technique such as Region Connection Calculus was used to denote five base relations (RCC-5): equivalence, proper part, inverse proper part, disjointness and overlapping. The number of relation types used in the mapping process is the major difference between Raunich and Rahm’s (2014) ATOM and the Euler/X approach.…”

“…(2016), with the Euler/X approach, attempted to align multiple taxonomies by aligning two at a time. The Euler/X approach could potentially align three or more taxonomies at a time, but at present the studies were still mainly working with pairwise alignments (Cheng and Ludaescher, 2020; Cheng et al ., 2017).…”

Systematic Comparison of Data Models used in Mapping Knowledge Organization Systems

“…In this exploratory study, we examine six KOS mapping projects and the data models used in their studies. The six different projects are selected to represent a diverse set of KOS: two projects are mapping metadata schemas (Harping et al, 2000; Library of Congress, 2008), two projects are mapping taxonomies (Franz et al, 2016; Cheng & Ludäscher, 2020), and two projects are mapping ontologies (Raunich & Rahm, 2014; Jung, 2008). The purpose of this study is to compare the data models used across mapping projects and ask the following research questions: What kinds of information are made explicit/implicit in the data model used in these projects? What are the potential information gains/losses from the data model used in these projects? …”

“…Tabular models are used in the Getty Research Institute project to crosswalk 15 different art‐related metadata standards (Harping et al, 2000), Library of Congress's MARC to Dublin Core Crosswalk (Library of Congress, 2008) and Franz et al (2016)’s study on aligning 11 biological taxonomies of Andropogon complex across 126 years. Tree models are used in Raunich & Rahm (2014)’s ATOM project, Cheng & Ludäscher (2020)’s Indigenous Taiwan project, and Jung (2008)’s Virtual Organization project. Below we discuss the differences between the models that are discovered by inductive qualitative analysis.…”