iDAI.chronOntology Wolfgang Schmidle, Nathalie Kallas, Sebastian Cuy - - PowerPoint PPT Presentation

CoDArchLab Cologne Digital Archaeology Laboratory i3mainz Institut für Raumbezogene Informations- und Messtechnik Hochschule Mainz

iDAI.chronOntology

Wolfgang Schmidle, Nathalie Kallas, Sebastian Cuy (DAI) Florian Thiery (i3mainz) Digital Classicist Seminar Berlin, 10.01.2017

Time gazetteers

Arachne

• no term definitions at all (on purpose)
• no dating information, can be added for individual items
• will move to chronOntology

Portraitbüste: Augustus

https://arachne.dainst.org/entity/1075871

Time gazetteers

Getty AAT

• very large, "Styles and Periods" facet has more than 5000 terms
• LOD
• descriptions but no real definitions
• spatial information only in free text
• inconsistent use of period types
• simple hierarchy

Augustan http://www.getty.edu/vow/AATFullDisplay?find=&logic=AND&note=&subjectid=300020543

Time gazetteers

PeriodO http://perio.do/

• LOD, used in ARIADNE project
• regions and timespans as part of the definition
• Periods as objects of discourse

Time gazetteers

FORTH thesaurus

• demonstrator for an "ideal" time gazetteer
• periods as concepts
• definitions/types and relations
• only 8 terms from one source
• bounding boxes (Spacetime volumes)
• will be ingested by ChronOntology

http://www.ics.forth.gr/isl/publications/paperlink/CIDOCpaper1_Doerr.pdf http://cidoc-crm.org/docs/period/xsl_output.html

iDAI.welt

ChronOntology

counterpart to iDAI.gazetteer (http://gazetteer.dainst.org/)

• part of iDAI.welt, guaranteed availability after end of project
• temporal norm data for the DAI
• LOD: stable URIs
• robust system for large amounts of data
• accept data of differing quality and completeness
• project partners can enter their data
• batch imports and user interface
• minimal goal: "get IDs for your temporal norm data"

funded by German Research Foundation (DFG), 2015 to 2017

ChronOntology

Implementation

Frontend and widgets: http://chronontology.dainst.org

• based on AngularJS
• beta version!
• Search
• Visualization of single datasets
• browse space and time
• on GitHub

Implementation

Backend http://chronontology.dainst.org/data/period/rjqJUoIToI7L

• LOD hub, REST API
• Java
• JSON as data format
• elasticsearch
• on GitHub

Goals

• take care of time-specific challenges, e.g. definitions
• ingest data from electronic sources and secondary literature
• accommodate conflicting definitions and opinions
• easy to use for minimal datasets
• linking with other (time) gazetteers

Design principles

attempt CIDOC CRM compatibility where possible:

• compatible with CIDOC CRM and extensions (periods as concepts)
• use E4 Period as basis (sets of coherent phenomena or cultural

manifestations occurring in time and space) —> but e.g. Murad II

• add properties only if there is evidence in the data

Design principles

P1: A dataset describing a temporal term is not identified by its name, but a unique ID.

This is a well-known general principle, just as a John Smith cannot be identified by his name alone. Using a unique ID allows creating different datasets for different usages of the same term, and one dataset can have more than one name associated with it.

Design principles

P2: A temporal term can not defined by dating information.

The definition and any dating information based on this definition should not be

• confounded. Only the definition determines whether we are talking about the same

thing or not (co-reference), and only when a term has a definition or at least a type such as “political” or “material culture” one can meaningfully associate dating information with it. Similarly to a dictionary entry, a temporal term can have more than one meaning, each potentially leading to different datings. A temporal term may not even have any known explicit dating information, for instance when it is only part of a relative chronology. Likewise, a temporal term can be defined without knowing anything about its spatial extent.

• However, we strongly encourage adding explicit temporal and spatial information

wherever it is missing

• Add information by reasoning (see gazetteer)

Design principles

P3: Each defined temporal term is actually a spacetime volume (STV)

I.e. the area in space and time where it happened, regardless of what we actually know about this STV. Any explicit spatial and temporal information given about the term approximates this STV. Follows from using E4 Period.

4500 BCE 3700 BCE 2900 BCE 2100 BCE 1300 BCE 500 BCE W

• r

l d w i d e N e a r E a s t C e n t r a l A s i a E a s t A s i a S

• u

t h A s i a S

• u

t h e a s t A s i a E u r

• p

a S u b

• s

a h a r i a n A f r i c a

IA BA SA

The Bronze Age is a time period characterized by the use of bronze.

Design principles

P4: Mark whether a period is ongoing or not.

This information will have consequences for reasoning over the data, for example for establishing bounding boxes for the period. Different extents may be due to looking at it at different points in time.

Design principles

P5: A link to the time gazetteer may contain additional information.

Especially time information. The details need to be worked out. In any case, if one simply links to a dataset in the time gazetteer without specifying any additional information, the semantics of the connection defaults to the usual “is part of

• r equal to” interpretation of a gazetteer link.

Design principles

P6: The data model should be robust enough to accommodate data with varying degrees of data quality and completeness.

Temporal term data that was not designed for this time gazetteer should fit

• nonetheless. All typical statements about temporal terms should fit in.

The data to be imported should be modeled semantically in a form that may contain imprecise statements, but explicitly wrong statements should be avoided. The data model should support a workflow for making the data more precise. Minimal set of information, first ingest without too much preparatory work

Data model

about the dataset:

• ID
• names in different languages, preferred names
• provenance, external ID
• types, including „all meanings“
• same term with different types: political, material culture > pottery
• definition: „standard definitions“ from name and type (real defs. rarely available)
• we would prefer explicit definitions (only for core?)
• description
• tags, notes
• created, modified, version
• dating information
• ongoing

Data model

Relations between datasets: Sense, place, time, matching, other

• by definition:
• hasSense / isSenseOf
• follows / isFollowedBy
• isPartOf / hasPart
• matching: „sameAs“, „same as or finer than“ (more will come as needed)
• lists

Senses

• hasSense / isSenseOf

all meanings, political, etc.

• must be easy to use! no barrier to entry
• matching of similar senses: e.g. style vs. material culture
• complementary to e.g. PeriodO: some ChronOntology data may be

added to PeriodO (as new statements), especially "all meanings"

Connection to the iDAI.gazetteer

Gazetteer: split between Gazetteer and chronOntology connection types: spatiallyPartOfRegion, hasCoreArea, isNamedAfter „core area“, „ca.“: problematic, but this is what is in the ingested data unify data models?

Time

• follows / isFollowedBy
• isPartOf / hasPart
• Allen relations (at the moment only inferred)

internal:

• hasTimespan: timeOriginal, begin At, AtPrecision, End etc., source
• can be repeated

Role of marker events Aegean Bronze Age: Use case FORTH thesaurus, different dating systems

Matching

Data refinement: the better the data, the better the matching Use case: DAI thesauri

Hochmittelalter (um 900 bis 1250) Arabische Münzen Barbarische Nachprägungen

The data so far

Getty AAT periods facet (the biggest chunk) Arachne geological epochs (several sources) Secondary literature: Bronze age in the Levant "chronontology core": still a little experimental

Ingestion workflow

• should be doable in reasonable time, faster than the Getty AAT data
• only model explicitly what is given in the source
• add explicit definition (or at least type) from free text notes
• add explicit explicit temporal and spatial information if possible
• Statements should be at least not wrong („equal to or restricted to

sense“)

• can be made more precise at any time (different dataset)
• tested with different levels of pre-processing for different parts of the

AAT

Ingestion workflow

• Core terms / ingested terms
• matching data: "sameAs"

http://chronontology.dainst.org/search?q=hadrianic

Ingestion workflow

Minimal information: similar to PeriodO ?

• Type?
• place?
• time? (becomes one opinion in the dataset)
• add provenance

Arachne: just the terms, "all meanings"

Challenges

Distinguish between e.g. “Neolithic, limited in a strictly geographic sense to the Levant” and “Neolithic in the Levant” as the latter may explicitly or implicitly have a different definition. Data sources inconsistent or unclear: „European Bronze Age coexists with Bronze Age (three-age system)“ List nodes different opinions about whether one can apply “Augustan” to the ancient Greek city of Aphrodisias (now Turkey)

• a matter of precise definition

Reasoning

pairs of properties inheriting the region (not yet implemented) isPartOf (by definition) => is temporally part of

Visualisation

Mapping of actual term usage

Visualisation

Mapping of actual term usage

Visualisation

Gazetteer-based search

Use case

Augustan: meanings of a term, Roman art by Greek artists

Use case

Viking Buddha: interplay of systems, formulate research questions

Named entities?

extend system to accommodate non-modern terms (P5: A link to the time gazetteer may contain additional information)

Next steps

prepare for hands-on session in March 2017: system should be ready to ingest data on the spot

neutral URL Implementation: reach 1.0 version, including data editor extend the data model as needed documentation: best practices, type thesaurus, property hierarchies