jskos-metrics

Scripts to create statistical analysis of JSKOS data

See https://observablehq.com/@nichtich/jskos-metrics for a demo.

Installation

You need to clone the repository or copy its content to a local directory.

The scripts require jq and standard Unix command line tools (bash, sort, uniq, perl...).

To call the main script jskos-metrics from anywhere, add a symlink from a directory in your $PATH, e.g.:

cd ~/.local/bin/
ln -s $DIRECTORY_OF_JSKOS_METRICS/jskos-metrics jskos-metrics

Usage

Run script ./jskos-metrics with a item type (concepts or mappings) and a .ndjson file. On success the statistics are emitted in JSON format:

Metrics of concepts contain the following keys:

• keys - histogram of JSKOS field names
• conceptNumber - total number of concepts
• broaderDistribution - histogram of number of broader terms
• narrowerDistribution - histogramm of number of narrower terms
• narrowerDistributionImplicit - histogramm of number of narrower terms, references broaderDistribution for Values
• topConceptOf - number the top concepts
• typeDistribution - histogram of concept type URIs
• levelDistribution - histogram of concepts per hierarchy level

Metrics of mappings contains the following keys:

• keys - histogram of JSKOS field names
• fromSchemeDistribution - histogram of source scheme URIs
• toSchemeDistribution - histogram of target scheme URIs
• typeDistribution - histogram mapping type URIs
• creatorNames - histogram of creator names
• creatorNumber - histogram of number of creators per mapping
• createdPerDay - number of mappings created per day
• modifiedPerDay - number of mappings modified per day
• fromNumber - histogram of number of source concepts
• toNumber - histogram of number of target concepts
• fromConceptsCount - number of distinct source concepts
• toConceptsCount - number of distinct target concepts
• mappingURICount - number of mapping uris

Implementation

Each metric id calculated with a script of its own. Each item type has a main script that executes its scripts and emits a JSON file:

• concepts/concept-metrics.sh - concept scheme metrics, main script
• mappings/mapping-metrics.sh - mapping metrics, main script

Run make to execute jskos-metrics with examples as unit test.

Examples

Directory examples contains sample files which are also used for testing.

./jskos-metrics concepts examples/concepts.ndjson
./jskos-metrics mappings examples/mappings.ndjson

More example results are include in the jskos-data collection of vocabulary data and their data visualization at https://observablehq.com/@nichtich/jskos-metrics.

Related work

KOS metrics have best been summarized by Stock (2015).

• Wolfgang G. Stock: (2015) Informetric Analyses of Knowledge Organization Systems (KOSs). https://arxiv.org/abs/1505.03671 (published in: C. R. Sugimoto (Ed.): Theories of Informetrics and Scholarly Communication. De Gruyter, 2015)

• Gangemi, Catenacci, Ciaramita, & Lehmann (2005): A theoretical framework for ontology evaluation and validation. (PDF available)

• Owens (2004): Thesaurus evaluation

Consistency checks of KOS have been implemented by

• Skosify

Some more background information can be found in the internal GBV wiki. We are going to also make public this drafts.

License

This scripts can be used without any restrictions (CC Zero).