We analyse the evolution of vocabularies on the Linked Open Data cloud. Based on the recent statistics of the LOD cloud, we have selected the twelve most dominant vocabularies in terms of their use in di erent pay-level domains. The number of versions we found for these vocabularies range between 2 to 11. While some ontologies exist for more than 10 years (e.g., FOAF) others are only online since a few years (like DCAT). Our analysis shows that many changes occurred on annotation properties. This re ects a need for more clari cation of the terms, especially at early versions of the vocabularies. The majority of changes in the vocabularies are due to changes in other, imported vocabularies. Thus, there is a co-evolution of di erent vocabularies. This insight has practical impacts to ontology engineers. They not only need to consider the evolution of the vocabularies they directly use, but also those they import and indirectly depend on.
The Semantic Web main objective is to give data in the web a well-de ned
meaning. Those meanings can be represented using ontologies, which can be de ned
as a descriptive form for the concepts and items in a speci c eld or domain
and provides speci cations for those items and its relations to other concepts
[
Creating a new version of an ontology requires processes to handle and
manage multiple versions of that ontology [
Copyright held by the authors.
In this paper, we focus on analysing the changes that occurred on a selected set of vocabularies. We analyse them from di erent perspectives and observe how vocabularies are in uenced by changes made in their dependency vocabularies. Furthermore, we analyse the types of changes that occurred in vocabularies terms (classes and properties). Those changes can be additions, deletions, modications, or renaming of vocabulary terms. Other changes such as splits/merges can be considered as additions/ deletions processes.
We clari ed the percentage of changes occurred on the examined vocabularies that caused by the external vocabularies they depend on establishing their vocabularies, and which vocabularies are depend on their terms on establishing and evolving the vocabulary. All those analyses are useful for ontology engineers when they are establishing a new ontology or updating an existing one by having a clear idea about dependencies and relations between vocabularies to choose vocabularies terms that meet their needs.
The remainder of the paper is organized as follows: In Section 2, we present and discuss related work. We present our methodology for analysing vocabularies in Section 3. In Section 4 we give an overview for the examined vocabularies. In Section 5 we make a discussion of vocabularies evolution. Conclusion and future work in presented in Section 6. 2
Current research focused on measuring what are the changes in the LOD cloud,
but not how they are changed. For example, Dividino et al. [
Schmachtenberg et al.[
Based on their report, we selected the top used vocabularies in their crawled
subset of the LOD Cloud and have di erent available versions to download. Our
methodology can be expressed in the following steps:
{ Selection criteria: We chose the vocabularies that have been used in more
than ve datasets (0.49% of 1014 datasets used in the statistics [
In the following gures and tables, we summarize the statistics of the selected
vocabularies and their di erent versions. According to the 1014 datasets crawled
in the State of the LOD Cloud report (Version 0.4, 08/30/2014) [
Table 2 shows the topical domains for our selected vocabularies. In addition,
we show the percentage of the datasets crawled in [
Table 3 shows the external vocabularies that are used in establishing each vocabulary in our study. From the external vocabularies listed in this table, we can note two things: First, there are three vocabularies (FOAF, DCterm and SWC) stuck on their external vocabularies list that used in establishing their rst edition until the recent one. Second, we can note that there are two vocabularies (a) Total number of created entities (b) Total number of deleted entities (c) Total number of modi ed entities (d) Total number of renamed entities (GN and ORG) have many changes on their external vocabularies list during their evolution period. GN partially changed its external vocabularies ve times during six years, and ORG four times during approximately four years.
The percentage of internal changes compared to external changes in the vocabularies versions are shown in Table 4. We calculate the total percentage for all internal changes occurred through the vocabulary evolution overall its versions; from rst to latest version. We can see that three vocabularies with more than 90% of internal changes (DCterm, GN, and PROV). Furthermore, there is just one vocabulary (Cube) with a percentage of internal changes less than 50%. 5
We observe that most of the vocabularies if they evolves, most of the changes occurred on annotation properties for more explanations (metadata) to clarify classes, properties or individuals. This can be expressed as a need for more clearance for terms, especially between the early versions. Another observation is that vocabularies are highly static w.r.t. the number of external vocabularies used to establish and develop them. We can conclude that the topical domains are fully covered with terms in the existing vocabularies, and if there is a need of change, ontology engineers can modify the existing vocabularies.
In most of the examined vocabularies, the terms that changed were internal terms, i.e. the terms that were created for the examined ontology by ontology engineers, not the terms that are imported from external vocabularies during establishing or evolving the vocabulary. On the other hand, some vocabularies such as Cube, bibo, DCAT, and ORG are changed because some changes in their external vocabularies occurred. For example, in the Cube vocabulary, the percentage of the internal change is 43%, and the remaining percentage of change is caused by other external terms over its four published versions of this vocabulary. The rst version from the Cube vocabulary was published in 27.11.2010, and some of its external vocabularies are DCterm and FOAF. DCterm published its latest version in 14.06.2012, before the next version of Cube (was in 02.03.2013) was published.
Another example is the bibo ontology. Analysing ten versions from that vocabulary, we conclude that 35% of changes are caused by external vocabularies. bibo uses many external vocabularies such as DCterm and SKOS, and both of them had versions between the two published versions of bibo ( rst version was in 03.06.2008 and the latest version was in 04.11.2009).
Vocabularies such as DCterm, GN, and PROV are dependent on their own terms, and most of the changes are made on those terms. For example, in the GN ontology, we analysed eleven versions, and the percentage of internal changes are 97%, which means that when the ontology engineers needs to change, they change their own terms. Another observation in DCterm and PROV vocabularies is that their external vocabularies as shown in Table 4 are small, and they are organized as upper level ontologies.
Cube dcterm/foaf/owl/rdf/rdfs/
scovo/skos/void/xml/xsd bibo Address/dc/dcterm/event/foaf/owl/rdf/rdfs/ skos/time/vann/vs/ wgs84 pos/xml/xsd Event/foaf/ns/owl prism/rdf/rdfs/schema/ skos/dcterm/vann/xml/xsd DCAT dcterm/owl/rdf/rdfs/xml/xsd dc/dcterm/dctype/foaf/owl/rdf/rdfs/ schema/skos/vcard/vann/voaf/xml/xsd GN skos/owl/rdf/rdfs/xml/xsd
The vocabularies in our study are di erent in their evolution period and the number of versions published so far. In our study, we are trying to analyse the evolution behaviour for those vocabularies. Some vocabularies such as FOAF, SKOS, GN, and ORG have many versions; 10, 8, 11, and 10, respectively. GN has eleven versions in the period from 05.10.2006 to 29.10.2012, and ORG have ten versions in the period from 06.06.2010 to 12.04.2014. We think this large number of versions is caused by their topical domain they are used in, especially in social web ontologies.
The last observation is that the vocabularies used in publications, geographic, social web, and government topical domains have the largest number of changes, this is obvious from Fig 1 using SKOS, GN, SWC, FOAF, and PROV vocabularies number of changes. For example, PROV added 29 classes and 62 object properties, and modi ed 51 classes and 80 object properties during its evolution period (Three versions published from 03.05.2012 to 11.01.2015). Another example is SWC, the ontology engineers modi ed 110 classes and 25 object properties through the three published versions. 6
Analysing the change behaviour of vocabularies can help ontology engineers in establishing new ontologies and evolve existing ones. This study can give a clear view about ontology dependencies (External vocabularies), and the relation between change in ontologies and their external ones.
Changes are mostly made for internal terms if they compared by external terms percentage. Furthermore, most of the vocabularies have a static number of external ontologies to depend on during the evolution period, and if they add or remove some vocabularies, the number of those additions and deletions is small.
Topical domains such as publications, geographic, social web, and government have a high percentage for change if they compared with other domains.
As a future work, we will analyse the usage of vocabularies' in the Dynamic Linked Data Observatory (DyLDO) dataset. We will select the vocabularies that have a version before and after a speci c snapshot of a DyLDO crawl (the rst snapshot was in 06.05.2012, the latest is until today). Furthermore, we will include the vocabularies not considered so far in this study. Establishing a framework for ontologies' concepts history tracking system will be useful for ontology engineers.
Acknowledgement This work was supported by the EU's Horizon 2020 programme under grant agreement H2020-693092 MOVING.