-
Notifications
You must be signed in to change notification settings - Fork 1
Ontology
- Pier Luigi Buttigieg
- Hiromi Kanegae
- Kouji Kozaki
- Robert Hoehndorf
- Chris Mungall
- Tatsuya Kushida
- Atsuko Yamaguchi
- Hiroshi Masuya
The core issue of this project was to create an application ontology to provide a semantic framework for experiments evaluating the effect of growth conditions and varietal on the properties of crops destined for agronomic markets. Importantly, we sought to ensure that this ontology interoperates with existing reference ontologies while showing local fitness-for-purpose.
We began from the development of draft ontology for soy bean which had placed elements of the Plant Anatomy Ontology (PO), Trait Ontology (TO), Phenotypic Quality Ontology (PATO), the Environment Ontology (ENVO), Flora Phenotype Ontology (FLOPO) and Interlinking Ontology for Biological Concepts (IOBC) [ANY OTHERS?].
In addition to the imported classes, the scheme included mappings to Japanese vocabulary standards such as Agricutlure Activity Ontology (AAO) and Crop VOcabulary (CVO) as well as new classes specific to our application.
While this ontology provided a firm basis, optimisation and further alignment with reference ontologies and more consistent use of object properties (logical relations) was needed to increase its value, especially to machine agents.
The working group created a refined and simplified application ontology scheme which 1) removed classes which were either extraneous or ambiguous, 2) specifies more consistent use of axioms across imported and locally minted terms, and 3) provides patterns to extend this work to other applications with differing crops, environmental conditions, and/or measured properties.
Further, the working group reported inconsistencies or points of semantic ambiguity back to the respective reference ontologies, which have marked them for review.
[ADD IMAGE]
In conclusion, our collaboration of local practitioners with developers of reference ontologies during this hackathon allowed us to rapidly refine our application ontology for soy bean cultivation in an experimental setting, further aligning it to reference ontologies in the process. We were also able to motivate discussions in several of those reference ontologies to refine parts of their representations where we detected ambiguity.
Integration with ontologies which represent protocols and planned processes is desirable, to allow methodology and reporting on experimental processes to be placed within our model. The same is true for sensor ontologies as well as more Japanese terminological resources to increase national relevance of our work.
Further attention to the semantics of the use case's environmental setting and experimental conditions is also desirable, to enable more automated discovery of and reasoning over links between environments, genotypes, and phenotypes.
We also intend to create a functional OWL artifact based on our model. We plan to do so using the Ontology Development Kit to 1) maximise interoperability with exisitng ontologies 2) allow us to use standardised development procedures, and 3) ensure that our imports are logically sound and, through continuous integration, checked for consistency.
To enhance community impact of our work, we will attempt to contribute any classes created within the application ontology that show more general value to the appropriate reference ontologies.
- Atsuko Yamaguchi
- Kouji Kozaki
- Keiichiro Nishiyama
RDFS/OWL can be useful to search data efficiently from LOD if data follows data models defined by them. To use restrictions from OWL/RDFS for search of RDF dataset, how real datasets follow restrictions should be investigated. To do so, RDF dataset validation using restrictions with owl/rdfs are considered.
We selected six types restrictions to validate as follows.
- ex:Domain rdf:type rdfs:Class. ex:Range rdf:type rdfs:Class. ex:property rdfs:domain ex:Domain; rdfs:range ex:Range.`
- ex:Domain rdf:type rdfs:Class. ex:property rdfs:domain ex:Domain; rdfs:range xsd:(datatype).
- ex:Class1 rdf:type rdfs:Class. ex:Class2 rdf:type rdfs:Class. ex:Class1 rdfs:subClassOf ex:Class3. ex:Class3 ex:p owl:Restriction; owl:onProperty ex:prop; owl:someValuesFrom ex:Class2.
- ex:Class1 rdf:type rdfs:Class. ex:Class1 rdfs:subClassOf ex:Class3. ex:Class3 ex:p owl:Restriction; owl:onProperty ex:prop; owl:someValuesFrom xsd:(datatype).
- ex:Class1 rdf:type rdfs:Class. ex:Class2 rdf:type rdfs:Class. ex:Class1 rdfs:subClassOf ex:Class3. ex:Class3 ex:p owl:Restriction; owl:onProperty ex:prop; owl:allValuesFrom ex:Class2.
- ex:Class1 rdf:type rdfs:Class. ex:Class1 rdfs:subClassOf ex:Class3. ex:Class3 ex:p owl:Restriction; owl:onProperty ex:prop; owl:allValuesFrom xsd:(datatype).
As a first challenge of development of IoT (Internet of Things) device which exchanges RDF data, we have developed a practical environment observation system which works under harsh conditions such as in the forest.
The goal of the implementation is collecting observed RDF data generated by sensor-hosting devices automatically into an RDF triple store via WiFi. The device is powered by solar battery and is capable of continuously working even in a dark forest. In order to realise such device, we employ Arduino (CPU: ATmega328P having 2kB RAM and 32kB ROM) and small-scale ontology for describing small observation RDF data.
The working group created a minimal, application-level Environment Observation Ontology, reusing content from the Semantic Sensor Network Ontology and the Ecosystem Ontology (which interoperates with ENVO for more global reach), which describes concepts including Platform, Sensor, Observation and Observable Properties. The description capability was evaluated by writing sample RDF for an Arduino platform having 2 digital sensors (BME280 air temperature, air moisture and air pressure sensor and TSL2561 light intensity sensor) and 2 analog sensors (LM25DZ soil temperature sensor and SEN1004 soil moisture sensor). Finally, we wrote Arduino program for hosting these 4 sensors.
We published the Environment Observation Ontology and Arduino program through GitHub (https://github.com/dbcls/bh19/tree/master/ontology/EnvironmentObservationOntology).
We will implement hardware and deploy it in a small forest in the RIKEN Wako Campus as a model forest, and prove that our semantic-web-based IoT concept is efficient through the deployment. We will also explore how to compress semantically qualified data packets to save more energy in the field, as well as how to flexibly extend the minimal model using design patterns in ENVO and related ontologies.
Biology has context: the ecosystems in which biological entities are embedded define many of their properties and behaviours. However, there is still much to do to link ontologies and semantic web resources serving the life sciences to those in Earth and planetary science, as well as those serving global development. A large part of this issue is the lack of exchange between developers and users of ontologies in these domains.
Many users of ontologies across the domains of life science, planetary science, and socio-economic development use reference ontologies to provide semantic consistency to their diverse efforts. However, often these communities are reluctant to engage with ontologists in domains outside their own, especially if those ontologies follow unfamiliar development methods.
The Envionment Ontology (ENVO) provides a rare point of overlap between these fields, with its content and userbase extending into both domains. It has thus become a nexus between life sciences (via the OBO Foundry), Earth sciences (via the ESIP Federation), and sustainable development (via efforts to provide semantic interfaces to the 2030 Agenda for Sustainable Development led by UN Environment).
By the end of this hackathon, the following phenomena were identified:
- Halo phenomena in Kazakhistan
- Red tide in San Diego
- Green flash
- Akagi-oroshi / Karakkaze an andesitic stratovolcano
- Erratics
- Clints
- Grykes
- Glyders
- Aits
- Meols
- Zawns
- Stors
- Karl the fog
- STEVE
- The Aral Sea shrinking phenomenon
- The Kerala backwaters
- Japanese rain expressions (including Guerilla rainstorms)
- A lake variant in Kazakhstan's Big Almaty Lake
- A canyon variant in Kazakhstan's Charyn Canyon
- Haar
- Atmospheric inversions, including smog inversions
- Shiranui (不知火)
- The Hikari no Michi at Miyajidake Jinjya
- The Kentucky River Pallisades
- A valley variant in La Vall de Núria
- A mountain variant in the Queralt mountains
Work is now underway to upgrade those sections of ENVO which contextualise these phenomena, before adding the phenomena themselves to the ontology. For examples, see issues for atmospheric ghost lights and meols.
This mini-project engaged a number of BioHackthon participants in the development of ontology content for Earth and planetary science, as well as sustainable development and culture. The familiarity of all participants on the role of Earth and development ontologies in contextualising biology was enhanced, and the ontologies themselves benefitted from the international perspectives of the BioHackathon participants.
The content captured in the BioHackathon will be woven into ENVO, SDGIO, and - as needed - associated ontologies to provide rich expression for multiple applications. They will be made available to all users over the next few release cycles, with the submitters and the BioHackathon event nano-credited.
In the near- to mid-term, we hope that efforts such as these brings communities across the domains of life, planetary, and development sciences and operations closer together around linked data technologies.