Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016 Short Talks IV

These are my notes for the fourth session of talks at the UK Ontology Network Meeting on 14 April, 2016.

Source https://lh3.googleusercontent.com/-Ql7RFDSgvxQ/AAAAAAAAAAI/AAAAAAAAAFA/pnoDTCze85Q/s120-c/photo.jpg 14 April 2016

Mining informative OWL axioms with DL miner
Viachaslau Sazonau, Uli Sattler

A reminder: TBox = OWL Class and property axioms, and ABox = individuals with labels and relations. A common question is “What is missing in my TBox?” The ABox might provide hints… A domain expert could scan the ontology and find the additional axioms “manually”.  DL-Miner automatically scans the ABox and generates hypotheses for the TBox.

If the hypothesis is correct, add to TBox. If incorrect, you then check the ABox as there might be something wrong there. An alternative would be go outside the ontology (to the laboratory, for example) to see if there’s another reason why the hypothesis has been suggested.

Justification and Reasoner Verification
Michael Lee, Bijan Parsia, Uli Sattler

Reasoners are vital and it is essential they are correct as they do tasks impossible to do ourselves. When there are disagreements that have occurred before (ORE) an error must have occurred. They want to find disagreements and resolve them. They will evaluate each justification for a disagreement either with a human or a reasoner.

They look at FaCT++ and HermiT. One reasoner would make a statement, and the other reasoner states whether it disagrees or not. If they can’t make a decision, it goes to a human. They did this with 4 reasoners (also Pellet and JFact) and looked at 190 ontologies. 181 had agreements on classifications, and 9 disagreements. This resulted in 1622 justifications for the disagreements. They found errors with data types and missing asserted axioms.

When you do ontology engineering, make sure you use more than one reasoner. However, reasoners are generally stable with a 95% level of agreement. In future, would be worth making a service where you can submit your ontologies to be compared across reasoners.

Antipattern Comprehension: An Empirical Evaluation
Tie Hou, Peter Chapman, Andrew Blake

Comprehension of justifications is known to be difficult for even experienced ontologists. Even with reasoners, understanding is difficult. They are trying to make things easier with visualization. Most visualization tools show only the hierarchical structure of an ontology, however incoherence in an ontology can arise from the interaction between concepts and properties. Therefore they use concept diagrams which can be viewed individually or merged.

Does visualization make it easier to examine incoherence? A set of antipattern categorizations were extracted from online TONES ontology repo. They focused only on the identification of logical contradictions. Participants using Protege statements did not perform any worse than those using diagrams. They want to extend the study to help debug ontologies. The study was performed with students with no knowledge of ontologies, and they’d like to do it again with experts.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016 Short Talks III

These are my notes for the third session of talks at the UK Ontology Network Meeting on 14 April, 2016.

Source https://lh3.googleusercontent.com/-Ql7RFDSgvxQ/AAAAAAAAAAI/AAAAAAAAAFA/pnoDTCze85Q/s120-c/photo.jpg 14 April 2016

Ontology-driven Applications with MEKON and HOBO
Colin Puleston

Currently used in a sepsis prediction system and a clinical trials design / design-retrieval system (early prototypes). MEKON  is based around a generic frames model with a plugin framework which allows the incorporation of ontology-like knowledge sources and associated reasoning mechanisms. HOBO extends MEKON to enable the creation of domain specific object models, bound to appropriately populated frames models, Instantiations of object models operate in tandem with instantiations of bound frames models.

They come with instance store mechanisms, delivered via plugins. Current plugins are based on OWL DL, RDF + SPARQL and the XML database BaseX. They also come with a Model Explorer GUI to allow the model developer to browse, explore dynamic behaviour of specific instantiations, and exercise the instance store (store instances and execute queries).

MEKON and HOBO provide layered architecture (structured means of combining generic knowledge sources and reasoners, and domain specific processing), handle much of the dirty work, and provide client access via appropriate APIs (generic frames model, domain-specific object models).

MEKON enables the creation of a skeleton application without having to write a single line of code.

The Synthetic Biology Open Language
Goksel Misrili, James Alastair McLaughlin, Matthew Pocock, Anil Wipat

Synthetic biology aims to engineer novel and predictable biological systems through existing engineering paradigms. Lots of information is required, including DNA sequence info, regulatory elements, molecular interactions and more. They are currently using SBOL 2.0. Modules allow hierarchical representation of computational and biological systems. A biological example would be a subtilin sensor.

SBOL utilises existing SW resources, e.g. BioPAX, SO, SBO, EDAM, Dublin Core, Provenance Ontology. SBOL is serialized as XML. SBOL allows the unambiguous exchange of synthetic biology designs, and is developed by a large consortium.

ConSBOL: A terminology for biopolymer design
Matthew Pocock, Chris Taylor, James Alastair McLaughlin, Anil Wipat

The presenter was unable to make the workshop today.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016: Modelling Dispositions

These are my notes for Alexander Carruth’s talk at the UK Ontology Network Meeting on 14 April, 2016.

Some ULOs, such as BFO, do try to model dispositions. What is a disposition? Fragility, solubility are canonical examples. Dispositions are capacities, tendencies, or causal powers (for example). They are the features in virtue of which things engage in particular causal interactions. Other examples are mass and charge.

Traditionally, the dominant account of dispositions is called the Conditional Analysis (CA). This basically says if S occurs, then O will M. S = stimulus, M = some manifestation. Example: if the vase is struck, it will break. This relationally captures a disposition’s nature as D(s,m). There have been some challenges to the CA method in recent years.

Source http://www.danielborup.com/wp-content/uploads/2013/01/Cracks1.jpg 14 April 2016

There are two ongoing debates about the nature of dispositions. The first is the Tracking Debate (single trackers versus multi trackers). This debate concerns the number and variety of manifestations that can be associated with a single disposition. Within multi tracking, there is quantitative and qualitative multi tracking. Multi trackers: Being ball-shaped has a variety of manifestations (e.g. rolling, making a dent in some clay). Therefore dispositions have multiple manifestations produced by multiple stimuli.

The second debate concerns how dispositions operate: CA assumes a stimulus-based account of how dispositions operate. The Mutual Manifestation view states that dispositions ‘work together’, with no distinction possible between the active disposition and the mere stimulus.

Therefore there are four accounts of disposition:

  • single-track stimulus manifestation (CA) D(s,m)
  • Multi-track stimulus manifestation
  • single-track, mutual manifestation D1(D2, m1)
  • Multi-track, mutual manifestation

How should we react? Monism (choose which of the four accounts to go with); pluralism (greater complexity but could pick and choose); or pragmatism (different responses for different purposes)?

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016 Short Talks II

These are my notes for the second morning session of talks at the UK Ontology Network Meeting on 14 April, 2016.

Source https://lh3.googleusercontent.com/-Ql7RFDSgvxQ/AAAAAAAAAAI/AAAAAAAAAFA/pnoDTCze85Q/s120-c/photo.jpg 14 April 2016

Dialogues for finding correspondences between partially disclosed Ontologies
Terry Payne and Valentina Tamma

and-

Dialogue based meaning negotiation
Gabrielle Santos, Valentina Tamma, Terry Payne, Floriana Grasso

Different systems have different ontologies, so you need to align them. Many approaches exist. There are 3 problems: different alignment systems produce different solutions; you may not want to align all of your ontology; part of that ontology might be commercially sensitive and therefore not want to expose or disclose it.

You can use 2 agents to negotiate possible alignments. Agents selectively identify what mappings should be disclosed if the agents are knowledge rich. If not knowledge rich, then the agents need to start exchanging segments of the ontology.

They have started to develop a formal inquiry dialogue that allows two agents to exchange knowledge about known mappings. If you’re looking at many different alignments, you many have many one-to-many mappings. Which mapping should be selected? Could it be resolved through objections within the dialogue? Through the dialogue each agent extends their ontology by including correspondences.

They’ve used a cognitive approach to reaching consensus over possible correspondences. Agents identify possible concepts that may be ontologically equivalent in their respective ontologies. Each then seeks further evidence. One agent asks the other agent if it can align an entity. Start with a lexical match. The second phase will then ask for evidence to support the correspondence, e.g. what are the structural similarities? The CID dialogue has been empirically evaluated using OAEI datasets.

The Cardiovascular Disease Ontolgoy (CVDO)
Mercedes Arguello Casteleiro, Julie Klein, Robert Stevens

CVDKB content: 34 publications, including human and mouse. How can we connect the biological information, e.g. HGNC, UniProt, MGI, ChEBI, miRBase? From this they have 86792 mouse proteins, 172121 proteins from human, and many metabolites and miRNAs. The CVDO reuses some ontologies, such as OBI, and parts of other ontologies including SO, PRO, GO, CL, UBERON, PATO…

There is a CVDO application which allows SPARQL queries over the OWL. They’re looking at including SOLR and elasticsearch to make searching fast, via the conversion of OWL to JSON-LD. Go to http://cvdkb.cs.man.ac.uk to try it out. The idea is to hide the ontological complexity from the end user.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016: The Use of Reformation to Repair Faulty Analogical Blends

These are my notes for Alan Bundy’s and Ewen Maclean’s talk at the UK Ontology Network Meeting on 14 April, 2016.

This talk is divided into two parts: Merging Ontologies via Analogical Blending, and Repairing Faulty Ontologies using Reformation.

Can you merge ontologies successfully using analogical blending? It would be quite easy to get things wrong, and therefore they are using the reformation technique to repair any mistakes made in the merging process.

T1 and T2 are the parent theories, and B is the blend between them. Suppose T1 and T2 are two retailer ontologies. T1 has relationships for owning, and part numbers, and product A and product B have the same part number as they are different instances of the same product. In T2, the relationship is sold_to and there are serial numbers rather than part numbers. So, things are similar but not identical. It would be easy to automatically align these concepts incorrectly. When the ontology is merged, the two products are incorrectly given the same serial number (when they only have the same part number). This makes the ontology inconsistent.

How can the reformation technique help you recover? Reformation works from reasoning failures. Here, we’re looking into inconsistencies. Using the proof of inconsistency, reformation tries to break the proof to prevent it from getting to the inconsistency, and therefore creating a suggested repair, in this case rename the two occurrences of the serial number. The resulting new blended ontology has replaced the serial number type with a part number type, and part and serial number are two different types, thus correcting the ontology.

Ontologies can be merged by analogical blending, but some blends can be faulty. Faults can be revealed by reasoning failures. Reformation uses such failures to diagnose and repair faulty ontologies. This work is still in the early stages.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016 Short Talks I

These are my notes for the first morning session of talks at the UK Ontology Network Meeting on 14 April, 2016.

Source https://lh3.googleusercontent.com/-Ql7RFDSgvxQ/AAAAAAAAAAI/AAAAAAAAAFA/pnoDTCze85Q/s120-c/photo.jpg 14 April 2016

Integrating literature mining and curation for ontology-driven knowledge discovery
George Demetriou, Warren Read, Noel Ruddock, Martyn Fletcher, Goran Nenadic, Tom Jackson, Robert Stevens, Jerry Winter

It is hard to keep up with the volume and complexity of data throughout its life cycle (search for content, collect it, read and analyse it, convert it into formal representations, integrate knowledge into computational models, use it to produce explanations, predictions or innovations). Therefore they have BioHub, which stores information on feedstocks, chemicals, plants, organisms, chemical transformation, and properties. The task is to extract, organise and integrate knowledge into models of chemical engineering. An example question: “Which chemicals come from which feedstocks?”

Where does the human come in for the curation task, and where the machine? DARPA Big Mechanism: a big project to compare based on text evidence from literature. In the study, they found humans are good for finding interactions and bad for grounding. Manchines were bad for interactions and good for grounding. A hybrid method had the best result.

In the BioHub Curation Pipeline, there are different types of annotation, with human and machine curation.

Integrating Concept Dynamism into Logitudinal Analysis of Electronic Health Records
Chris Smith and Alex Newsham

Policies that determine the data captured in EHRs is subject to change over time for a variety of reasons, including updated clinical practice, improved tests, and the introduction or cessation of PH initiatives. EHRs may capture different clinical concepts or use different representations. Longitudinal analysis of EHRs aims to identify patterns in health and healthcare over time to inform the design of interventions. The analysis predicated on the ability to robustly identify specific clinical concepts.

A set of policies determine recording by clinicians. These policies define a set of quality indicators. Updates are provided every 3 months, and need to be taken into account, and the changes need to be recorded.

Dynamism in presence and representation of clinical concepts in policies needs to be integrated into the longitudinal analysis of EHRs. This will improve accuracy with which patients, interventions and outcomes can be characterised over time.
INSPIRE: An Ontological Approach to Augment Careers Guidance
Mirko Michele Dimartino, Vania Dimitrova, Alexandra Poulovassilis

Build an intelligent tool to inspire career paths. They want to build the tool on top of semantic web technologies. A GUI Tool would interface with the user and with a SPARQL endpoint. Other SPARQL endpoints are attached to RDFS data from LinkedIn and L4All (and others). They are joined up and integrated, and then the user queries the system through federated querying. The integration of the data happens with an ontology-based rewriting for integration.

They have one ontology to describe LinkedIn. The user is asked to create a profile, then can explore the next career step or explore a long-term career goal. The user can select time intervals, and the response is matched against those intervals.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!

Categories
Meetings & Conferences Semantics and Ontologies

UKON 2016: Great North Museum

These are my notes for Dan Gordon’s talk at the UK Ontology Network Meeting on 14 April, 2016.

Dan Gordon is the Keeper of Biology at the Great North Museum. He has about one million objects in his collection, ranging from taxidermy to microscope slides. One problem they face at the museum is that there are 52 million records in the museum, and classification of those objects is very challenging.

The Great North Museum
Source: https://twmuseums.org.uk/images/900/7ELR-4087-original.jpg 14 April 2016

In the Biology Collection, in some ways he starts off better than with the other collections, as there are already many classification systems available (e.g. taxonomic classification). Taxonomy is easier for larger animals, and harder for insects and plants which can be either (or both) small and highly diverse.

There are 40,000 plant specimens in his collection. When new research comes in, rather than re-classifying and moving all of the specimens, he leaves them in their existing system (there are loads of obsolete systems!). One example is a lichen, where you have two completely different organisms living symbiotically (algae and fungus) – here you have two completely different phylogenies, and the position in the system is constantly being revised. Therefore the best way for him to organize things is… Alphabetically!

Another example is dynamos. There are many in the collection, and often relate to different academic discipline, and therefore tend to get organized along those lines. In terms of trying to classify them, their historical use is very important. They store some un-ground lenses for lighthouse lamps, and these are very important for the history of lighthouses and industry in general. There is a system for classifying this kind of industrial object which isn’t univerally used, called the SHICS system, which works a bit like the Dewey decimal system.

There are SHIC numbers for wedding dresses, marriages etc. However they have a dress to store which was worn for a wedding during WWI. Therefore its primary importance is relating to WWI, so in theory you could assign many different SHIC numbers to note its different roles. However, many people tend to choose what they believe is the most important role and assign a number just for that. This tends towards an “incomplete” number of axes of importance.

What about the art collection? There’s Flat Art, 3D Art, etc. In the physical store, everything that’s framed is on racks, and unframed things are in drawers – that’s the primary classification 🙂 . After that, title and artist are important classifications. But what about things like pots, where there are designers, makers, and manufacturers? There are several layers to the cataloguing which are not obvious when you initially get starting classifying.

Please note that this post is merely my notes on the presentation. I may have made mistakes: these notes are not guaranteed to be correct. Unless explicitly stated, they represent neither my opinions nor the opinions of my employers. Any errors you can assume to be mine and not the speaker’s. I’m happy to correct any errors you may spot – just let me know!