Selecting Ontology Entailments for Presentation to Users

Artemis Parvizi, Chris Mellish, Kees van Deemter, Yuan Ren, Jeff Z. PanKEOD 2014 · DOI

TL;DR

Presenting the vast number of logical consequences (entailments) of OWL ontologies to users is a significant challenge that current tools largely ignore. This paper argues for a more principled approach to selecting and presenting these entailments, drawing inspiration from linguistics and pragmatics to make ontology development more intuitive and less overwhelming.

What Problem Does This Paper Address?

Developing and maintaining ontologies, especially those using expressive languages like OWL, can lead to an overwhelming number of logical consequences, or "entailments." When a user adds or modifies an axiom in an ontology, a reasoner can deduce many new facts. The problem is that this set of entailments can be infinite, either due to logically equivalent ways of stating facts or because there are infinitely many distinct, yet true, statements that can be derived. Current ontology authoring tools often fail to address this, either by showing a limited, arbitrary subset of entailments or by not presenting them in a way that helps users understand their significance. This can lead to user confusion, misunderstanding of logic, and inefficient ontology development.

Methods

The paper proposes a multi-faceted approach to selecting and presenting ontology entailments, moving beyond purely syntactic methods. The authors categorize potential solutions into four main areas:

  • Syntactic Approaches: These methods select axioms based on their explicit structure, such as focusing on axioms of the form "A is a subclass of B" where A is a named concept. While common in tools like Protégé, they are seen as a starting point rather than a complete solution.
  • Logic-Based Approaches: These methods leverage logical properties like entailment and specificity. For example, they might prioritize more specific entailments or avoid tautologies. The paper also suggests using the derivation path of entailments (e.g., how many reasoning steps are involved) as a selection criterion.
  • Discourse Structure Approaches: This novel approach suggests using the context of the user's interaction and the "discourse" of ontology development to infer relevance. If a user is discussing subclasses of a particular concept, entailments related to those subclasses might be more pertinent.
  • Pragmatic Approaches: This category emphasizes understanding the user's goals and intentions. The authors highlight the use of "Competency Questions" (CQs) – natural language questions an ontology is expected to answer. Entailments that are crucial for answering these CQs, or whose violation would make a CQ unanswerable, are deemed highly relevant.

Datasets

Not discussed in the paper. The paper uses examples from the "pizza ontology" to illustrate its points, but does not present formal experiments or datasets.

Key Insights and Ideas

  • The "Problem of Logical Equivalence" is a Major Hurdle: The paper identifies that simply presenting all logically equivalent forms of an axiom is unhelpful and overwhelming.
  • User Intent is Key: The most promising approaches (discourse and pragmatics) focus on understanding why a user is developing an ontology and what they aim to achieve, rather than just what the logic dictates. Competency Questions are a concrete mechanism for capturing user intent.
  • Linguistics and Philosophy Offer Valuable Tools: The authors note that many useful ideas for entailment selection come from linguistics (e.g., Gricean maxims, presuppositions) and philosophy of language, suggesting that effective ontology tools need to consider the communicative aspect of ontology development.
  • Selection, Ordering, and Grouping are Intertwined: The way entailments are presented (ordered and grouped) significantly impacts their understandability and interacts with the selection process itself.

Key Results

The paper is a position paper and does not present quantitative results from experiments. Its contribution lies in identifying a critical, under-addressed problem in ontology engineering and proposing a framework of principled approaches to tackle it. The "results" are conceptual: a categorization of methods and a strong argument for integrating user-centric, linguistic, and logical principles into ontology authoring tools.

Limitations and Future Work

The authors acknowledge that their ideas are tentative and incomplete, particularly regarding the integration of these approaches into practical systems. They suggest that future work should focus on:

  • Developing more sophisticated discourse-based approaches for ontology authoring environments.
  • Extending pragmatic approaches by more deeply modeling user requirements through competency questions and their associated authoring tests.
  • Indicating to the user why certain entailments were selected or omitted to avoid misunderstandings.
  • Providing mechanisms for users to explore entailments hypothetically, allowing them to understand the consequences of their actions before committing them.
  • Integrating these selection, grouping, and ordering mechanisms more deeply into authoring environments, allowing the environment to make more intelligent decisions rather than relying solely on the reasoner.

Summary updated

BibTeX
@inproceedings{parvizi2014selecting,
  author = {Artemis Parvizi and Chris Mellish and Kees van Deemter and Yuan Ren and Jeff Z. Pan},
  title = {{Selecting Ontology Entailments for Presentation to Users}},
  booktitle = {KEOD},
  year = {2014},
  doi = {10.5220/0005136203820387},
}
Figure