Learning Formal Definitions for Snomed CT from Text

Ma, Yue; Distel, Felix

doi:10.1007/978-3-642-38326-7_11

Cited by 12 publications

(19 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is complemented by our proposed formalism in [8] for extracting other named relationships (so-called DL roles) between concepts. For example, if Baritosis is the target concept, then from the sentences given in Table 1 the Table 1: Sentence Examples "Baritosis is a benign type of pneumoconiosis, which is caused by long-term exposure to barium dust."…”

Section: Definition Candidatesmentioning

confidence: 99%

“…Intuitively, (7) ensures that a solution must contain S l for at least one literal l from each clause in f . The constraint (8) ensures that a solution cannot contain both Sx i and S¬x i for some i ∈ {1, . .…”

Section: Np-hard Variantsmentioning

confidence: 99%

“…Following [8], we use the distance supervision [10] approach for the definition candidate extraction. This approach is independent of manual annotation of textual data, which would be costly.…”

Section: Learning Definition Candidates From Textsmentioning

confidence: 99%

“…Using DL reasoning we first generate the set of all triples A|R|B that are entailed by SNOMED CT in the following sense: RB = {A|R|B : SNOMED CT |= A ∃R.B}. Reasoning provides a way to use implicit information encoded in SNOMED CT 8 . Because the inferred role base RB contains the relationship Baritosis | Causative_agent | Barium_dust, the sentence is aligned to Causative_agent.…”

Section: Learning Definition Candidates From Textsmentioning

confidence: 99%

“…Different approaches have been proposed for assisting ontology generation. Among them are machine learning and natural language processing based approaches such as [4,18,8]. In most settings, several natural language sentences are available for a single target concept, and one concept description is obtained from each of these sentences.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Concept adjustment for description logics

Distel

2013

Proceedings of the Seventh International Conference on Knowledge Capture

Self Cite

View full text Add to dashboard Cite

There exist a handful of natural language processing and machine learning approaches for extracting Description Logic concept definitions from natural language texts. Typically, for a single target concept several textual sentences are used, from which candidate concept descriptions are obtained. These candidate descriptions may have confidence values associated with them. In a final step, the candidates need to be combined into a single concept, in the easiest case by selecting a relevant subset and taking its conjunction. However, concept descriptions generated in this manner can contain false information, which is harmful when added to a formal knowledge base. In this paper, we claim that this can be improved by considering formal constraints that the target concept needs to satisfy. We first formalize a reasoning problem for the selection of relevant candidates and examine its computational complexity. Then, we show how it can be reduced to SAT, yielding a practical algorithm for its solution. Furthermore, we describe two ways to construct formal constraints, one is automatic and the other interactive. Applying this approach to the SNOMED CT ontology construction scenario, we show that the proposed framework brings a visible benefit for SNOMED CT development.

show abstract

Section: Definition Candidatesmentioning

confidence: 99%

Section: Np-hard Variantsmentioning

confidence: 99%

Section: Learning Definition Candidates From Textsmentioning

confidence: 99%

Section: Learning Definition Candidates From Textsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Concept adjustment for description logics

Distel

2013

Proceedings of the Seventh International Conference on Knowledge Capture

Self Cite

View full text Add to dashboard Cite

show abstract

On the Complexity of Learning Description Logic Ontologies

Ozaki

2020

Reasoning Web. Declarative Artificial Intelligence

View full text Add to dashboard Cite

Ontologies are a popular way of representing domain knowledge, in particular, knowledge in domains related to life sciences. (Semi-) automating the process of building an ontology has attracted researchers from different communities into a field called "Ontology Learning". We provide a formal specification of the exact and the probably approximately correct learning models from computational learning theory. Then, we recall from the literature complexity results for learning lightweight description logic (DL) ontologies in these models. Finally, we highlight other approaches proposed in the literature for learning DL ontologies.

show abstract