Actions

Ontolog Forum

Purpose

Automated and Machine Learning for Knowledge Extraction, Ontology Development and Enhancement

Organized by Gary Berg-Cross (Ontolog Board Member)

One problem with the wider use of ontologies in such things as data integration and enhancing machine intelligence is that engeering ontologies and related knowledge bases remains time consuming and requires special and still limited abilities (Simperl et al, 2012). This makes creating relevant and reusable quality ontologies difficult. It is therefore natural to look to some help automating or semi-automating ontological engineering with techniques that can discover, extract and formalize knowledge for ontology construction or enhancment. This idea and work has old roots in AI where researchers investigated using inductive inference for machine learning (i.e. the generalizations of class definitions & relations from a set of observed instances). Inductive logic programming that "learned" classes and helped form logical theories was already underway by the mid 1980s. By 2000 the Ontology Learning ECAI-2000 Workshop could note the:

"surge of interest in other fields than ontology engineering that tackle the discovery and automatic creation of complex, muli-relational knowledge structures. For example, the natural language community tries to acquire word semantics from natural language texts, database researchers tackle the problem of schema induction, and people building intelligent information agents research the learning of complex structures from semi-structured input (HTML, XML)."

Ontology learning from text, semi-structured documents as well as Linked Data remains a notable topics in artificial intelligence and related knowledge engineering research area. Within Ontology Learning efforts a hierarchy of 6 types of relevant knowledge are noted (Cimiano 2016):

  • acquisition of the relevant terminology
  • identification of synonym terms/linguistic variants (possible

across languages)

  • formation of concepts
  • hierarchical organization of the concepts (concept hierarchy)
  • learning relations, properties or attributes, together with the

appropriate domain and range

  • definition of arbitrary axioms/rules

The objective of this Ontology Summit 2017 track is to explore work that bridges the realm from symbols to concepts and examine the current status and role of automated and machine learning (ML) to support the development and improvement for ontologies - populating them with instantiations of both concepts and relations, often broadly called ontology learning (Lehmann & Volker, 2014). Machine learning (ML) technology, both symbolic and now non-symbolic, seems to be advancing rapidly as a diverse and interdisciplinary research field. A cited example is the progress of Google Translate to lever observations of patterns in languages. As reported if Google Translate knew English to Korean, and English to Japanese translations, it could actually get "pretty good" initial results translating from Korean to Japanese using a "common ground of English".Efforts such as DL-FOIL and OCEL use generic supervised machine learning approaches for capturing knowledge in description logic form, while onto-relational learning combines methods for learning OWL axioms with rule learning approaches. But how practical is the use of advanced automation and ML to support the development, enrichment and broadening of ontologies and broadly in the Semantic Web? Can we, for example, in the face of big, but noisy data, expect that machine learning will increasingly be employed for building axiom rich ontologies of quality as well as adding light semantics for such related things as metadata annotation? What ML best practices exist to analyze distributed data & information resources ranging for ontology construction using word patterns in unstructured text or domain dependencies in highly structured database? Can we reliably detect meaningful structural patterns in RDF graphs to develop ontologies? Although many ontology ML methods now exist, it seems that much remains to be done. There are, for example, no comprehensive models across the whole ontology engineering process.

This track will examine the status of work, particularly in regard to its strengthens, weaknesses and prospects to support richly axiomatized ontology formalizations at different levels (top-level, domain level, application level etc,) as well as modular ontology patterns.

Machine Learning Methods

ML: Active Learning ML: Bayesian Learning ML: Big Data / Scalability ML: Case-Based Reasoning ML: Classification ML: Clustering ML: Data Mining and Knowledge Discovery ML: Deep Learning/Neural Networks ML: Dimensionality Reduction/Feature Selection ML: Ensemble Methods ML: Evaluation and Analysis (Machine Learning) ML: Evolutionary Computation ML: Feature Construction/Reformulation ML: Graphical Model Learning ML: Kernel Methods ML: Learning Theory ML: Online Learning ML: Preferences/Ranking Learning ML: Recommender Systems ML: Reinforcement Learning ML: Relational/Graph-Based Learning ML: Time-Series/Data Streams ML: Transfer, Adaptation, Multitask Learning ML: Semisupervised Learning ML: Structured Prediction ML: Supervised Learning (Other) ML: Unsupervised Learning (Other) ML: Machine Learning (General/other)

References

Buitelaar, Paul, Philipp Cimiano, and Bernardo Magnini. "Ontology learning from text: An overview." Ontology learning from text: Methods, evaluation and applications 123 (2005): 3-12.

Cimiano, P, Ontology Learning and Population from Text: Algorithms, Evaluation and Applications, Springer, 2006.

Gómez-Pérez, Asunción, and David Manzano-Macho. "An overview of methods and tools for ontology learning from texts." The knowledge engineering review 19.03 (2004): 187-212.

Lehmann, Jens, and Johanna Voelker. "An introduction to ontology learning." Perspectives on ontology learning. AKA/IOS Press, Heidelberg (2014): 9-16.

Maedche, Alexander, and Steffen Staab. "Discovering conceptual relations from text." Ecai. Vol. 321. No. 325. 2000. Maedche, Alexander, and Steffen Staab. "Semi-automatic engineering of ontologies from text." Proceedings of the 12th international conference on software engineering and knowledge engineering. 2000.

Maedche, Alexander. Ontology learning for the semantic web. Vol. 665. Springer Science & Business Media, 2012.

Simperl, Elena , Tobias Buerger, Simon Hangl, Stephan Woelger, and Igor Popov. Ontocom: A reliable cost estimation method for ontology development projects. Web Semantics: Science, Services and Agents on the World Wide Web, 16(0):1 – 16, 2012.

Wong, Wilson, Wei Liu, and Mohammed Bennamoun. "Ontology learning from text: A look back and into the future." ACM Computing Surveys (CSUR) 44.4 (2012): 20.