Topic: Identify new relations from KBs.
keyword: Statistical Relational Learning
Engineering the problem:
- Constructing onto for SemWeb is a time consuming task, therefore automating the process is a demand .
- As the domain of a system scales, new data will be introduced; thus identifying and inferring (and/or removing noise of) the new relations need to be performed systematically.
- Machine learning can be applied in this case (learn from previous data and apply on new ones).
- Statistical Relational Learning “a field of ML” deals with relational structures, and typically in KR and reasoning.
- Probabilistic soft logic is a new framework ‘language’ (based on SRL) for modeling probabilistic and relational domains. PSL is applicable for ML tasks such as link prediction and ontology alignment.
- Machine Learning and Ontology Engineering
- Ontology learning
- Probabilistic Soft Logic
- A Probabilistic-Logical Framework for Ontology Matching
- A Short Introduction to Probabilistic Soft Logic
- To visualize relations and data (ontology) structures, use Ayasid SW
- PSL for for Semantic Textual Similarity (2014)
- Folksonomy used in matching for annotated classification, see wiki
( see: Semantic Matching thesis implemented in python)
- Unsupervised Learning of Semantics Relations between concepts pdf
PSL vs. Fuzzy logic:
- PSL works with continuous variables, while
- Fuzzy logic works with discrete values.