ter Horst, Hendrik Roman: Information extraction from text for deep domain knowledge graph population. Extracting pre-clinical outcomes in the domain of spinal cord injury. 2021
Inhalt
- Abstract
- Acknowledgements
- 1 Introduction
- 2 Foundations
- 2.1 Knowledge Representation
- 2.1.1 Knowledge Graphs
- 2.1.2 Resource Description Framework
- 2.1.3 Web Ontology Language
- 2.1.4 SPARQL Protocol And RDF Query Language
- 2.2 Conditional Random Fields
- 3 Related Work
- 3.1 Historical Situation
- 3.2 Related Information Extraction Problems
- 3.2.1 Entity Recognition and Linking
- 3.2.2 Relation Extraction
- 3.2.3 Slot-Filling
- 3.2.4 Co-Reference Resolution
- 3.3 Knowledge Graph Population in the Medical Domain
- 4 Application Domain: Spinal Cord Injury
- 5 Model-Complete Text Comprehension
- 5.1 Conditional Random Fields and Factor Graphs
- 5.2 Inference and Parameter Estimation
- 5.3 Sampling from the State Space
- 5.4 Feature Engineering
- 5.5 Entity and Literal Annotation
- 6 Deep Domain Knowledge Graph Population
- 6.1 Ontology-Specific Problem Modelling
- 6.2 Special Case: Experimental Group
- 6.3 Special Case: Result
- 7 Experiments and Evaluation
- 7.1 Evaluation Metrics and Experimental Settings
- 7.2 Experimental Results and Error Analyses
- 7.2.1 Organism Model
- 7.2.2 Injury Device
- 7.2.3 Injury Location
- 7.2.4 Delivery Method
- 7.2.5 Anaesthetic
- 7.2.6 Injury
- 7.2.7 Treatment
- 7.2.8 Experimental Group
- 7.2.9 Trend
- 7.2.10 Investigation Method
- 7.2.11 Result
- 7.3 Discussion
- 8 Applications
- 8.1 Annotating Complex Relational Data with SANTO
- 8.2 System Application: Populating a Knowledge Graph
- 8.3 Exploration of Knowledge with SCIExplorer
- 8.4 Answering Competency Questions
- 8.5 Automated Grading
- 9 Conclusion
- List of Figures
- List of Tables
- Abbreviations
- A Group Name Recognition Expressions
- B Regular Expressions for Literal Extraction
- Bibliography