Intelligent Database Systems Lab Presenter: CHANG, SHIH-JIE Authors: Kevin Meijer, Flavius Frasincar, Frederik Hogenboom 2014.DSS. A semantic approach for extracting domain taxonomies from text
Intelligent Database Systems Lab Outlines Motivation Objectives Methodology Experiments Conclusions Comments
Intelligent Database Systems Lab Motivation Manually creating a taxonomy is difficult and time consuming process and may not be high quality.
Intelligent Database Systems Lab Objectives This paper presents a framework using a semantic approach for the automatic building of a domain taxonomy, called Automatic Taxonomy Construction from Text (ATCT).
Intelligent Database Systems Lab Methodology
Intelligent Database Systems Lab
Methodology – term filtering lexical cohesion domain pertinence domain consensus domain score of term t in domain corpus Di
Intelligent Database Systems Lab WSD on text corpora WSD on existing taxonomies
Intelligent Database Systems Lab Methodology – Concept hierarchy creation score(pricing | ‘pricing behavior’) = ½*0.4 +1/3*0.3= 0.9 score(trading | ‘pricing behavior’) = ½*0.3 = 0.85
Intelligent Database Systems Lab Implementation WSD result
Intelligent Database Systems Lab Implementation hierarchy creation result
Intelligent Database Systems Lab Experiments semantic precision semantic recall
Intelligent Database Systems Lab Experiments – core taxonomy V.S. reference taxonomy
Intelligent Database Systems Lab Experiments – two measures taxonomic precision taxonomic recall global taxonomic precision global taxonomic recall taxonomic F-measure
Intelligent Database Systems Lab Experiments
Intelligent Database Systems Lab Experiments
Intelligent Database Systems Lab Conclusions –ATCT framework can be successfully applied to other domains than economics and management. –Our approach works well in capturing the broader–narrower relation between concepts.
Intelligent Database Systems Lab Comments Advantages –Define concepts well. Applications –Built taxonomies 、 Term extraction and filtering.