Classifying Entities into an Incomplete Ontology Bhavana Dalvi

Classifying Entities into an Incomplete Ontology Bhavana Dalvi , William W. Cohen , Jamie Callan School Of Computer Science, Carnegie Mellon University Experimental Results Motivation q Existing Techniques Ontology 1 v Semi-supervised Hierarchical Classification: Carlson WSDM’ 10 v Extending knowledge bases: Finding new relations or attributes of existing concepts Mohamed et al. EMNLP’ 11, Reisinger et al. ACL’ 09 v Unsupervised Ontology Discovery: Adams et al. NIPS’ 10, Blei et al. JACM’ 10, Reisinger et al. ACL’ 09 Ontology 2 q Evolving Web-scale datasets v Billions of entities and hundreds of thousands of concepts v Difficult to create a complete ontology v Hierarchical classification of entities into incomplete ontologies is needed v Our technique is in between Semi-supervised Hierarchical Classification and Unsupervised Ontology Discovery v Class hierarchy is a tree. Assumptions v Classes at any one level are mutually exclusive. Hierarchical DAC Exploratory EM Entity Features Pittsburgh Washington DC Spinach Lives in _ : 200 , City of _ : 156, was born in _ : 250 City of _ : 230, was born in _ : 150, _ is capital of : 50 _ is a green vegetable : 150, _ contains iron : 100 Example of features extracted from Clueweb 09 Locatio n 0. 1 State Adds to class constraints Food 0. 9 Condime nt Vegetabl e Countr y Example of DAC Exploratory EM Coke 1. 0 Root 0. 55 C 8 0. 45 A Dataset #Classe #Levels #NELL s Entities DS-1 DS-2 Methodology Details 11 39 3 4 #Contexts #Entity, #NELL Context Entity Pairs Labels 2. 5 K 3. 4 M 8. 8 M 6. 7 K 12. 9 K 6. 7 M 25. 8 M 42. 2 K Datase #Train Leve t /Test l Point s #Seed/ Macro-averaged Seed Class F 1 #Ideal FLAT DAC Class Semisup Explore. E es EM M DS-1 2 2/3 43. 2 78. 7 * 69. 5 77. 2 * 3 4/7 34. 4 42. 6 * 31. 3 44. 4 * 3. 9/4 64. 3 53. 4 * 65. 4 68. 9 * 9. 4/24 31. 3 33. 7 * 34. 9 41. 7 * 2. 4/10 27. 5 38. 9 * 43. 2 42. 4 * DS-2 335/ 2. 2 K 1. 5 K/1 2 1. 4 K 3 4 Conclusions Acknowledgements : This work is supported by Google and the Intelligence Advanced Research Projects Activity (IARPA) via Air Force Research Laboratory (AFRL) contract number FA 8650 -10 -C-7058.