CPSC 477 Lecture 7: nlp_3.8
24 May 2018
School
Department
Course
Professor
Named entity recognition
Segmentation vs classification
Ex: Donald Trump went to Washington
- Named entities: Donald Trump, Washington
o And then classify
▪ SVM
Classification
- Use gazetteers, large database of names
- Gale
Relation extraction
Unsupervised, find similar things, get the same later
Semi-supervised learning
- A little bit of training data, but a lot of unlabeled data
- Eg: build classifier for named entities
- Have sentences that contain these names
o Pierre Vinken
▪ Bill Gates is the executive whatever of whatever
▪ Look for known patterns
• Person and the company are connected with “is CEO of”
o Train system to recognize this middle point, learn pattern,
connect the two words
ACE
- Build taxonomy of relations
Is-A relations
