|
Courses and Workshops
|
| Title: | Symbolic learning of natural language |
| Lecturer(s): | Dimitar Kazakov (University of York) and James Cussens (University of York) |
| Type: | Introductory Course |
| Section: | Language and Computation |
| Week: | Second | | Time: | 11.00-12.30 (Slot 2) |
| Webpage: | http://www-users.cs.york.ac.uk/~kazakov
|
| Reader: | Download | | Room: | A3 |
Description Machine Learning (ML) of language studies the acquisition of linguistic
knowledge from examples. While statistical learning methods are widely
spread nowadays, they have their limitations. The need for adequate
modelling of semantics, and acquisition of human-comprehensible theories
has led to an increased interest in symbolic learning.
The course demonstrates how Symbolic ML (SML) is used across the whole
range of natural language processing (NLP) tasks. Word morphology
learning is chosen to introduce a number of SML techniques and compare
supervised and unsupervised learning. Of these, Transformation-Based
Learning is further focussed on and its use for tagging and bracketing
discussed. Tagging is also used to illustrate the principles of
Inductive Logic Programming (ILP). Detailed discussion of ILP learning
of syntax, lexical semantics and ontologies follows. Hybrid SML is the
last topic covered.
Prerequisites: good practical knowledge of logic programming and
familiarity with the basics of NLP, but no knowledge of ML.
/Further particulars/
The first lecturer has taught an undergraduate (3rd, final-year) course
with the same title and content at York University for 2 years.
Both lecturers have an active research record in the area, and have
given a number of AI-related international summer school tutorials and
courses in the past (ESSLLI-2000, ACAI-01, ECAI-02).
|
© 2003, 2004 LORIA, Carlos Areces. |
2004-09-07
| |
