GSLT: Machine Learning

The purpose of this course is to give a solid methodological foundation in machine learning for language technology, an overview of the most widely used approaches to learning, and an in-depth understanding of a subset of these approaches. The course is aimed at students with a basic knowledge of natural language processing and/or speech technology (at least the equivalent of a GSLT level 1 course in one of these areas, see NLP, Speech technology). Basic programming skills are useful as well as a rudimentary knowledge of basic statistics and probability theory.

The course consists of three parts:

The first part of the course (taught in the first intensive week) gives a general introduction to machine learning, covering basic methodological principles and introducing the major learning paradigms used in language technology.
The second part of the course (taught in the first and second intensive week) consists of three advanced tutorials on specific learning methods:
1. Transformation-based learning (week 1)
2. Inductive logic programming (week 2)
3. Memory-based learning (week 2)
Each tutorial will include a lab session with an assignment to be completed after the session.
The third part of the course is a practical project (reported at the closing seminar) applying one or more of the learning methods covered in the course to one or more areas of language technology.

All assignments and projects should be carried out in groups of two students.

The main text book for the course is Mitchell (1997) Machine Learning, which will be used especially in the first part of the course. Students are recommended to read at least chapters 1 and 5 before the first intensive week.

NB: The official language within GSLT is English but we can decide to have lectures, seminars and discussions in Swedish instead, provided of course that all participants are comfortable with this. In any case, participants are free to formulate their contributions to discussions, whether oral or written, in any language that can be understood by the other participants (which in most circumstances means Swedish or English).

Schedule

Lectures/Tutorials

Date	Time	Room	Contents	Teacher
12/9	10-12, 13-15, 15-17	G412 (10-12) H421 (13-15, 15-17)	Introduction	JN
13/9	8-10, 10-12, 13-15	H421	Transformation-based learning	TL
26/10	8-10, 10-12, 13-15	H420	Inductive logic programming	JC
27/10	8-10, 10-12, 13-15	H421	Memory-based learning	WD

Teachers: Joakim Nivre (JN)

Torbjörn Lager (TL)

Walter Daelemans (WD)

James Cussens (JC)

Study Periods

The period between the first and second intensive weeks will be devoted to two tasks:
- Read the text book and do the written assignment.
- Complete the TBL assignment and write a short report.
Deadline: 21 October

The first period after the second intensive week will be devoted to three tasks:
- Complete the MBL assignment and write a short report.
- Complete the ILP assignment and write a short report.
- Write a proposal for the final project (maximum 1 page).
Deadline: 25 November

The second period after the second intensive week will be devoted to the course project, which is to be reported in a term paper.
Deadline: 21 December.

Closing Seminar

The closing seminar will take place at Växjö University 12-13 January 2006.

Teachers:	Joakim Nivre (JN)
	Torbjörn Lager (TL)
	Walter Daelemans (WD)
	James Cussens (JC)