University of Tsukuba | Grad. Scho. Syst. and Info. Eng. | Dept. Comp. Sci. | List of Courses
Mikio Yamamoto
E-Mail myamaAtcsDottsukubaDotacDotjp
Office hours SB908, 11:00-12:00, Monday
Cource# 01CH603, 01CJ223
Area Intelligent Systems
Basic/Advanced 専門科目
Course style lecture (in Japanese)
Term Fall A,B
Period Tue5,6
Room# SB0110
Keywords Language models, Ngram models, Smoothing, Backoff-smoothing, Interpolation.
Prerequisites Elemental level of probability, statistics and information theory. Programming skill is needed for a final project.
Outline This course will introduce students to several modern techniques for generative models of natural human language such as Japanese. In particular, we will focus on methods for estimating probabilistic models of languages.
Course plan 1.Statistical properties of natural languages:
review of probability and statistics,
units and statistical measures for languages.

2.Probabilistic language models:
N-gram models, smoothing, frequency discountings,
measures for language models.
Textbook pdf files on the web.
References (1)Kenji Kita, "Kakurituteki-gengo-moderu", 1999, (in Japanese).
Evaluation Dependent on year, I impose writing examination and/or submitting a project report. The evaluation will be made by writing examination (and/or a project report) and degree of attening the lecture.
Misc. Open in an odd number year.
Identical to 01CJ223.
2015年度まで開講された「自然言語処理特論」(01CH603, 01CJ223)の単位を修得した者の履修は認めない。