Computer Science / BCS 530

Methods in Data-Enabled Research

Fall 2015

Syllabus

8/31	Intro
9/2	Lab
9/9	Power Laws	Clauset et al., 2007	HW 1 due
9/14	Lab
9/16	Kneser Ney	Chen and Goodman, 1998	HW 2 due
9/21	Lab
9/23	Processing Efficiency	Gildea and Jaeger	HW 3 due
9/28	IBM Model 1	Brown et al., 1993
9/30	Lab
10/5			HW 4 due

Homework 1, Due 9/8 10am

Plot rank vs. frequency on a log-log scale, and fit parameters for the generalized power law.

Data are in: /u/cs248/data/mt/training.eng on the instructional network. Note: on grad network, the same files are available in /u/cs448.

Homework 2, Due 9/16 10am

Use the Kolmogorov-Smirnov statistic to fit x_min and alpha for the data from the previous assignment. Discuss how the values compare with the results of HW 1, and plot both models against the data.

Homework 3, Due 9/23 10am

Implement a Kneser Ney bigram language model, training on the file training.eng and testing on test.eng. See test files kn.train, kn.test, kn.out.

Homework 4, Due 10/5 10am

Implement IBM Model 1, and plot log probability of training and test files. See test files train.eng, train.fra, test.eng, test.fra.

To turn in: /u/cs530/TURN_IN dir_name

gildea @ cs rochester edu

September 30, 2015