Linguistics 439/539
Statistical Natural Language Processing

Handouts:

• Syllabus
• Survey
• Assignment #1
• Assignment #2
• N-gram approximation handout
• Revised Assignment #3
• HMM N-gram handout
• Assignment #4
• Assignment #5
• Forward probabilities for a PCFG
• Sample from Welsh-English corpus
• Assignment #6
• Sample alignment for Welsh-English corpus
• Final assignment

Matlab and text files:

• Simple PCFG weights
• Efficiency demo
• Austen text
• Russian text
• The+N sequences in Brown
• Two-dice simulation
• Binomial distribution
• Entropy of a die with up to 8 sides
• Entropy of an unfair die with 4 sides
• Distribution of word lengths in Brown
• Mutual information
• Zipfian distribution in Brown
• Special log2 function
• N-gram approximation
• Laplace and Lidstone demo
• Unseen unigram demo
• Comparing Laplace and Lidstone for Austen
• Comparing MLE and Lidstone for Austen
• Second Austen text
• Comparing smoothing techniques for one sentence in second Austen text
• Deleted estimation demo
• One answer to #3 on HW#2
• Translation of Good-Turing
• Demo of regression for Good-Turing
• Applying Good-Turing to Austen
• Multiple tags in Brown
• Dumb tagger for Brown
• One answer to #2 on HW#3
• Forward probabilities
• HMM visualization (Kludgy, platform-dependent, and Graphviz must be installed.
• One answer to #1 on HW#4
• One (partial) answer to #3 on HW#4
• One answer to #2 on HW#5
• Quick and dirty xml parser for Welsh National Assembly data
• Code for setting different colors
• K-means demo
• EM demo
• One answer to #1 on HW#6
• Vector space demo
• Poisson distributions
• Poisson demo
• two-Poisson distribution demo
• k-mix demo
• least-squares demo
• Binary independence model demo

Links:

• Corpora and readings (for enrolled students only)
• Textbook website
• Matlab site license page at UA (free for UA folks)
• Matlab website
• Octave (open-source free alternative to Matlab)
• Graphviz (Graph visualization software, for HMMs)
• Various Matlab files from 478/578 (Spring 2013)
• Various Matlab files from 408/508 (Fall 2013)