Week 5: Welcome and checklist
Welcome to Week 5! We are now switching lecturers again in the course, and I (Shay) expect to teach you in the next three weeks. We continue to see an active use of Piazza, which is excellent.
Overview of this week
This week, we focus on the topic of part-of-speech tagging. Parts of speech refer to categories such as "noun", "verb", and "adjective". There are various ways to define such categories; we will see some of them in the lectures. We will also touch on algorithms for actually performing part-of-speech tagging -- taking an input sentence and returning the part of speech of each word. If you want to get a head start, try to think why such a problem is non-trivial to begin with (why not just take all occurrences of "dog" with "noun"?)
For POS tagging, we are looking at Hidden Markov Models, which combine some of the ideas from n-gram models (sequence modelling) and those of the Naive Bayes model (latent or hidden variables). We'll also see our second example of a dynamic programming algorithm, and we are inching our way towards models of syntax, which we will get to for real next week.
Week 5 materials
Slides:
- Parts of speech (Lecture 1 - [pdf])
- A model for POS tagging (Lecture 2 - [pdf]; note that slides 31-38 are for enrichment only, and are not examinable)
- Inference of POS tags (Lecture 3 - [pdf]; Viterbi intuition [pdf]; HMM Viterbi spreadsheet [xlsx])
Readings (reminder: * indicates "essential"):
- JM3 8.0-8.4.3 (*), 8.7
- JM3 8.4.4-8.4.6 (*), JM3 Appendix A.2-A.5
Additional materials:
- Week 4 tutorial solutions [pdf]
- Jason Eisner's ice cream consumption spreadsheet for the HMM Baum-Welch algorithm
Week 5 checklist:
- Throughout: Continue working on the assignment, which is due this Wednesday. In addition, follow the quizzes on Gradescope for each lecture.
- Tutorial: Go over the answers for the tutorial that we provide, and check to what extent you understand them.