Semester page for INF4820 - Autumn 2016
As we speak, we are finalizing the list of students qualified for the exam. To get this communicated to the faculty administration, we actually have to record preliminary feedback in Devilry; for reasons of time, we are just publishing the minimum number of points you will receive, which will determine whether or not you will be registered for the exam. More complete feedback (and possibly a higher number of total points) will be published as quickly as possible, certainly before the exam date next Friday.
To reward everyone who has worked with us through the semester, we include the points from the short quiz we ran last week against the total sum for the third obligatory exercise, and we lower the passing threshold to a minimum of ten points. In total, all but one of the students who submitted for problem set (3b) have qualified for the exam. Congratulations! In absolute numbers, we receiv...
In the lecture today, we will complete our introduction to generalized chart parsing and Viterbi decoding over a packed parse forest. To consolidate our newly acquired knowledge (and offer some bonus points), we hope to squeeze in one final quiz for this term.
Because we are a little behind in our lecture schedule, we are extending the submission deadline for our final obligatory problem set to the end of the day (midnight) on Thusday this week (November 17). For the theoretical parts of question (4), we are missing some background on the inner workings of the generalized chart parsing framework; and even though it should be possible to work things out from just the code (with its generous comments), this part of the assignment will (hopefully) be much clearer after our final lecture on chart parsing this Thursday.
With a delay of about 24 hours, we have just posted the second (and final) part of our third obligatory exercise, which has its submission date a little less than two weeks from today (November 16). The first part of this problem set builds on the theory that was the topic of the most recent lecture, while remaining theoretical background for the 'middle' part of this problem set will be lectured next week.
This exercise also offers an optional third part, which calls for combination of our chart parser implementation with the HMM tagger and implementation of additional optimizations; submissions to this optional part will be rewarded with up to ten bonus points-so it should really be possible for everyone who starts work on this assignment about now to qualify for the exam, if you just work hard enough on it. Besides offering bonus points, working on t...
As results for exercise (2b) are being posted these days, a few students are just a point or two short of the minimum threshold of 12 points (in sum) that we require to pass each obligatory assignment. To make it possible to make up for such minor deficits, we offer a 'debt' scheme, but everyone who wants to take advantage of that should think hard about it, please. In a nutshell, we will allow students to 'borrow' up to two points from the final assignment (3). In other words, someone whose sum for (2a) and (2b), and possibly the bonus points for the half-way evaluation, was for example ten points can continue to submit for exercises (3a) and (3b), but they will need a minimum sum of fourteen points for assignment (3) then, to qualify for the exam. Exercises (3a) and (3b) will not be 'easier' or less implementation-heavy than the two parts of assignment (2), however; and we will need to apply the final minimum point threshold rigidly, so everyone who...
We have just published the first half of the final oblgatory exercise, which will address the theory and implementation of Hidden Markov Models. We will introduce the required background in the next two lectures and urge everyone to start reading up on the recommend sections from Jurafsky & Martin (2008).
We cannot quite publish a model solution for (2b) yet, since our schedule has been delayed because of dagen@ifi, and a few students had their submission deadlines extended for (2b) because of illness. Thus, we will use the laboratory session tomorrow (Monday, October 17) to review a few common mistakes in Common Lisp coding and to look through (parts of )the the final exams from previous years.
1- Starting from tomorrow, and for the rest of the semester, we will run one lab session only in room Sed.
2- The lecture this week (October 13) is canceled, as it happens to be on the same day as Dagen@Ifi.
- There will be only one laboratory session tomorrow, in room Sed (i.e. there will be noone in Module), where Elena will go through the solution for Exercise 2a and provide assistance for working with Exercise 2b.
- We haven't published the model solution for Exercise 2a yet due to deadline extensions related to illness. We have therefore pushed back the deadline for Exercise 2b to October 16 (noon 12:00). We will publish the model solution for Exercise 2a tomorrow.
We have published Exercise (2b); the submission deadline for this assignment will be at noon (12:00, in the middle of the day) on Wednesday, October 12. Good luck!
There will only be one laboratory session today, in room Module (i.e. there will be noone in Sed), where Elena will use the first half (or so) of the session to review our ‘model’ solution for Exercise (1) and then provide assistance for working with Exercise (2a).