Grading of the final exam is complete (Herman Ruge Jervell served as the second reviewer). Results might be visible already, or should show up in the next few days. In total, everyone who had qualified for the exam succeeded (there were two As, five Bs, four Cs and Ds each, and one E). Godsommer!
The previous post about how obligatory assignments are registered in the central university system left some room for confusion; i am sorry about that! The minimum requirement of 200 points from the four assignments is unchanged, and only those who accumulated at least as many points qualify for the exam. But the central registration of obligatory assignments only allows a binary distinction (pass or fail) per assignment, hence we had to map the actual results into this scheme, and just looking at the central registry, the 200-point requirement is no longer visible (but still valid). For those who want to re-use partial credit on assignments in the future, the actual results are recorded internally. So, someone taking this class a year from now could opt to have the exact number of points credited, per assignment, for those assignments that they actually delivered.
The results of obligatory assignments have been registered with the university administration; in total, 16 (of 19 who started the class) qualified for the exam, including everyone who submitted for the last assignment. Mapping our point system into four binary decisions in the official system was a bit tricky: any submission that scored more than 50 points is registered as successful, and in total anyone who had at least two successful submissions can take the exam. Numerically, this makes the requirements look a little easier than what we actually asked (a minimum of 200 points in total), and should anyone take the class in the future, wanting to re-use credit from earlier submissions, they would be credited the actual number of points that they scored (so, historically, there would be justice). André will email individual feedback later today, and I am about to post the model solution and slides copies from the MT lecture, even though I did not get to complete the slide s...
André offers an extra, optional laboratory consultation tomorrow, Friday, May 23. Between (roughly) 12:00 and 13:00, he will be in the IT-SLP laboratory (known as 100m-skogen) on the third floor of Forskningsparken.
We have posted a model solution for the past assignment, and the fourth (and final) obligatory assignment today. For you convenience, the command solution6 on the IFI Linux environment will create a new sub-directory for you, containing the files of the model solution. In the lecture tomorrow, we will spend some time reflecting on agentive nominalization, and then move on to aspects of semantics and meaning composition.
Remember that the deadline for electronic submission of parts (1) through (5) of the current exercise is tomorrow, Tuesday, May 7, at 9:00 in the morning. We will publish a model solution sometime mid-day tomorrow, hence cannot accept late submissions. For the theoretical parts (6) and (7), on the other hand, feel free to hand in non-electronic, pencil-and-paper solutions. Please give your results for these parts to André and me, but no later than during the next lecture, i.e. Thursday this week.
To make it easier for you to retrieve and study our model solution, we have created an additional command on the IFI Linux environment: running solution5 from the shell prompt will create a new sub-directory by that same name, containing our model solution (there is a script file there, to load into the LKB).
The third obligatory (and sixth, in total) assignment is now available. The starting package for this assignment will be provided on the IFI Linux environment after the lecture today. The starting grammar for this exercise, essentially, is the model solution to the previous assignment. However, we have made a few more adjustments in the new starting package, so please also take a look at our separate (and pure) model solution for the second obligatory assingment. We will talk about our analysis of modification during the lecture today; after that will be an excellent time for you to compare our solution to yours. We plan to email comments on your submissions prior to the next laboratory session, i.e. probably late on Monday this coming week.
The moderately revised version of the slides on typed feature structures is now available, as used in the lecture yesterday. I deleted one slide (the extended version of InteractionofLexiconandPhraseStructureSchema) and added one (on the introduction of appropriate features in the hierarchy).
There was an error in the assignment of points to the second obligatory assignment: our original version did not allocate any points (but merely useless pointes) to part (4). We have corrected this error by mildly re-assigning points from other parts of the assignment (the total remains 100 points). Also, a sample exam is now available; please note that there are slightly more questions in this example than we would expect to ask in the actual exam. Also, observe that for most of the questions we really want quite short answers: when we say inasentenceortwo, we really mean it!
Right before easter, we asked you to answer a five-minute questionnaire on the course experience so far (which is a standard component of quality assurance for teaching at IFI). The results were encouraging. Nevertheless, there will be a few adjustments to course organization for the rest of the semester.
The slide copies from the lecture today are available on the course page; please take some time to (a) read through the obligatory background reading (Chapter 3 of Sag, Wasow, and Bender, 2003; and soon Chapter 4 also), (b) compare the slide copies and starting grammar for the current exercise to the book (in same cases, we have further simplified things), and (c) aim to complete parts (1) through (at least) (5) of the current assignment prior to the laboratory next week. Remember that (on the IFI Linux machines) the shell command lkbdoc allows you to consult on-line documentation for the LKB software, i.e. essentially parts of the Copestake (2002) book. Please familiarize yourself with this documentation.
The second obligatory (and fifth, in total) assignment is now available. In the laboratory tomorrow, we will start working on this assignment, and the lecture later this week will provide additional background.
The model solution for the first obligatory assignment is now available, and André is in the process of emailing feedback. The meeting on Thursday this week is devoted to an in-depth review of this assignment. Please compare our solution to yours beforehand.
We have completed about half of the class (and the first obligatory assignment), and we would like to ask for your feedback on things so far. See some general information about half-way evaluation, and please take a look at the questionnaire we use for this class. Those of you who were not in class, you can (and should) still get your comments to us. Either hand your filled-in questionnaire to André when you run into him, drop it into the mailbox for Stephan (in the LNS printer room on the fourth floor of Forskningsparken), or bring it to the next lecture. Our next meeting is on Thursday, March 27. Godpåske!
The laboratory today is kind of canceled. Stephan is staying home with a cold, but André will be there to work with people individually (on the Lisp environment and the last assignment), for those who want it. No new exercise this week. The lecture on Thursday this week will be given as usual; please go through the assigned reading from Jurafsky & Martin (2008).
Remember that the deadline for submitting your work on the current assignment is Monday morning at 9:00. We cannot accept late submissions, i.e. no points will be awarded for work on this assignment unless you get it to us before the deadline. Please put your answers to the theoretical questions into a file README (or similar) in the same folder as your Lisp code, pack up the entire directory as a tar(1) or zip(1) archive, and email it to both André and me. On the IFI Linux environment, you can have all of this done for you by typing submitexercise4 at the shell prompt (assuming you are in the directory containing exercise4 as a sub-directory, typically your home directory). It is perfectly reasonable to submit more than once, say to make sure you send a snapshot of your results early on, but then re-send everything again later (but before the Monday morning deadline), after you have worked on your co...
We have found no less than three (rather) minor glitches in the second part of the current assignment. First, the body of fundamental-rule() returns whatever agenda-push() returns; contrary to what we suggest in our calling example for fundamental-rule() on the assignment sheet, that return value will be the complete current agenda, i.e. a list whose first element should be very similar to what we show. Second, towards the end of part (B2), we fail to say the obivous: test your implementation of the generalized chart parser on a number of examples and compare to the results you get from the built-in LKB parser; the function to call for our own generalized parser is, predictably, parse(). Third, in part (B3) we have opted for a simplified notation of trees as lists: (S (NP kim) ...). This is different from the return value of our own function edge-to-tree(), because the leaf nodes are not wrapped into additional singleton lists. Thus, what we show on the assignmnet shee...
The second part of the first obligatory assignment is now available. This coming Tuesday, we will have a (really) brief lecture, consoloditing our understanding of the generalized chart parser and its data structures; the bulk of the laboratory session, however, we will use to work with you individually on the current exercise.
We have made a few changes to the LKB and Lisp installation on the IFI Linux cluster; excessive warnings about missing fonts should be reduced now, and the compiler settings have been changed to yield code that executes a little slower but should be easier to debug, including giving more meaninful error messages. Please let us know immediately (in email to both instructors) in case you notice anything surprising.
We have (finally) posted model solutions for the second and third exercise; see the detailed course schedule. Please compare our solutions (and comments on the code) with your own, and feel free to ask for clarification or to point out to us where we could further improve these examples.
Also, note that the first bug in our fourth assignmnent has been found: one of the skeleton functions that we provide in chart.lsp should have three arguments: (defun chart-adjoin (from to object) ???)
The machine(s) under the name login.ifi.uio.no were switched over to a different (64-bit) version of Linux this weekend. The effect was that our course environmnet failed to start up the LKB today. This should now be fixed again; if you continue to encounter problems with launching our LKB and Lisp environment, please email us.
The (first part of the) first obligatory assignment is now available, and the starting package will appear magically in your home directory, once you log into the IFI Linux environment. Please read the instructions carefully and see whether you can get a heads start prior to our next laboratory. This coming Tuesday, we will have a brief lecture, introducing the basics of chart parsing and the CKY algorithm, which we will discuss in more detail in the following lecture.
Regrettably, there was a bug in the file parse.lsp that we provided as part of the current exercise. Our file actually contained part of the solution to part (4) of the exercise, viz. the recursive rules VP --> VP PP and NP --> NP PP. As some of you have discovered, rules of this type make our current parser go into an infinite loop. It is good to remember that C-c C-c (pressing the Control and C keys jointly, twice in a row) interrupts the Lisp system. To correct our mistake, either delete these rules from your current version, or retrieve a corrected version of our file
Several of you have asked about the first obligatory assignment (on chart parsing with a simple context-free grammar). We will post the assignment (on this web page) on Friday this week, i.e. after the next lecture. You will then have until Wednesday, March 5, to submit your results (more on that on the assignment), and we will devote the two laboratory meetings between now and March 5 to this exercise.