We explained how from the relational perspective, rule
application corresponds to whether the left or right contexts of a
rule need to match on the input or output string.
We discussed 3.3.3. the Double Plurals in Breton in depth.
31 Oct 2018
Jon led discussion on chapter 3 where we considered Roark and Sproat's arguments that lexical/incremental and realizational/inferential theories of morphology are refactorizations of each other.
We studied the formal derivation of the equivalence for English past tense [d] in 3.4.
We questioned whether this kind of derivation could really be extended when there is multiple exponence.
29 Oct 2018
TBR
24 Oct 2018
The updated python file for English plurals: [EngPlurals.py].
Here is the output from the Jupyter qt3 session ipython2018-10-24.html Unfortunately, it looks like the beginning of the session was cut off.
Here
is installation.pdf, which
explains how to set up the CompLingLab's Virtual Machine, and then
install openfst and pynini onto it.
For next Wednesday, your goal is to install this software
successfully.
01 Oct 2018
We discussed pages 29-35 in RS2007 where the concept of morpheme as a transformation/regular relation is shown to handle prosodic circumscription and prosodic governance.
On Wednesday, the following students will present the material from the following subsections: 2.2.4 (Andrija), 2.2.5 (Ji Yea), 2.2.6 (Alex), 2.2.7 (Kalina), 2.2.8 (Jon).
Ali and Andrija volunteered to present 2.2.10, and we are still looking for volunteers for 2.2.9, 2.3, and 2.4.
Install Oracle Virtual Box on your computer if you don't have any VMware installed already.
26 Sep 2018
Ali and Ayla presented
[Karttunen1993.pdf]
on intermediate representations and 2-level morpho-phonology.
We discussed the different ways a complex regular relation can be factored into simpler ones.
24 Sep 2018
We reviewed the definition of composition carefully.
We saw how the composition of a lexical item as a regular
relation with a morpheme as a regular relation results in regular
relation representing the form of the original item subject to the
morphological process.
19 Sep 2018
We discussed pages 27-29 in RS2007 where the notion of morpheme as a transformation/regular relation is introduced.
For next Monday: consider the regular relation L = {(kat,kat)} (L is for lexicon) and the regular relation P = [ID][e:z] (where P is for Plural, [ID] is the identity relation, e is epsilon, and z is the phoneme 'z'). Following Lothaire's definition carefully, compute both P o L and L o P. [kaet.pdf]
We had some discussion regarding questions from the readings.
We explained how product constructions for regular langugae intersection and union as well as regular transduction composition. [cm03.pdf]
For next class: write out the product construction which gives the composition of the two literal transducers A and B in the handout at the bottom of page 2.
10 Sep 2018
Tony presented slides showing examples of the use of regular expressions in every day programming.
We reviewed the automata product which computes intersection.
We introduced the idea behind transducer composition.
Read for next class: pages 40-44 on transducer composition from Lothaire, and pages 47-50 from Sipser on non-determinism.
Today we went over regular expressions (REs), generalized regular expressions (GREs), star-free expressions (SFE), and "cat-union" expressions. [cm02.pdf]
We introduced epsilon transitions, which have an epsilon on the input side, and discussed how they also introduce non-determinism.
We mentioned the following theorems.
Kleene's theorem: The formal languages (sets of strings)
describable with regular expressions are exactly the ones
recognizable by finite-state acceptors.
The formal languages recognizable with non-deterministic
acceptors are exactly the ones recognizable with determinisitc
finite-state acceptors.
With the epsilon transitions, we explained how finite-state
acceptors are closed under union, concatenation, and Kleene
star.
Assignment 2, due Monday September 10
read page 13 of Roark and Sproat and study the definition of acceptor product.
Do the following:
Write down two acceptors. (For example: one which recognizes all strings without 2 consecutive "a"s and one which recognizes all strings with an odd number of "b"s.) Each acceptor should have at least two states.
Calculate the acceptor product using the definition.
We will go over some of your work in class and I will collect it.
Here is an example answer that would receive full credit.
Re-read the assigned pages from last time and come to class prepared to ask questions.
27 Aug 2018
First day of class!
Today we went over how weighted transducers map strings to values [cm01.pdf].
Assignment 1, due Wednesday:
read the brief entry by Stephen Anderson on morphology
read pages 1-12 and pages 23-27 of Roark and Sproat. It is NOT important to understand EVERYTHING! Instead, try to recognize in pages 1-12 some of the things we talked about in class today. Yes, we will discuss it on Wednesday
do the four sets of four problems each in the today's handout. These are on pages 2 and 3 subtitled Boolean, Counting, Probability, and String Transducers. Draw out the paths!!