Assignment 9: Construction Grammar Learning
Assigned Tuesday, April 24th
Due Tuesday, May 1st, in class.
This assignment involves two parts. The first is to get you familiar with
the sentence analysis process, and the second to understand the grammar learning
algorithm. You will be using a construction grammar analyzer to analyze sentences
using a simple grammar written in ECG (Embodied Construction Grammar). You will
then be asked to predict the result of grammar learning by examining sets of
input tokens.
Please read O8, O9,
Book Ch 25
before beginning this assignment.
java -jar ECGLnew.jar
jar xf ECGL.jar
java -classpath classes ecgs.ECGL -s
The older version of the gui is much more of a work in progress, so a lot of the menus aren't
yet fully functional. There are a few bugs as well. So: please follow the directions closely.
You can feel free to poke around, of course, but be patient with non-functional
or buggy stuff if you do.The CHART pane will then generate a table which displays the parsing of the sentence according to the known constructions. So, for example, bring is labeled with the "Bring" construction (Bring-Cn1). Notice that the analyzer only understands some of the words; "will" and "the" are not labeled because it does not know these words (i.e., they are not in the grammar). Sometimes the grammar will also find constructions which it doesn't use in its final parsing; these will show up in the CHART but will be left unshaded.
Note also that the CHART is hierarchical: lower level (level 0, in fact: see the lefthand column) constructions have been found for the individual words I, bring, and cup; and a sentence-long Transitive construction has been found to span the aforementioned ones (level 2--don't worry about the numbering for now but, roughly, higher numbers correspond to constructions which span more words). In finding the spanning construction, the analyzer skipped the words it didn't recognize--hence the "Skipped: 1 3" in the Transitive-Cn1 box. This indicates that the first and third words have been skipped (I is the zeroth word). You can change the number of words the analyzer will skip by pulling down the Settings > Set learning parameters menu and adjusting the appropriate meter (the default is 1).
Clicking on any one of the constructions that do appear will display its associated
Semantic Specification (SemSpec). So, for example,
if you click on the Transitive construction, its SemSpec will be displayed,
showing (inter alia) three schemas: Referent1 (Speaker), CauseMove,
and Referent2 (Cup). Check to see that the speaker is correctly identified with
the causer and agent, and that the cup is correctly identified with the mover.
Finally, try typing in the sentence Bring the cup. Notice that the
analyzer fails to find a construction which spans the sentence--only single-word
constructions appear in the CHART. Check this against the grammar: the transitive
construction requires three constituents, the agent, the verb and the object.
We know of course that the sentence is grammatically well-formed, but this grammar
has not yet learned the imperative-mood
form of bring.
Part I: Analysis
Using a9.ecg, analyze the following sentences:
(a) I will bring you the cup
What is the difference between this analysis and the one for I will bring the cup? How is this difference captured in the grammar?
(b) I will bring the cup here
Does the analyzer find a spanning analysis? Why or why not? How does
the analysis account for the destination of the bring action, if at
all?
(c) * Nomi give I the ball
Does the analyzer find a spanning analysis here? If it does not, explain why. If it does, propose a change to the grammar that will exclude this sentence from being admitted as grammatical.
Part II: Learning
Now load in nomi.ecg. The file is identical to a9.ecg except it does not contain the following two constructions: the transitive construction and the ditransitive construction. It contains a number of basic schemas and mostly lexical constructions (and a phrase red ball). This approximates the single-word stage in child language acquisition.
Assume that this is the initial grammar that the child learner (Nomi) is using at the beginning of a play time one day. Remember that when a child acquires language, she has a lot of background knowledge as well as contextual knowledge at her disposal. Below is a pictorial description of the scene in which the utterances were said. In the learning system code, this scene will be encoded as a number of schemas linked together. The scene includes the following descriptive knowledge:
Over the course of interacting with her parents and playing with toys, Nomi hears the following six utterances from her parents, accompanied by some context:
Assume that Nomi is a linguistically precocious child, so that she
hypothesizes new constructions (albeit lexically specific ones) with
every utterance she hears. According to the verb-island hypothesis,
children tend first to learn verb-specific constructions (e.g. get-ball
construction, or throw-block construction) and then later on to
generalize across verbs. Based on this principle, as well as the three
learning mechanism we discussed in lecture (relational mapping,
merging, composing), simulate by hand what Nomi might learn from these
sentences.
level 1
constructional constituentsbring: Bring-Cnform constraints
juice: Juice-Cn
bringf before juicefmeaning
evokes Father as father
evokes Nomi as nomi
constraints
bringm.agent <-> father
bringm.direction <-> nomi
bringm.patient <-> juicem