Due: Friday, March 13 (by noon)
Submit assignments electronically to all three teachers
(ron.kaplan "at" microsoft.com, tracy.king "at" microsoft.com,
mforst "at" parc.com)
| Turn in: | 1. the final grammar you end up with (eng-week7.lfg) |
| 2. a testsuite with sentences that your grammar does and does not cover | |
| Please name your grammars with name-eng-week7.lfg |
| Choose one of the following: | |
| CHOICE 1: | Interrogatives |
| CHOICE 2: | Auxiliary Stacking |
| CHOICE 3: | Predicative Adjectives |
You can use the grammar eng-week7.lfg or any other grammar used in the earlier exercises (other than the big English grammar used in the MT assignment last week) as your starter grammar.
The aim of this exercise is to add matrix interrogatives to the grammar. You will add both polar (yes-no) interrogatives and wh interrogatives. When doing these, do not worry too much about constraining the stacking of the auxiliaries (if you are particularly interested in this, do choice 2 instead), but do be sure that you can handle the basic patterns of the interrogatives.
In general, these exercises will involve changes to both the c-structure rules and the f-structure annotations, as well as adding a few lexical entries. Do not worry about the behavior of interrogatives as embedded clauses (e.g. They asked who he had seen, They asked whether he had left).
You will need to add in lexical entries for do as an auxiliary and for the interrogative mark, as well as alter the c-structure rules to allow the subject-auxiliary inversion. You may also need to change the equations that the lexical entries for the bare forms of the verbs (e.g. sleep) contain.
Declaratives should have an f-structure feature STMT-TYPE declarative, while interrogatives should have a feature STMT-TYPE interrogative.
You may want to create two S categories: one for declaratives and one for interrogatives.
Wh-interrogatives involve a fronted wh-phrase and for non-subject wh-questions there is subject-auxiliary inversion.
Add entries for who and what and alter the c-structure rules to allow these types of interrogatives. This will be similar to the relative clause rules that you did earlier in the course: you will need to be sure both that appropriate c-structures can be built and that you have the correct functional uncertainty equations.
Extend your rules to include extraction of wh-phrases out of embedded clauses. The patterns for these can get very complicated. Here, just be sure that your grammar allows for extractions that you know are good; do not worry about blocking extractions that are marginal or bad.
Add negation to your grammar. First create a lexical entry for not and n't using a unique c-structure category (e.g. NEG). You can either have negation provide a feature in the f-structure (e.g. NEG +) or be an ADJUNCT (e.g. with PRED 'not').
Then alter the c-structure rules to allow negation in declaratives. Negation appears after the first (tensed) auxiliary.
Now check whether your negation rules cover the interrogatives that you added previously. You will want to be able to parse things like:
In addition to all of the above sentences, add a few more interrogatives (with and without negation) that your grammar covers to a testfile.
English requires that auxiliaries appear in a particular order and each auxiliary requires a particular form after it.
The idea behind this task is to capture the basics of auxiliary stacking, including the ordering and form requirements of the auxiliaries. In addition, each auxiliary should provide features to the f-structure that encode the meaning of the auxiliary (e.g. the PASSIVE + feature from the earlier homework).
Modals such as should, would, could are generally analyzed as verbs that take XCOMPs which are headed by bare infinitives (e.g. infinitive which are not preceded by to). Add lexical entries for these modals, giving them a distinct c-structure category, and alter the c-structure rules to allow them. Modals are generally considered to be tensed even though they do not inflect and so provide a TENSE feature for them. Note that you may need to alter the equations in the lexical entries of the bare forms of the verbs.
Note that in most dialects of English you cannot stack modals. Constrain your rules to block sentences like:
This will be the hardest part of the task. You will need to create entries for the auxiliaries (there are already tensed entries for some of the forms of be). In addition to the tensed forms, you will need the forms that occur when they are stacked under other auxiliaries. The tricky part is to get each form to occur in the correct places and to require the correct forms under it. You can do this with a combination of c-structure categories (feel free to create very specific categories like AUXtensed) and additional features (similar to the features used to constrain the PASSIVE).
For the tensed auxiliaries, only worry about the third person singular forms, although if you would like to include the other forms for completeness, that is fine. (That is, all the examples can have third person singular subjects.)
There are two schools of thought on how to analyze English auxiliaries:
Choose one of these approaches and implement it.
Your grammar will need to allow sequences like:
Make sure that your grammar can embeded passive verbs under all the auxiliary combinations. The basics of the passive are already in place (using the same analysis as in your previous homework assignment). You will want to cover things like:
Create a testsuite with the above sentences and with a set of bad examples. Make sure that your grammar does not parse the bad examples. For example:
Add negation to your grammar. First create a lexical entry for not using a unique lexical category (e.g. NEG). You can either have negation provide a feature in the f-structure (e.g. NEG +) or be an ADJUNCT (e.g. with PRED 'not'). (Do not worry about n't for this task.)
Then alter the c-structure rules to allow negation in declaratives. Negation appears after the first (tensed) auxiliary. Note the use of do support in the first example; you will probably need to add this as a special auxiliary which only occurs with negation (and in interrogatives).
Make sure that your rules block negation from occurring lower in the stack. (These do appear sometimes in English as constituent negation, but they should be blocked as sentential negation.)
This task adds more complicated adjectival constructions than we have seen so far in the course. For the purposes of this task, we will focus on predicative adjectives that appear after the verb be.
First add in the copular verb be in its various tensed forms; some of these already exist in the lexicon as auxiliaries, but copular be should be a verb. It should take a subject and an adjectival argument. There is much controversy in the LFG literature as to the nature of the adjectival argument. To make this exercise easier, use ACOMP which is like an XCOMP in that its subject is controlled but which can only be an adjective. For a sentence like it is purple you want a structure roughly like:
[ PRED 'be' SUBJ [ PRED 'it' ]1 ACOMP [ PRED 'purple ' SUBJ []1 ] ]
Make sure that you can parse various forms of the copula.
Certain adjectives can take PP arguments, which are usually analyzed as OBL (obliques). Some examples of this type are:
Alter your grammar to allow this construction. You will need to change the AP rule as well as adding in new lexical entries and templates for this subcategorization frame. Since these constructions do not occur in attributive position in English, you will need to constrain the grammar to allow them only in predicative position.
Create a testsuite for this task that includes the sentences listed in the task as well as additional forms that your grammar can and cannot parse. Be sure that your grammar can still parse simple forms (e.g. he is proud as well as he is proud of her). Add to this testsuite as you complete the exercises.
Just as certain verbs take infinitival arguments, as you saw in earlier homeworks, some adjectives allow infinitival arguments in copular position. Some examples of this type are:
Once again, alter your AP rule to allow these constructions. Like with the obliques, these only occur in predicative position. You will need to make sure that the control equations for these adjectives supply a subject to the infinitival argument.
In this exercise you will add comparative forms to adjectives. First create an entry for happier that looks like an adjective but adds in a new feature DEGREE comparative; also add an entry for better. Comparatives should have a PRED that is identical to that of their positive (non-comparative) form. Comparative adjectives can occur in both predicative and attributive position. Make sure that your grammar can parse:
Comparatives in predicative position can take than clauses. These clauses can have all types of complements. For this exercise, we will only worry about NP complements. Create a lexical entry for than and add to the grammar a rule which combines than with an NP complement (you can assign this an OBJ or a COMP role, whichever you feel is more appropriate). Make sure your rule works by parsing than bananas and than it.
Alter the AP rule to allow the than clause to occur with comparatives in predicative position. Decide what grammatical function the than clause should have.
Check whether the comparatives are working properly with the other adjectival constructions.
For whichever task you chose, turn in:
If you have any questions, you can send us email (ron.kaplan "at" microsoft.com, tracy.king "at" microsoft.com, mforst "at" parc.com), call us (Ron: 650-245-6865, Tracy: 415-848-7276, Martin: 650-812-4788), or set up office hours with us.