Practical Grammar 3

Features

The lexical entries in Grammar 1 all looked like the following:

this D;
those D;
bottle N;
bottles N;

Together with the rule

NP -> D N;

the grammar accepts all the following strings as grammatical, even though (2) and (4) are ungrammatical:

(1) this bottle
(2) *this bottles
(3) those bottles
(4) *those bottle

In this unit, we will change the grammar so that it makes the correct predictions about (1)-(4).

The source of the problem pointed out above is easy to spot: English determiners and nouns can both be singular or plural and in an NP of the form 'D N' the two words have to agree in number: either they are both singular or both plural. (1)-(4) illustrate this.

What this shows, is that so far, the grammar does not contain enough information about words. Besides a part of speech, words also have inflectional features and the values of these features are regulated in the syntax (this is why these features are also called morphosyntactic features: they determine the morphological shape of words, but their distribution is determined by the syntax).

Adding features to words

So, let us begin adding features to words. First, some terminology. We make a distinction between a feature and its value(s). This is illustrated for nouns and determiners below:

1. Nouns and determiners carry the feature NUM(ber).
2. The possible values of the feature NUM are: sg, pl.

With these concepts, we can add features to the words in (1)-(4) which will lead to the correct predictions for these sentences by our grammar.

Exercise 3.1

1. Go to https://xlfg.labri.fr/ and log in.
   

Then you locate our workshop:

1. Choose Workshops|All Workshops.
2. Look for the workshop "Syntax 1" for the current semester.
3. In the column "Action", click the button "Consult the workshop".

Now you find a project of the workshop:

1. On the right side of the screen is the list of projects of our workshop.
2. Look for "Grammar 2" and click on "Create your own version".
3. Now click on "Edit your version".

• Go to the lexicon and change the lexical entry of the word this exactly as is shown below:

this D
     [NUM:sg];

• Parse.

You will notice a difference from Grammar 1. When you parsed a string in the previous grammar, xlfg showed you one window per string which was labeled 'Constituent Structure'. Now there is a second window which is labeled 'Functional Structure'. Unfortunately, that second window is empty. So, where is the inflectional information that we added to the word this?

• Click on the D node in the Constituent Structure.

Ahh! Now the D node has a colored block around it and simultaneously the Functional Structure shows the feature NUM and its value we added for the word this earlier.

What we are seeing here for the first time is that the grammar associates two different kinds of information with words (and also phrases, as we will soon see):

C(onstituent)-structure information consists of the syntactic category (= parts of speech) of an expression in a phrase structure tree.

F(unctional)-structure information consists of features and values that express the functions of an expression.

Exercise 3.2

• Add NUM features (and their correspondig values) to the three other words in the lexicon.
• Parse.
• Click on the D or N node in Constituent Structure to display the f-structure information associated with that node in Functional Structure.

Exercise 3.3

• Display the Constituent Structure of those bottle.
• Click on the D and the N node to display their f-structures.
• They differ, as expected.
• Yet, the grammar does not mark the string as ungrammatical! Think about why that is! Determiners and nouns are supposed to agree in NUM. Here, they don't agree in NUM and yet the grammar does not seem to mind. Again, why?

The answer is the following: you know that determiners and nouns agree in English, but the grammar doesn't know! Why? Because you haven't told it!

You are the grammar writer: the grammar contains all and only the information that you have given it!

Annotations of phrase structure rules

In stating the determiner-noun agreement rule, we are going to encounter a central tool of Lexical-Functional Grammar, the grammatical theory that xlfg is based on: annotations of phrase structure rules.

The first c-structure rule that we are going to annotate is the NP rule below:

NP -> D N;

Recall from Unit 1 that this rule accepts a local non-lexical tree, as follows:

Mother = NP
Daughter 1 = D
Daughter 2 = N

This is just c-structure information. But each of the three nodes also has an f-structure and these 3 f-structures need to be related in a certain way.

First, some important notation:

In c-structure rules, 
the up-arrow ↑ refers to the f-structure of the mother of the local tree;
down-arrows ↓1, ↓2, ... ↓n with a number refer to the f-structure of the first, second, or nth daughter of the local tree.

The following is one of the most important constraints in any lexical-functional grammar:

↑=↓n;

Here is what it means:

the f-structure of the mother of the local tree licensed by the current rule (= ↑) is identical to the f-structure of the nth daughter of that local tree (↓n).

Another way of saying the same thing is: the mother's f-structure is also the daughter's f-structure and the other way round. There is only one f-structure and it describes the functional information associated with both mother and daughter.

With this tool, we can change the NP rule to the following one, which will have the consequence that D and N agree in NUM:

1. NP -> D N
2. {
3.  ↑=↓1;
4.  ↑=↓2;
5. }

This rule now says all and only the following:

1. Line 1: the rule licenses a local tree whose mother is NP, whose first daughter is D, and whose second daughter is N.
2. Line 2: the opening parenthesis signals that what follows are functional annotations of the nodes in line 1.
3. Line 3: it says that the NP mother and the D daughter have one and the same f-structure.
4. Line 4: it says that the NP mother and the N daughter have one and the same f-structure.
5. Line 5: the closing parenthesis signals the end of the functional annotations and therefore the end of the c-structure rule.

Exercise 3.4

• Go back to https://xlfg.labri.fr/.
  

Part 1:

• Parse.
• Click on "this bottle".
• Click on the D and N nodes and inspect their f-structures.
• Click on the NP node and inspect its f-structure. There is none.

Part 2:

• Go to the grammar editor.
• Change the NP -> D N rule so that it looks exactly as above.
• Repeat Part 1.
• Note the difference to the result in Part 1 when you click on NP.

Part 3:

• Repeat Part 2, but now click on "those bottle".
• Click on the NP.
• Try to interpret what you see.

Exercise 3.5

• Add a single lexical entry for the determiner a and a single lexical entry for the determiner the and make NO other changes to the grammar, except to add the following test sentences:

(1) a bottle
(2) *a bottles
(3) the bottle
(4) the bottles

• Check whether your grammar makes the correct predictions for (1)-(4).

Exercise 3.6

Instead of the f-structures that the grammar currently produces for the whole NP in (1) and (3) above, we would like to obtain the following ones (note: in xlfg, the feature structures below will appear in 2 lines):

(4) a bottle: [NUM sg, DEF -]

(5) the bottle: [NUM sg, DEF +]

• Change the grammar so that it yields the desired f-structures.

What our grammar covers so far:
- lexical entries of the most important parts of speech and their features
- NPs, VPs, APs, and PPs
- agreement in number between determiners and nouns
- definite vs. indefinite NPs

   Main page  Week 2  Week 3  Week 4  Week 5  Week 6  Week 7  Week 8  Week 9