Practical Grammar 6: Difference between revisions

From English Grammar
Jump to navigation Jump to search
No edit summary
 
(193 intermediate revisions by the same user not shown)
Line 1: Line 1:
<font face="Arial, Helvetica, sans-serif">
<font size="3">
== Governable grammatical functions ==
== Governable grammatical functions ==


In linguistics, a head '''governs''' a property of another expression if it requires this expression to have that property. Governable properties include grammatical functions, cases, and particular prepositions.  
In linguistics, a head '''governs''' a property of another expression if it requires this expression to have that property. Governable properties include grammatical functions, cases, and particular prepositions.  
   
   
In Lexical-Functional Grammar, the '''governable grammatical functions''' are those which may be listed in the PRED values of meaningful words The textbook (p. 21) gives a complete list:
In Lexical-Functional Grammar, the '''governable grammatical functions''' are those which may be listed in the PRED values of words. In this course, we will use the following functions:


  Governable grammatical functions: SUBJ, OBJ, OBJ-θ, OBL-θ, POSS, COMP, and XCOMP.
  Governable grammatical functions: SUBJ, OBJ, OBJ-θ, OBL-θ, POSS, COMP, and XCOMP.
Line 9: Line 12:
There are other GFs, for instance TOPIC and FOCUS, but heads cannot govern those.  
There are other GFs, for instance TOPIC and FOCUS, but heads cannot govern those.  


The Greek letter θ in the functions OBJ-θ and OBL-θ stands for the name of a thematic role (the thematic roles are listed on p. 13 in the textbook). Thus, OBL-LOC is the GF of a locative PP (''sit <u>at the window</u>''), OBL-BEN the GF of a benefactive PP (''bake a cake <u>for Mary</u>''). The only thematic role that appears in OBJ-θ is THEME, the second object of verbs of giving and showing.  
The Greek letter θ (= theta) in the functions OBJ-θ and OBL-θ is an abbreviation. It stands for the name of a thematic role.


Here are some comments on the functions other than SUBJ and OBJ:
Here are some comments on the functions other than SUBJ and OBJ:


* OBJ-THEME: the GF of the second NP object of a ditransitive verb like ''give'' (e.g. ''give Mary <u>a present</u>'')
* OBJ-THEME: the GF of the second NP object of a ditransitive verb like ''give'' (e.g. ''give Mary <u>a present</u>'')
* OBL-θ, for instance OBL-LOC: the GF of a PP which is a locative complement of a verb like ''sit'' (e.g. ''sit <u>at the window</u>'')
* OBL-LOC: the GF of a locative complement of a verb like ''sit'' (e.g. ''sit <u>at the window</u>'')
* OBL-BEN: the GF of a benefactive PP (''bake a cake <u>for Mary</u>'')
* POSS: the GF of the possessor within an NP (e.g. ''<u>my</u> book'')
* POSS: the GF of the possessor within an NP (e.g. ''<u>my</u> book'')
* COMP: the GF of a finite complement clause complement of a verb like ''say'' (e.g. ''said <u>that Mary called</u>'')
* COMP: the GF of a finite complement clause complement of a verb like ''say'' (e.g. ''said <u>that Mary called</u>'')
Line 29: Line 33:


  '''The Completeness Condition''' (p. 21)
  '''The Completeness Condition''' (p. 21)
  All governable functions which are part of th value of a PRED feature must be present in the local f-structure. All functions that have a θ-role must themselves have a PRED value.
  All governable functions which are part of the value of a PRED feature must be present in the local f-structure. All functions that have a θ-role must themselves have a PRED value.


This is the "no less" part from above: the f-structure of a PRED must contain at least the GFs which it selects in its PRED value.
This is the "nor less" part from above: the f-structure of a PRED must contain at least the GFs which it selects in its PRED value.


The "no more" part is stated in the  
The "neither more" part is stated in the  


  '''Coherence Condition''' (p. 22)
  '''Coherence Condition''' (p. 22)
  All governable functions present in an f-structure must occur in the value of a local PRED feature. All functions that have a PRED value mst have a θ-role.
  All governable functions present in an f-structure must occur in the value of a local PRED feature. All functions that have a PRED value must have a θ-role.


So, an f-structure is only '''complete''' and '''coherent''' only if there is a 1:1 relation between the GFs listed in its PRED value and the actual GFs present in the f-structure! If this is not the case, then the f-structure is ill-formed.
So, an f-structure is '''complete''' and '''coherent''' only if there is a 1:1 relation between the GFs listed in its PRED value and the actual GFs present in the f-structure! If this is not the case, then the f-structure is ill-formed.


== Governable features ==
== Governable features ==


Above, it was said that not only the GFs of its dependents, but other properties can be governed by a selecting head. These include inflectional features like PERS, NUM, and CASE. In other words, a head can require that its dependents have particular values for such features. In this section, we will see how this is done.
Above, it was said that not only the GFs but also other properties  of its dependents can be governed by a selecting head. These include inflectional features for person and number, and CASE. In other words, a head can require that its dependents have particular values for such features. In this section, we will see how this is done in XLFG.


In Week 3, we already saw how features can be added to words themselves. There we formulated lexical entries like the following:
We will begin with subject-verb agreement.
 
=== Subject-verb agreement ===
 
In Week 3, we already saw how features can be added to words. There, we formulated lexical entries like the following:


  this D
  this D
       [PERS:3,
       [NUM:sg];
      NUM:sg];
 
English determiners, nouns, and verbs are not only marked for NUMBER, however, but also for PERSON. And agreement between finite verbs and their subjects involves both features. Therefore, from here on, we will replace the NUM feature by the feature AGR, which can take 6 values:
 
AGR values: _1sg, _2sg, _3sg, _1pl, _2pl, _3pl
 
The underscore is part of the feature name and is necessary, since XLFG throws an error without it (because the system doesn't allow feature values to start with a number).


We can do the same for defining pronouns:
We can now extend our lexicon by adding lexical entries for pronouns, as follows:


  she N
  she N
     [PERS:3,
     [AGR:_3sg];
      NUM:sg,
      CASE:nom];


me N
== Past tense verbs ==
    [PERS:1,
    NUM:sg,
    CASE:acc];


This is the easy case, since (except for the auxiliary ''be'') all English verbs have only one form in the past tense and this form is compatible with subjects of all AGR values. The most economical treatment for this is to make words like ''disappeared'' impose no constraints on the AGR value of their subject.


==== <span style="color: blue>Exercise 6.1</span> The past tense ====


In order for the word above to impose feature values on its SUBJ, it can simply refer to the f-structure of that SUBJ, as follows:
1. Go to <span class="newwin">https://147.210.117.56</span> and log in.<br>
2. Open ''Grammar 6''.<br>
3. Copy the verb ''disappeared'' from your version of Grammar 4 and add it to the lexicon.<br>
4. Add lexical entries for all English personal pronouns, i.e. ''I'', ''you'', ... with their correct AGR values. You will need two entries for ''you''.
5. For each pronoun, add a sentence to the testsuite that consists of the pronoun followed by ''disappeared'' and nothing else.<br>


disappears D
6. Parse and make sure that all 6 sentences come out as grammatical.
            [PRED &nbsp; 'DISAPPEAR<SUBJ>',
            PERS:3,
            NUM:sg,
            '''SUBJ &nbsp; [...]'''];


The presence of a SUBJ in the verb's f-structure is guaranteed by '''Completeness''': since the verb selects a SUBJ in its PRED value, there has to be a SUBJ. And that subject has its own f-structure. Any feature-values pairs in the place of the dots inside the embedded brackets above must be present in the f-structure of the SUBJ. This is an example of feature government.
== Present tense verbs ==


<span style="color: blue>Exercise</span>
=== Subject-verb agreement ===


1. Go to https://xlfg.labri.fr/ and log in.<br>
2. To the verb ''disappears'', add constraints that ensure that its SUBJ is third person singular and bears nominative case.<br>
3. Add the following test sentences:<br>
: She disappears
: *We disappears
: *Her disappears
4. Parse.


== Macros ==
Next, we need to ensure that only expressions whose AGR value is _3sg can act as subjects of a present tense verb like ''disappears''. This is easy to accomplish: in order for ''disappears'' to impose on its SUBJ that the SUBJ's AGR value must be _3sg, the verb can simply specify the f-structure of that SUBJ as follows:


The treatment of the present tense forms other than the third person singular is more complicated. An unmarked form like ''disappear'' can be 1sg, 2sg, 1pl, 2pl, and 3pl. We need an efficient way of encoding this informtion. To this end, we introduce macros, i.e. abbreviations that we define.
disappears V
            [PRED:'DISAPPEAR<SUBJ.THEME>',
            '''SUBJ:[AGR:_3sg]'''];


We start small. Since there are so many third person singular verbs, we often have to write
The presence of a SUBJ in the verb's f-structure is guaranteed by '''Completeness''': since the verb selects a SUBJ in its PRED value, there has to be a SUBJ in the verb's f-structure. And that SUBJ has its own f-structure. The f-structure of the SUBJ of ''disappears'' is well formed only if it contains the attribute AGR and the value of AGR is _3sg. This is an example of '''feature government'''.


[...,
==== <span style="color: blue>Exercise 6.2</span> The third person singular present tense ====
  PERS:3,
  NUM:sg,
  ...]


This is annoying. To make life easier on us, we can define a macro which allows us to encode this information in a shorter way. A macro is a statement of the following form:
1. Go to <span class="newwin">https://147.210.117.56</span> and log in.<br>
2. Open your version of ''Grammar 6''.<br>
3. Add an entry for the verb ''disappears'' and ensure that its SUBJ is _3sg. Note that the verb in this system does NOT need to have the feature AGR itself! Rather, the verb form with the shape ''disappears'' requires its SUBJ to have a particular value for this feature.<br>
4. Add the following test sentences:<br>


Macro format
(4) she disappears<br>
@macro-name: attribute:value;
(5) *i disappears<br>
(6) it disappears<br>


You can also have several attribute-value statements on the right. The statements are separated by commas. We will see examples presently.
5. Parse.


So, let us define our first two macros:


@s: NUMBER:sg;
The present tense verb form ''disappear'' is less convenient to handle, since it is inbetween allowing only one kind of subject or all subjects. Rather, it allows subjects with 5 out of 6 AGR values:
@p: NUMBER:pl;


The first macro says that we can use the expression "@s" to abbreviate "NUMBER:sg".
(7) i disappear<br>
(8) you disappear<br>
(9) *he/she/it disappear<br>
(10) we disappear<br>
(11) you disappear<br>
(12) they disappear


Here are macros for different person values:
This is inconvenient to state in any formal grammar of English, but that is not the grammarian's fault. Languages aren't always completely regular! One way of capturing the generalization would be to state in each non-third-person singular verb's lexical entry that its subject's AGR value can be


@_1: PERS:1;
_1sg or _2sg or _1pl or _2pl or _3pl
@_2: PERS:2;
@_3: PERS:3;


It get better: the definition of one macro may contain other macro calls (a macro calls is the use of a macro name in an environment where that macro is not defined). The examples below illustrate this:
But that would mean that the grammar would generate 5 different words from each of these lexical entries.


@_1s: @_1, @s;
Obviously, it would be more economical in this case to state which AGR value the subject of the verb is NOT allowed to have. To express this, XLFG offers the inequality operator:  
@_2s: @_2, @s;
 
@_3s: @_3, @s;
@_1p: @_1, @p;
@_2p: @_2, @p;
  @_3p: @_3, @p;


The first line says that "@_1s" is short for "@_1, @s", which, when we unpack the two macros on the right means that "@_1s" is short for "PERS:1, NUM:sg". And the same for the five other person-number combinations.
With this operator, you can say things like


Finally comes another cool feature. Macros can contain '''the symbol "|"''' (= stroke), which means "or". Here is an example:
AGR ≠ _3sg


@_12: PERS:1|2;
which means that the AGR value can be anything other than _3sg. That is the condition that we what to impose on the SUBJ value of non-third-person singular present tense verbs in English.


Of course, this says that "@_12" is short for "the PERS value can be 1 or 2".
Unfortunately, XLFG requires a particular format for using the inequality operator. Therefore, I give you the correct way of stating the condition in the lexical entry of ''disappear'' below:


Now we have all the ingredients to say in a relatively short manner that forms like ''disappear'' are compatible with subjects that are non-third person singular:
disappear V
[...]
{(↑ SUBJ AGR) ≠ _3sg;};


@non3s: [@_12] | [@_3p] ;
So, just specify all the information about the lexical entry as usual, but before you signal the end of the lexical entry with a semicolon, add the statement


Read this carefully: "@non3s" is the name of the macro. This name abbreviates a feature structure which is either first or second person or it is third person plural. Note that these are precisely the person-number combinations other than third perso singular!
{(↑ SUBJ AGR) ≠ _3sg;};


<span style="color: blue>Exercise</span>
as the last line. Use the buttons for the up arrow and the inequality symbol from the XLFG lexicon editor or otherwise XLFG will complain!!!


1. Go to https://xlfg.labri.fr/ and log in.<br>
==== <span style="color: blue>Exercise 6.3</span> The non-third person singular present tense ====
2. Add all the macros above in the '''Declarations''' editor.
 
3. Add the verb ''disappear'' to the lexicon and add constraints that ensure that its SUBJ is non-third person singular and bears nominative case.<br>
1. Go to <span class="newwin">https://xlfg.labri.fr/</span> and log in.<br>
4. Add the following test sentences:<br>
2. Open your latest version of ''Grammar 6''.<br>
: I disappear
3. Add the two versions of the verb ''disappear'' to the lexicon as given above<br>
: You disappear
4. Add (7)-(12) as test sentences:<br>
: *She disappear
5. Parse.
5. Parse.
=== Case ===
The treatment of English case is pretty straightforward. For now, we will only deal with arguments of the verb so that we have to account for nominative and accusative pronouns. We thus postulate a feature CASE with the two possible values ''nom'' and ''acc''. Some pronouns are only compatible with one of these values, but other pronouns and all non-pronominal nouns are compatible with both cases.
==== <span style="color: blue>Exercise 6.4</span> Nominative and accusative ====
1. Where necessary, add the appropriate case information to all the nouns and verbs in your current grammmar.
2. Add the following test items::<br>
(13) *her disappeared<br>
(14) lilly disappeared<br>
(15) *me disappeared<br>
3. Parse!
==== <span style="color: blue>Exercise 6.5</span> Nominative and accusative ====
1. . Add the following test items:<br>
(16) lilly liked her<br>
(17) *lilly liked she<br>
(18) lilly liked it<br>
(19) lilly liked john<br>
(20) lilly liked the cat<br>
(21) the cat liked lilly<br>
3. Parse!
<br>
<br>
<font face="Arial, Helvetica, sans-serif">
<font size="2">
<div align="center">
    [[Practical_Grammar|'''Main page''']]  [[Practical_Grammar_2|'''Week 2''']]  [[Practical_Grammar_3|'''Week 3''']]  [[Practical_Grammar_4|'''Week 4''']]  [[Practical_Grammar_5|'''Week 5''']]  '''Week 6'''  [[Practical_Grammar_7|'''Week 7''']]  [[Practical_Grammar_8|'''Week 8''']]  [[Practical_Grammar_9|'''Week 9''']] 
</div>

Latest revision as of 07:26, 15 June 2026

Governable grammatical functions

In linguistics, a head governs a property of another expression if it requires this expression to have that property. Governable properties include grammatical functions, cases, and particular prepositions.

In Lexical-Functional Grammar, the governable grammatical functions are those which may be listed in the PRED values of words. In this course, we will use the following functions:

Governable grammatical functions: SUBJ, OBJ, OBJ-θ, OBL-θ, POSS, COMP, and XCOMP.

There are other GFs, for instance TOPIC and FOCUS, but heads cannot govern those.

The Greek letter θ (= theta) in the functions OBJ-θ and OBL-θ is an abbreviation. It stands for the name of a thematic role.

Here are some comments on the functions other than SUBJ and OBJ:

  • OBJ-THEME: the GF of the second NP object of a ditransitive verb like give (e.g. give Mary a present)
  • OBL-LOC: the GF of a locative complement of a verb like sit (e.g. sit at the window)
  • OBL-BEN: the GF of a benefactive PP (bake a cake for Mary)
  • POSS: the GF of the possessor within an NP (e.g. my book)
  • COMP: the GF of a finite complement clause complement of a verb like say (e.g. said that Mary called)
  • XCOMP: the GF of non-finite phrases with understood subjects (e.g. Lilly tried to leave)

The Completeness and Coherence Conditions

Words may specify in their lexical entries that they govern one or more GFs. These GFs are mentioned in the word's PRED value, e.g.

(1) disappear: [PRED   'DISAPPEAR<SUBJ>']
(2) see: [PRED   'SEE<SUBJ, OBJ>']

This means that the verb disappear governs one GF, namely a SUBJ and the verb see governs two GFs, a SUBJ and an OBJ. As a result, these verbs must appear with exactly these GFs, neither more, nor less. This is ensured by the following two principles:

The Completeness Condition (p. 21)
All governable functions which are part of the value of a PRED feature must be present in the local f-structure. All functions that have a θ-role must themselves have a PRED value.

This is the "nor less" part from above: the f-structure of a PRED must contain at least the GFs which it selects in its PRED value.

The "neither more" part is stated in the

Coherence Condition (p. 22)
All governable functions present in an f-structure must occur in the value of a local PRED feature. All functions that have a PRED value must have a θ-role.

So, an f-structure is complete and coherent only if there is a 1:1 relation between the GFs listed in its PRED value and the actual GFs present in the f-structure! If this is not the case, then the f-structure is ill-formed.

Governable features

Above, it was said that not only the GFs but also other properties of its dependents can be governed by a selecting head. These include inflectional features for person and number, and CASE. In other words, a head can require that its dependents have particular values for such features. In this section, we will see how this is done in XLFG.

We will begin with subject-verb agreement.

Subject-verb agreement

In Week 3, we already saw how features can be added to words. There, we formulated lexical entries like the following:

this D
     [NUM:sg];

English determiners, nouns, and verbs are not only marked for NUMBER, however, but also for PERSON. And agreement between finite verbs and their subjects involves both features. Therefore, from here on, we will replace the NUM feature by the feature AGR, which can take 6 values:

AGR values: _1sg, _2sg, _3sg, _1pl, _2pl, _3pl

The underscore is part of the feature name and is necessary, since XLFG throws an error without it (because the system doesn't allow feature values to start with a number).

We can now extend our lexicon by adding lexical entries for pronouns, as follows:

she N
    [AGR:_3sg];

Past tense verbs

This is the easy case, since (except for the auxiliary be) all English verbs have only one form in the past tense and this form is compatible with subjects of all AGR values. The most economical treatment for this is to make words like disappeared impose no constraints on the AGR value of their subject.

Exercise 6.1 The past tense

1. Go to https://147.210.117.56 and log in.
2. Open Grammar 6.
3. Copy the verb disappeared from your version of Grammar 4 and add it to the lexicon.
4. Add lexical entries for all English personal pronouns, i.e. I, you, ... with their correct AGR values. You will need two entries for you. 5. For each pronoun, add a sentence to the testsuite that consists of the pronoun followed by disappeared and nothing else.

6. Parse and make sure that all 6 sentences come out as grammatical.

Present tense verbs

Subject-verb agreement

Next, we need to ensure that only expressions whose AGR value is _3sg can act as subjects of a present tense verb like disappears. This is easy to accomplish: in order for disappears to impose on its SUBJ that the SUBJ's AGR value must be _3sg, the verb can simply specify the f-structure of that SUBJ as follows:

disappears V
           [PRED:'DISAPPEAR<SUBJ.THEME>',
            SUBJ:[AGR:_3sg]];

The presence of a SUBJ in the verb's f-structure is guaranteed by Completeness: since the verb selects a SUBJ in its PRED value, there has to be a SUBJ in the verb's f-structure. And that SUBJ has its own f-structure. The f-structure of the SUBJ of disappears is well formed only if it contains the attribute AGR and the value of AGR is _3sg. This is an example of feature government.

Exercise 6.2 The third person singular present tense

1. Go to https://147.210.117.56 and log in.
2. Open your version of Grammar 6.
3. Add an entry for the verb disappears and ensure that its SUBJ is _3sg. Note that the verb in this system does NOT need to have the feature AGR itself! Rather, the verb form with the shape disappears requires its SUBJ to have a particular value for this feature.
4. Add the following test sentences:

(4) she disappears
(5) *i disappears
(6) it disappears

5. Parse.


The present tense verb form disappear is less convenient to handle, since it is inbetween allowing only one kind of subject or all subjects. Rather, it allows subjects with 5 out of 6 AGR values:

(7) i disappear
(8) you disappear
(9) *he/she/it disappear
(10) we disappear
(11) you disappear
(12) they disappear

This is inconvenient to state in any formal grammar of English, but that is not the grammarian's fault. Languages aren't always completely regular! One way of capturing the generalization would be to state in each non-third-person singular verb's lexical entry that its subject's AGR value can be

_1sg or _2sg or _1pl or _2pl or _3pl

But that would mean that the grammar would generate 5 different words from each of these lexical entries.

Obviously, it would be more economical in this case to state which AGR value the subject of the verb is NOT allowed to have. To express this, XLFG offers the inequality operator:

With this operator, you can say things like

AGR ≠ _3sg

which means that the AGR value can be anything other than _3sg. That is the condition that we what to impose on the SUBJ value of non-third-person singular present tense verbs in English.

Unfortunately, XLFG requires a particular format for using the inequality operator. Therefore, I give you the correct way of stating the condition in the lexical entry of disappear below:

disappear V
[...]
{(↑ SUBJ AGR) ≠ _3sg;};

So, just specify all the information about the lexical entry as usual, but before you signal the end of the lexical entry with a semicolon, add the statement

{(↑ SUBJ AGR) ≠ _3sg;};

as the last line. Use the buttons for the up arrow and the inequality symbol from the XLFG lexicon editor or otherwise XLFG will complain!!!

Exercise 6.3 The non-third person singular present tense

1. Go to https://xlfg.labri.fr/ and log in.
2. Open your latest version of Grammar 6.
3. Add the two versions of the verb disappear to the lexicon as given above
4. Add (7)-(12) as test sentences:
5. Parse.

Case

The treatment of English case is pretty straightforward. For now, we will only deal with arguments of the verb so that we have to account for nominative and accusative pronouns. We thus postulate a feature CASE with the two possible values nom and acc. Some pronouns are only compatible with one of these values, but other pronouns and all non-pronominal nouns are compatible with both cases.

Exercise 6.4 Nominative and accusative

1. Where necessary, add the appropriate case information to all the nouns and verbs in your current grammmar. 2. Add the following test items::
(13) *her disappeared
(14) lilly disappeared
(15) *me disappeared

3. Parse!

Exercise 6.5 Nominative and accusative

1. . Add the following test items:

(16) lilly liked her
(17) *lilly liked she
(18) lilly liked it
(19) lilly liked john
(20) lilly liked the cat
(21) the cat liked lilly

3. Parse!