From Structural Syntax to Constructive Adpositional Grammars

F. Gobbo & M. BeniniUniversity of Insubria, Italy

CC© BY:© $\© C©

1 of 14

What is dependency?

The posthumous book by Tesniere (1959) is considered a masterpiece,as it introduces the two key concepts of dependency and valency.

Nonetheless, unlike valency, there is no agreement among scholars andspecialists on how to treat precisely the concept of dependency.

� How Tesniere really defined dependency?

� What can be saved – and adapted – from his work nowadays?

2 of 14

What is dependency?

The posthumous book by Tesniere (1959) is considered a masterpiece,as it introduces the two key concepts of dependency and valency.

Nonetheless, unlike valency, there is no agreement among scholars andspecialists on how to treat precisely the concept of dependency.

� How Tesniere really defined dependency?

� What can be saved – and adapted – from his work nowadays?

2 of 14

Tesniere talked about connection, not dependency!

parle

IAlfred

Stemma 1

parle

Iami

Imon

Stemma 2

In Alfred parle (‘Alfred speaks’), the verb parle is the governor(regissant), the noun Alfred being the dependent (elementsubordonne).

Their relation “indicated by nothing” (1, A, ch. 1, 4) is theirconnection (connexion). Connections are recursive (stemma 2).

3 of 14

Tesnierian Structural Syntax triple

For instance, in Alfred parle (Alfred speaks, stemma 1):

1. governor (parle)

2. dependent (Alfred)

3. connector (ε) – empty? yes, but it does exist indeed!

Tesnierian unary trees – even if recursive – tend to obscure theconnector in the triple, especially when it is collocational (syntactic)instead than morphological.

4 of 14

Tesnierian Structural Syntax triple

For instance, in Alfred parle (Alfred speaks, stemma 1):

1. governor (parle)

2. dependent (Alfred)

3. connector (ε) – empty? yes, but it does exist indeed!

Tesnierian unary trees – even if recursive – tend to obscure theconnector in the triple, especially when it is collocational (syntactic)instead than morphological.

4 of 14

Tesnierian Structural Syntax triple

For instance, in Alfred parle (Alfred speaks, stemma 1):

1. governor (parle)

2. dependent (Alfred)

3. connector (ε)

– empty? yes, but it does exist indeed!

Tesnierian unary trees – even if recursive – tend to obscure theconnector in the triple, especially when it is collocational (syntactic)instead than morphological.

4 of 14

Tesnierian Structural Syntax triple

For instance, in Alfred parle (Alfred speaks, stemma 1):

1. governor (parle)

2. dependent (Alfred)

3. connector (ε) – empty? yes, but it does exist indeed!

Tesnierian unary trees – even if recursive – tend to obscure theconnector in the triple, especially when it is collocational (syntactic)instead than morphological.

4 of 14

Tesnierian Structural Syntax triple

For instance, in Alfred parle (Alfred speaks, stemma 1):

1. governor (parle)

2. dependent (Alfred)

3. connector (ε) – empty? yes, but it does exist indeed!

Tesnierian unary trees – even if recursive – tend to obscure theconnector in the triple, especially when it is collocational (syntactic)instead than morphological.

4 of 14

From unary to binary trees

Stemma 2 in Constructive Adpositional Grammars (CxAdG):

mon

D

����

p↔εF

AAAA

ami

G

����

p↔εF

AAAA

parle

G

� G indicates the grammar character of governors

� D indicates the grammar character of dependents

� F indicates the grammar character of adpositions (= connectors)

5 of 14

From unary to binary trees

Stemma 2 in Constructive Adpositional Grammars (CxAdG):

mon

D

����

p↔εF

AAAA

ami

G

����

p↔εF

AAAA

parle

G

� G indicates the grammar character of governors

� D indicates the grammar character of dependents

� F indicates the grammar character of adpositions (= connectors)

5 of 14

From unary to binary trees

Stemma 2 in Constructive Adpositional Grammars (CxAdG):

mon

D

����

p↔εF

AAAA

ami

G

����

p↔εF

AAAA

parle

G

� G indicates the grammar character of governors

� D indicates the grammar character of dependents

� F indicates the grammar character of adpositions (= connectors)

5 of 14

From unary to binary trees

Stemma 2 in Constructive Adpositional Grammars (CxAdG):

mon

D

����

p↔εF

AAAA

ami

G

����

p↔εF

AAAA

parle

G

� G indicates the grammar character of governors

� D indicates the grammar character of dependents

� F indicates the grammar character of adpositions

(= connectors)

5 of 14

From unary to binary trees

Stemma 2 in Constructive Adpositional Grammars (CxAdG):

mon

D

����

p↔εF

AAAA

ami

G

����

p↔εF

AAAA

parle

G

� G indicates the grammar character of governors

� D indicates the grammar character of dependents

� F indicates the grammar character of adpositions (= connectors)

5 of 14

Dependency in adpositional trees

In adpositional trees (adtrees):

� governors are put on the right, dependents on the left;

� adpositions are put in evidence; they define the structure ofconstructions through the adtree final grammar character (F);

� left-to-right indicators (→) sign dependency, where theinformation prominence is in the dependent;

� right-to-left indicators (←) sign government, where theinformation prominence is in the governor;

� left-to-right & right-to-left indicators (↔) signunderspecification, where the information prominence is notrelevant.

6 of 14

The only mention of “dependency” in Tesniere (1959)

ruisseaux

Ipetits

Sl ettrma 2L

rulsseaux

J,petits

INCIDENCESTRUCTURALE

Stemma 22

ruisseauxfpetits

INCIDENCESÉMANTIQUE

Stemma 23

In adtrees, indicators are interpretations of incidence structural andincidence semantique (a kind of “dependency”) in terms ofinformation prominence, adapted from the dichotomy trajectors(tr) vs. landmarks (lm) by Langacker (1987).

7 of 14

The role of grammar characters

chante

cousrne délicie usement

T/4.OE,4AA

Sternma virtuel

Stemrna 44

votrc je,une

S Lern ln a réel

Stenlma 43

Tesniere borrowed from Esperanto final suffixes the letters of the fouruniversal grammar characters (same characters already in Whorf1945).

8 of 14

Adtrees put all Tesnierian structure together

�

D

�����

q→-ment

E

AAAAA

delicieuse

A

�����

q←εI

AAAAAAAAAA

votre

A

�����

q←εO

AAAAA

jeune

A

�����

q←εO

AAAAA

cousine

O

�����

q→εI

AAAAA

chante

I

This adtree renders both stemmas 43 (reel) and 44 (virtuel) in one.

CxAdGrams are a derivative work of Tesniere’s...

� the original concept of valency is preserved

� the Structural Syntax triple gives the form to CxAdTrees

� dependency is “only” a form of connection, as put by Tesniere

� the four grammar characters are general in CxAdGrams

� information prominence is adapted from Langacker’s tr/lmdichotomy, (at least) sketched by Tesniere himself

10 of 14

CxAdGrams are a derivative work of Tesniere’s...

� the original concept of valency is preserved

� the Structural Syntax triple gives the form to CxAdTrees

� dependency is “only” a form of connection, as put by Tesniere

� the four grammar characters are general in CxAdGrams

� information prominence is adapted from Langacker’s tr/lmdichotomy, (at least) sketched by Tesniere himself

10 of 14

CxAdGrams are a derivative work of Tesniere’s...

� the original concept of valency is preserved

� the Structural Syntax triple gives the form to CxAdTrees

� dependency is “only” a form of connection, as put by Tesniere

� the four grammar characters are general in CxAdGrams

� information prominence is adapted from Langacker’s tr/lmdichotomy, (at least) sketched by Tesniere himself

10 of 14

CxAdGrams are a derivative work of Tesniere’s...

� the original concept of valency is preserved

� the Structural Syntax triple gives the form to CxAdTrees

� dependency is “only” a form of connection, as put by Tesniere

� the four grammar characters are general in CxAdGrams

� information prominence is adapted from Langacker’s tr/lmdichotomy, (at least) sketched by Tesniere himself

10 of 14

CxAdGrams are a derivative work of Tesniere’s...

� the original concept of valency is preserved

� the Structural Syntax triple gives the form to CxAdTrees

� dependency is “only” a form of connection, as put by Tesniere

� the four grammar characters are general in CxAdGrams

� information prominence is adapted from Langacker’s tr/lmdichotomy, (at least) sketched by Tesniere himself

10 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:

� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;

� (linguistic) information can be hidden and recalled entirely in a veryprecise way, without being lost, with every piece clearly described;

� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;

� it was never done before.

11 of 14

...based on a up-to-date formal model

� adtrees and construction together form a (mathematical) category

� the possible finite sequences of morphemes of a language are amonoid M

� the presheaf over M mapping in the monoid gives thelexicalizations of adtrees

� the presheaves space is a Grothendieck topos, so languagestructure can be analysed through the power of the up-to-datemathematical methods of topos theory, which makes sense as:� it is the most general and formal mathematical theory we have;� (linguistic) information can be hidden and recalled entirely in a very

precise way, without being lost, with every piece clearly described;� it was never done before.

11 of 14

How to delve into CxAdGramsOur book published by Cambridge Scholars (C-S-P). Available now.

Warning! This Is An Advertisement

How to delve into CxAdGramsOur book published by Cambridge Scholars (C-S-P). Available now.

Warning! This Is An Advertisement

Conclusion: there is always more in languages...

Figure: from Monty Python’s The Meaning of Life

...than in grammars!

13 of 14

Conclusion: there is always more in languages...

Figure: from Monty Python’s The Meaning of Life

...than in grammars!

13 of 14

Conclusion: there is always more in languages...

Figure: from Monty Python’s The Meaning of Life

...than in grammars!13 of 14

¡Thanks for your attention!

¿Questions?

For proposals, ideas & comments:

{federico.gobbo,marco.benini}@uninsubria.it

Download & share these slides here:

http://www.slideshare.net/goberiko/

CC© BY:© $\© C© Federico Gobbo & Marco Benini 2011

14 of 14