Upload
cedep
View
2
Download
0
Embed Size (px)
Citation preview
DYNAMIC SYNTAX 152900093
ASSIGNMENT 1
Submitted By: Darshita Shah (537969)
Submitted To: Hannah Gibson
Due Date: 24/02/2014
Joel helped Miriam, who he knows dislikes syntax.
Dynamic Syntax (DS) claims that humans’ knowledge of language is essentially their
ability to parse spoken language. It is a formal model designed to capture, how a
listener interprets spoken language step by step. DS illustrates the parsing in human
brain, through tree structures and shows how a hearer progressively builds
semantic representation from lexical and contextual information. The tree structure
grows, as each word is parsed. The steps by which, we reach the final output, i.e. the
final tree, are as important as the final tree in DS.
IN DS each derivation begins with an axiom, which represents the goal to build a
tree with propositional content and reflects the hearer’s expectation to derive a
proposition (Cann et al. 2005). The type of expression is its semantic category,
associating an expression with a particular sort of denotation. This is indicated by a
Type label – Ty(x). A proposition is of type t. Thus, tree building begins with a
requirement, indicated by a question mark, to derive an expression of type t. (Cann,
et al. 2005: 35)
(1) Tn(0),?Ty(t), ◊
The pointer (◊) in (1) indicates the current node, which is under development. Tn is
the label for a tree node. It is the mean by which the address of a node is given. The
Tn address gives a node a definite position in the tree. Nodes in a tree may be
identified by a numerical index ranging over 0 and 1 ( Cann et al. 2005: 38). The
address of the current root node is Tn(0). In a DS tree, the daughter on the left hand
side is decorated with an argument formula, and daughter node on the right is
decorated with a functor formula. The value of any argument daughter is ‘0’
appended to the Tn value of its mother. And the value of any functor daughter is ‘1’
appended to the Tn value of its mother. Thus, the address of the root node’s
argument daughter is ‘00’ and its functor daughter is ‘01’.
(2) INTRODUCTION
{…{…? Ty (Y) …◊}…}
{…{…?Ty(Y),? <↓0> Ty(X), ? <↓1> Ty (X → Y),…◊}…}
The Introduction rule, adds requirement for an argument and functor daughter
decorated with expressions of a certain type, which combined can give the
expression required by the original goal. <↑> ,<↓> are basic modalities, <↓>
corresponding to the daughter relation ‘down’ and <↑> corresponding to the
mother relationship ‘up’. In the rule given above, <↓0> indicates an immediately
dominated node on the left-‐hand side, the argument daughter and <↓1> indicates an
immediately dominated note on the right hand side, the functor daughter. After
applying the introduction rule we get:
(3) Tn(0),?Ty(t), ?<↓0>Ty(e),?<↓1> Ty( e → t), ◊
The tree in (3) has not actually grown the daughter nodes, it has merely acquired
requirements to have such nodes. Hence, the tree or the input consists of only one
node. An expression of type e is a term that denotes some entity. The functor type is
expressed as (e → t), which denotes that it is a one-‐place predicate and when
combined with a term e, it yields a proposition type t. Next, the rule of Prediction is
applied, which builds the required nodes and leave the pointer at the argument
node.
(4) PREDICTION
{Tn(n),…,?<↓0>φ,?<↓1>ψ,◊}
{Tn(n),…? <↓0>φ,?<↓1>ψ}, {<↑0>Tn(n),?φ,◊}{<↑1>Tn(n),?ψ}
(4a) Tn(0),?Ty(t),?<↓0>Tye(e), ?<↓1> Ty (e → t)
Tn(00), ?Ty(e), ◊ Tn(01), ?Ty (e → t)
Note that instead of using the modalities <↓0 ,↓1> in the daughter nodes, in (4a), I am
using the Tn to identify a node by its numerical index (Tn(00), Tn(01)). Thus
introduction licenses the introduction of modal requirements and prediction
transforms them into non-‐modal type requirement, by constructing the appropriate
nodes. These rules are the most general rules to unfold tree structure (Cann et al.
2005:43).
Next we need information from the lexicon to continue with the parsing process.
Lexical items provide information about node annotation and give tree building
instructions. Once the hearer encounters Joel, he scans the lexical information
provided by the expression and adds it into the derivation. Lexical entries in DS are
represented as conditional statements (Cann et al. 2005: 45). IF introduces the
requirements. If the requirement is true, further instructions are given by THEN. If
either the IF and THEN are false, than ELSE specifies what to do next. If ELSE gives
the ABORT instruction, then the parse fails and stops there. The consequent of the
initial condition ‘make’ gives instruction to make new nodes, ‘go’ moves the pointer
to the node specified in the value and ‘put’ annotates a node with certain
information (Canne et. al. 2005:45). The lexical entry for Joel is given in (5):
(5) Joel IF ?Ty(e)
THEN put (Ty(e),Fo(Joel’),[↓]⊥)
ELSE ABORT
Proper name in DS are of type (e) and are required to be a current requirement
?Ty(e) at the stage at which lexical entry is scanned. The pointer was left at the
argument node in (4a) which holds a requirement for Ty(e), after we applied
introduction and prediction rules. Therefore, the condition imposed by lexical entry
for Joel is satisfied and the node is decorated with Ty(e),Fo(Joel’) and the bottom
restriction [↓]⊥. The bottom restrictions indicate that the word decorates a terminal
node, which means that the current node has and will have no daughters with any
properties at all (Cann et al. 2005:46). If at the current node, there was no
requirement for a Ty(e), the phrase would have been Aborted. Following is the
resulting tree structure:
(6) Tn(0),?Ty(t), ?<↓0>Ty(e), ? <↓1> Ty(e→t)
Tn(00), ?Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥,◊
In (6) we see that the current node has both the fact and the requirement to fulfill
the fact. Thus, we apply ‘Thinning’ rule, which eliminates the fulfilled requirement
(Cann et. al. 2005:49).
(7) THINNING
{…{…,φ…,?φ,…,◊}…}
{…{…,φ,…,◊}…}
After applying Thinning we get:
(7a) Tn(0),?Ty(t), ?<↓0>Ty(e), ? <↓1> Ty(e→t)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥,◊
Now the ‘Completion’ rule moves the pointer up to the mother node and annotates it
with the information that it has a daughter of a certain type.
(8) COMPLETION
{Tn(n)…}, {<↑i>Tn(n),…, Ty(X), …,◊ }
{Tn (n),….<↓i>Ty(X),…, ◊}, { }
i ∈ {0,1 *}
When we apply the completion rule, we get the structure in (8a). In the root note,
the decoration <↓0> Ty(e) is added, which shows us that this node has an argument
daughter of type e.
(8a) Tn(0), ?Ty(t), ? <↓0>Ty(e), <↓0> Ty(e),? <↓1> Ty(e→t), ◊
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥
Now the requirement, that there is an argument daughter of Type e is fulfilled. We
again apply the Thinning rule and get the following structure:
(8b) Tn(0), ?Ty(t), ? <↓0>Ty(e),? <↓1> Ty(e→t), ◊
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥
“The Anticipation rule moves the pointer down from a mother node to a daughter
node which has an outstanding requirement.” (Cann et al. 2005:51)
(9) ANTICIPATION
{Tn(n),…,◊},{<↑>Tn(n),…,?φ…}
{Tn(n),…{ <↑>Tn(n),…,?φ…},…,◊}
(9a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t)
Tn(00), Ty (e) Tn(01), ?Ty (e→t), ◊
Ty (e), Fo (Joel’), [↓]⊥
Now that the pointer has been moved by the anticipation rule, to a node decorated
with requirement type (e→t). Thus, now its time to parse helped. Below is the lexical
entry for help.
(10)
help IF ?Ty (e→t)
THEN go(<↑1>?Ty(t)),
put (Tns(PAST)),
go(<↓>?Ty(e→t)),
make (<↓1>),
put (Fo(help’), Ty(e→(e→t), [↓]⊥;
go(<↑1>);
make (<↓0>);
go (<↓0>);
put(?Ty(e))
ELSE ABORT
In (9a) the pointer is at a node decorated with a requirement for Ty (e→t) and thus
the condition imposed by the lexical entry of the verb help is satisfied. As a result,
the pointer is instructed to go up a functor relation to the node which holds a
requirement for an expression of Ty(t) and decorate it with tense. Then the pointer
is instructed to go down the same relation, build a new functor node from there and
annotate it with the specified formula, type values and the bottom restriction. Now
the pointer is instructed to go up that functor relation and build from there a new
argument node, decorated with a requirement for Ty(e). This is where the pointer is
left after parsing helped. Next, this transitive verb builds two nodes, one for a two
place predicate and other node for the internal argument. This is illustrated in the
following tree:
(10a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010)?Ty(e), ◊ Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
The pointer is at the argument node, decorated with Ty(e), thus, another proper
name – Miriam, can be parsed. Miriam has the same lexical entry as Joel and the
same instructions are followed. After the lexical entry is parsed, we decorate the
node with the information provided by the lexical entry and next we apply Thinning
again to eliminate the fulfilled requirement.
(11) Miriam IF ?Ty(e)
THEN put (Ty(e),Fo(Miriam’),[↓]⊥)
ELSE ABORT
(11a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥,◊
According to our sentence Joel helped Miriam who he knows dislikes syntax, next the
relative pronoun who is encountered. In DS, when the pointer encounters a relative
pronoun, a new tree is constructed and node of the host tree is linked to the root
node of the second tree. This linking is done through a Link relation <L> and its
inverse <L-‐1>, introduced by the rule of Link Adjunction.
(12) LINK ADJUNCTION
{…{Tn(α),Fo(A),Ty(e), ◊}…}
{…{Tn(α),Fo(α),Te(e)}…}, {<L-‐1>Tn(α),?Ty(t), ?<↓*>Fo(α),◊}…}
This rule specifies that the linked node consists of three decorations: < L-‐1>, which
expresses that it is an inverse link relation to the head, which is indicated by the Tn
value; it is annotated with a requirement for propositional content -‐ ?Ty(t);
?<↓*>Fo(α) expresses that within the link structure there should be somewhere
down a copy of the formula that decorates the head node, from which the head
structure is built. Thus, encountering who, makes us build a link node and moves the
pointer to that node.
(12a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>For(Miriam’), ◊
Now the *Adjunction rule is applied, which introduces an unfixed node and leaves
the pointer there. The unfixed node cannot remain unfixed, so to find a fixed
position, it is decorated by the requirement ?∃x.Tn(x). We apply this rule right after
building a linked node to reflect the expectation to find a copy of the formula
somewhere within the tree. Following is the *Adjunction rule and (13a) is the
structure, the rule leaves us with.
(13) *ADJUNCTION RULE:
{…{Tn(α),...,?Ty(t), ◊}…}
{…{Tn(α),…,Te(t)}…}, {<↑*>Tn(α),?∃x.Tn(x),…,?Ty(e),◊}…}
(13a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>Fo(Miriam’)
<↑*><L-‐1>Tn(010), ? ∃x.Tn(x),?Ty(e), ◊
Now we can parse who. And to do so, we first provide the lexical entry for the relative pronoun who.
(14)
whorel IF ?Ty(e),∃x.Tn(x),
<↑*><L-‐1>Fo(x)
THEN IF
THEN
ELSE
↑⊥
ABORT
Put(Fo(x),Ty(e), [↓]⊥
ELSE ABORT
The conditional statement in the lexical entry states that up the unfixed node there
is a LINK relation defined to some node decorated by a formula Fo(x), if however the
pointer is at a fixed node (↑⊥), then the parse fails, if not, then a copy of the formula
needs to be provided for the current node and ty(e) and [↓]⊥ need to be annotated.
Thinning also applies and we get the following tree structure:
(14a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>Fo(Miriam’)
<↑*><L-‐1>Tn(010), ? ∃x.Tn(x),?Ty(e), Fo(Miriam’)◊
Next Completion, introduction and prediction is applied and we get the following:
(14b) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>Fo(Miriam’)
<↑*><L-‐1>Tn(010), ?Ty(e), ◊ ?Ty(e→t)
? ∃x.Tn(x),?Ty(e),
Fo(Miriam’), [↓]⊥
Next, the pronoun he has to be parsed. DS claims that pronouns are placeholders for
a logical expression constructed within the context of the utterance (Cann et al.
2005:68). A Pronoun’s Fo value is indicated by a metavariable, which is latere
replaced by Substitution with a comparative Fo value from the context (Cann et al.
2005:69).
(15)
he IF ?Ty (e)
THEN put (Ty(e),
Fo(Umale’),
? ∃x.Fo(x),
?<↑0>(Ty(t)∧∃x.Tns(x)),
[↓]⊥)
ELSE ABORT
The lexical rule decorates the node with ? ∃x.Fo(x), a requirement to find the
formula value and ?<↑0>(Ty(t)∧∃x.Tns(x)) to indicate the case condition. After
parsing the decorations as instructed in the lexical entry for he and applying
Thinning we get (15a):
(15a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>Fo(Miriam’)
<↑*><L-‐1>Tn(010), Ty(e) ?Ty(e→t)
? ∃x.Tn(x),?Ty(e), Fo(Umale’)
Fo(Miriam’), [↓]⊥ ? ∃x.Fo(x),
[↓]⊥,◊
Next, we apply Substitution. The metavariable will be substituted with Joel’ as it is
the only compatible Fo value satisfying male’ presupposition. After substituting the
metavariable, the value of Fo(Joel’) allows the requirement ? ∃x.Fo(x) to be thinned.
(15b) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t),? <↓*>Fo(Miriam’)
<↑*><L-‐1>Tn(010), Ty(e) ?Ty(e→t)
? ∃x.Tn(x),?Ty(e), Fo(Joel’),
Fo(Miriam’), [↓]⊥ [↓]⊥,◊
Next the pointer will move to the functor node through Completion and
Anticipation. Thus, we can parse knows. The verb to know can either take an
expression of Ty(e) (Joel knows him/John), or it can take an expression of Ty(t) (Joel
knows she dislikes syntax). For this assignment, the second type (Ty(t)) is the
relevant one, thus only lexical entry for know with expression of Ty(t) will be
provided.
(16)
know IF ?Ty (e→t)
THEN go(<↑1>?Ty(t));
put(Tns(Pres));
go(<↓0>?Ty(e→t));
make(<↓1 >),fo(<↓1>);
put(Ty(t→( e→t)),Fo(know’), [↓]⊥)
go(<↑1>);
make(<↓0>);
go(<↓1>);
put(?Ty(t))
ELSE ABORT
The verb know takes as its object an expression of Ty(t). It builds an argument
daughter of type t and a functor daughter of type (t→( e→t)). When both the
argument daughter and the functor daughter come together, we get an expression of
ty(e→t) . Thus, satisfying the requirement present for the verb know of ty(t). After
parsing know, the pointer is left at the argument daughter holding a requirement
type (t). Then, Introduction and Prediction are applied once again leaving the
pointer at a node with a requirement for type e.
(16b) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e),? <↓*>Fo(Miriam’), Tns(PRES)
<↑*><L-‐1>Tn(010), Ty(e) ?Ty(e→t)
? ∃x.Tn(x),?Ty(e), Fo(Joel’),
Fo(Miriam’), [↓]⊥ [↓]⊥
?ty(t), Ty(t→( e→t))
?<↓0>Ty(e) Fo(know’), [↓]⊥
? <↓1> ty(e→t),
?Ty(e), ◊ ?Ty(e→t)
The requirement for a Ty(e) at the current node matches the type of the unfixed
node and hence they both can be merged. Thinning is applied after the unification of
two nodes ({<↑*><L-‐1>Tn(010) ? ∃x.Tn(x),?Ty(e)Fo(Miriam’), [↓]⊥, ?Ty(e)}).
Thinning eliminates the type requirement and also removes the requirement, which
was first needed to find a fixed position, as it is currently at a position with defined
position in the tree. The Tree after Merge and Thinning is shown in (17).
(17) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e),? <↓*>Fo(Miriam’), Tns(PRES)
Ty(e) ?Ty(e→t)
Fo(Joel’), [↓]⊥
?ty(t), Ty(t→( e→t))
?<↓0>Ty(e) Fo(know’), [↓]⊥
? <↓1> Ty(e→t),
<↑*><L-‐1>Tn(010) ?Ty(e→t)
?Ty(e),Fo(Miriam’), [↓]⊥, ◊
Next Completion is applied and the pointer moves up to the mother node and
Annotates the node that it has a daughter of Ty(e). Thinning then removes the
fulfilled requirement and through Anticipation the pointer moves down to the
daughter with an outstanding requirement. Now, dislikes is parsed. Below (18) is the
lexical entry for dislike.
(18)
dislike IF ?Ty (e→t)
THEN go(<↑1>);
put(Tns(Pres));
go(<↓1>);
make(<↓1 >);
put (Fo(know’),(Ty(e→( e→t)), [↓]⊥)
go(<↑1>);
make(<↓0>);
go(<↓0>);
put(?Ty(e))
ELSE ABORT
The pointer is now at a node, that holds a requirement for Ty(e→t). After the lexical
entry for dislike is given, a series of actions take place mentioned in the lexical entry.
A functor daughter is built and annotated with Fo, Ty values and the bottom
restriction. Also and argument daughter is built and decorated with a requirement
for ty(e) where the pointer is left. The tree is illustrated in (18a)
(18a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e),? <↓*>Fo(Miriam’), Tns(PRES)
Ty(e) ?Ty(e→t)
Fo(Joel’), [↓]⊥
?ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
? <↓1> Ty(e→t),
<↑*><L-‐1>Tn(010) ?Ty(e→t)
Ty(e),Fo(Miriam’), [↓]⊥
?Ty(e), ◊ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
Now lexical entry for Syntax is parsed (19). Then information from the lexicon is
added to the tree (19a)and Thinning and Completion apply.
(19) Syntax IF ?Ty(e)
THEN put (Ty(e),Fo(syntax’),[↓]⊥)
ELSE ABORT
(19a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e),? <↓*>Fo(Miriam’), Tns(PRES)
Ty(e) ?Ty(e→t)
Fo(Joel’), [↓]⊥
?ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
? <↓1> Ty(e→t),
<↑*><L-‐1>Tn(010) ?Ty(e→t), <↓0>Ty(e), ◊
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
Next Elimination takes place. Elimination rule takes the information from the two
daughters of a mother node, performs functional application and annotates the
mother with the result in (20a) (Cann et al. 20005:45).
(20) ELIMINATION
{…<↓0>(Fo(α),Ty(Y)), <↓1>(Fo(β),Ty(Y→X)),…, ◊ }
{...Fo(β(α)),Ty(X), <↓0>(Fo(α),Ty(Y)), <↓0>(Fo(β),Ty(Y→X)),…, ◊ }
(20a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e),? <↓*>Fo(Miriam’), Tns(PRES)
Ty(e) ?Ty(e→t)
Fo(Joel’), [↓]⊥
?Ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
? <↓1> Ty(e→t),
<↑*><L-‐1>Tn(010) ?Ty(e→t), Ty(e→t),Fo(dilsike’(syntax)), ◊
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
In (20a), where the pointer is, at that node the requirement for a Ty (e→t) is
satisfied and Thinning can be applied followed by Completion, Elimination and
Thinning again. Thus we get the following tree in (21) after we go through all the
steps mentioned above, we reach up until the top of the linked tree.
(21) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e), <↓1>Ty(e→t),? <↓*>Fo(Miriam’), Tns(PRES),
Fo (know’(dislike’(syntax’)(Miriam’)(Joel’)), ◊
Ty(e) Ty(e→t),<↓1>Ty( t→( e→t)),< ↓0>Ty(t),
Fo(Joel’), [↓]⊥ Fo(know’(dislike’(syntax’)(Miriam’))
Ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
<↓1> Ty(e→t), Fo(dislike’(syntax’)(Miriam’)
<↑*><L-‐1>Tn(010) Ty(e→t), Ty(e→t),Fo(dilsike’(syntax))
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
Through a revised rule of completion, the pointer can be moved back to the head
node allowing for the host tree to be completed.
(22) COMPLETION (Revised)
{…{Tn(n)…}, {µ-‐1} Tn(n),…,Ty(X),…,◊}…}
{…{Tn(n),…,<µ>Ty(X),…,◊},{< µ-‐1>Tn(n),…,Ty(X),…,}…}
µ-‐1 ∈ {↑0, ↑1, ↑*, L-‐1} µ ∈ { ↓0, ↓1, ↓*, L}
(22a) Tn(0), ?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST)
Tn(00), Ty (e) Tn(01), ?Ty (e→t)
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥,◊
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e), <↓1>Ty(e→t),<↓*>Fo(Miriam’), Tns(PRES),
Fo (know’(dislike’(syntax’)(Miriam’)(Joel’))
Ty(e) Ty(e→t),<↓1>Ty( t→( e→t)),< ↓0>Ty(t),
Fo(Joel’), [↓]⊥ Fo(know’(dislike’(syntax’)(Miriam’))
Ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
<↓1> Ty(e→t), Fo(dislike’(syntax’)(Miriam’)
<↑*><L-‐1>Tn(010) Ty(e→t), Ty(e→t),Fo(dilsike’(syntax))
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
We again apply Completion, Elimination and Thinning and make our way up to the
root node in (23)
(23) Tn(0),?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST),Fo(help’(Miriam’)(Joel’)), ◊
Tn(00), Ty (e) Tn(01), ?Ty (e→t), <↓0>ty(e),Fo(help’(Miriam’))
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e), <↓1>Ty(e→t),<↓*>Fo(Miriam’), Tns(PRES),
Fo (know’(dislike’(syntax’)(Miriam’)(Joel’))
Ty(e) Ty(e→t),<↓1>Ty( t→( e→t)),< ↓0>Ty(t),
Fo(Joel’), [↓]⊥ Fo(know’(dislike’(syntax’)(Miriam’))
Ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
<↓1> Ty(e→t), Fo(dislike’(syntax’)(Miriam’)
<↑*><L-‐1>Tn(010) Ty(e→t), Ty(e→t),Fo(dilsike’(syntax))
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
Next we apply Link Evaluation rule. This rule conjuncts the Fo value of the
propositional tree (Tn(0)) and the Fo value of the propositional structure linked to
the Tn(010) node.
(24) LINK EVALUATION1 (Non-‐restrictive construal)
{…(Tn(α),Fo(φ),Ty(t),◊}},{<L-‐1>MOD(Tn(α)),…Fo(ψ),Ty(t)}
{…{Tn(α),…,Fo(φ ∧ ψ),Ty(t), ◊}}, {< L-‐1>MOD(Tn(α)),…,Fo(ψ),Ty(t)}
MOD ∈ {<↑0>, <↑0>}*
With this final rule, it is now possible to express the content of Joel helped Miriam,
who he knows dislikes syntax as a conjunction of the propositional formulae (25).
(25) Tn(0),?Ty(t), <↓0>Ty(e),? <↓1> Ty(e→t),Tns (PAST),
Fo(help’(Miriam’)(Joel’))∧ know’(dislike’(syntax’)(Miriam’)(Joel’)), ◊
Tn(00), Ty (e) Tn(01), ?Ty (e→t), <↓0>ty(e),Fo(help’(Miriam’))
Ty (e), Fo (Joel’), [↓]⊥)
Tn(010),Ty(e), Tn(011),Ty (e→(e→t), Fo(help’), [↓]⊥
Fo(Miriam’), [↓]⊥
<L-‐1>Tn(010),Ty(t), <↓0>Ty(e), <↓1>Ty(e→t),<↓*>Fo(Miriam’), Tns(PRES),
Fo (know’(dislike’(syntax’)(Miriam’)(Joel’))
Ty(e) Ty(e→t),<↓1>Ty( t→( e→t)),< ↓0>Ty(t),
Fo(Joel’), [↓]⊥ Fo(know’(dislike’(syntax’)(Miriam’))
Ty(t), Ty(t→( e→t))
<↓0>Ty(e),Tns(Pres) Fo(know’), [↓]⊥
<↓1> Ty(e→t), Fo(dislike’(syntax’)(Miriam’)
<↑*><L-‐1>Tn(010) Ty(e→t), Ty(e→t),Fo(dilsike’(syntax))
Ty(e),Fo(Miriam’), [↓]⊥
Ty(e), Fo(syntax’), [↓]⊥ Fo(dislike’), Ty(e→(e→t)), [↓]⊥
BIBLIOGRAPHY:
• Cann, R, Kempson, R & Marten, L (2005). The Dynamics of Language. Oxford: Elsevier.