so first thank you for coming to my presentation
in the first part i'm going to introduce a if you a notion of all
of the problem i want to tackle
in the second part i'm going to show how to model and solve this problem
in linguistic terms
and finally i'm going to implement a solution
in a formal system
so here
we have a very simple example involving a discourse relation that you didn't come to
the party to one out of style so we expected to infer an explanation relation
between the two sentences
obviously this explanation can be made explicit with a discourse connective
jane didn't come to the problem because she one out of style
so in english you have three main categories of discourse connectives the first are coordinating
conjunctions with and all other than subordinating conjunction with because or even though
and obvious instead of the white ball for
so one of the phenomena i'm interested in it what we call the none alignment
of syntactic and discourse structure
so there is an alignment or mismatch
when there is no mapping between the syntactic arguments of a discourse connective and the
arguments of the discourse relation that is lexical
so a very basic example
a concern at the bills because they have only one syntactic arguments
but you have to discourse arguments for the discourse relation
for example blah than jane went to paris
and so this
for those women under we also consider does not structure all
and are often a mobile and of course
but i'm interested in other kind of mismatch so to understand that is first look
at this discourse
jane is very talkative as she told me for example that you had an apple
for access to the there is no mismatch here because an example of j being
very talkative is telling me about breakfast and so
but now
i see a trend is are lucky jane told me for example but he won
the lottery last week now one is an example of friends being very likely is
not change holding anything but that he one of the let's stick to here we
have a mismatch and on the line
yes in this body blocking the discourse argument respond to a corpus of the syntactic
so you have this is actually supporting conjecture
so here no mismatch between the j thing that here in europe
so them becomes less
if
i think you indicated here
but a fragment time for a fragment of parallel data think that you know which
will email
this kind of sentences are
and the light with usually with a mismatch because we
the speaker intends to
express the contrast between going to pair and not going to t
so those high enough examples are discussed
like the intention context
no relation to the penn discourse treebank so this kind of example annotated with that
such mismatch in the pdtb
but it has been question by a hard to support their uniformity hypothesis according to
uniformity there is no such mismatch
according to you do uniformity the them the contrasting for be would be between found
when comparing and for jane think that knowledge
but the reasoning in that
you need if a and b are in contrast you can say that a and
someone saying we are in contrast but the problem in that this kind of preventing
the times in to extend to all the kind of a discourse relation
if an example of a is b
you don't wanna say that an example of a is someone saying be the women
going to follow the uniformity purpose of this work considering kind of mismatches
you have also very similar in a minute with implicit relations
a friend did not come to the party jane told me he was taking care
of his gets
so when you have such a mismatch
you have something and income taxes or according to don't listen rambo in i there
is a commitment of the speaker to once the content of the report
you agree with one jane told it is what we call the religion affected information
but for sensor and after there is no official religion faculty but you expected to
infer
hedge relations or possible explanation and because explanation is a very the correlation that it
may be that it is possible that afraid was taking care six get
i so
i don't know exactly what is true do we
do you like in five when you want to find you necessarily agree with what
jane told you or do you
at least come to their it's possible i don't know exactly so we're going to
talk about that like
and see how we can almost constant
bit of technology so those mismatches in the whole action verbs like to be used
to know all those of people would like to say and tell a these four
and with distinguishing between the parenthetical user but it's okay one predicate is not possible
of the discourse argument so when there is an on the alignment and on the
other hand the and attentional use of a t v so when the predicate is
included in the discourse segments so when there is no that the line
yes very important not all not all connectives are compatible with a session of alignment
for example we are cool lexicalising an explanation you can say that did not come
to the party because he was taking care of its good at you cannot instead
an evidential weight you cannot say predicting outcomes of partly because jane says he was
taken care
and the problem comes from the connective because you can do that with an implicit
relation you can say for a given that comes partly j said he was taking
care of is good
the problem comes from the connective blocking their the infinite shown in fig
so our we're going to mortars so as to the to kind of other we
also is that are distinguished by a high amount so on the one hand you
have sensible a direct orders so they show good syntactic integration tool a measure close
and they provide some time for local information on the intensity in the matrix
for example jane left of the written because you have an appointment this is a
sensible close according to like them
on the other hand you have paris for and the realtors as they are much
less into were integrated to the matrix level and that's function is more to structure
the discourse so they can express concession or some but one relation for interpretation of
the matrix roles for example in eight
fragile estimates for less the rest and even though here not finished eating so you've
got some the concession this is a powerful and real close to record
so she wants many differences in syntax-semantics probability and so on between those two kind
of data is that what we see that another different it was performed so that
are compatible with last year's while sensible one or not
so we already thing that in the previous example but
please consider you know and b goes
so you can you can have
the intangible like in nine so no mismatch for the case the policy
even though jane said he'd and should be one thing that you can have a
mismatch like in we fight change the party even though j said you hadn't finished
writing his research couple
but you can and do that we because so you can have to be like
in ten eight contains the partly because jane said she wanted to talk to him
but you cannot have any and then eventually likely to be fed into the partly
because james as he finished writing is research cable
so how could one expenses so well stream no i tell you audio that are
middle connective a considered and of course not to what i'm going to say that
in fact even conjunctions are not for the structure of the kind of selection mechanism
able syntax but with more the that i'm going to model as an unable to
the y about syntax
we couldn't better use examples of all useful so a connective sse
and so i don't want to postulate any syntactic n b greedy in this kind
of sentence that sentences so that the if there is a mismatch or not
the this syntactic structures same but there is this selection mechanism involved in with the
connective that will explain the different kind of interpretation
yes and obviously this selection mechanism always different constraints force and for every from all
sentences
those can find a
first
a discourse arguments must have been introduced by the corresponding syntactic argument but you cannot
select
anything like when you're discourse connective you need to select something that is related to
you that syntactic arguments and a central connective cannot they compose a closed headed by
a navy so that you to access the report but to parry for one can
so with us to a constraint you can really account for the data we would
seen that before
but a in addition are considered as a something that is related to sequences of
conjunctions
a congestion one b conjunction juicy but i'm not going to dwell
yes so i'm going to implement that it is now i mean in a formal
system called a continuation semantics so it's based on lambda calculus it has been used
to model the syntax-semantics interface and
also some aspects of the semantics pragmatics interface and that's the kind of it we're
going to use right now
so and you are then used to model f one for nominal for
so in well you got a very standard high priced for proper noun like jane
is a function with an argument pete's and individual predicate and j just take that
these two of the consulting
now for context sensitive
so like of the pronoun she
you need some expected the context arguments
and now p is not true of any specific constant a viable it's true and
individual that has to be retrieved from the context
you think what people selection function
and in this case the output context use the same every
for a predefined expressions such as cats
what happened these we introduced by a factor the cat is to this
and we aren't information that x is a cat in the context so it can
be this information is from might be relevant for subsequent enough
i don't and the context c
is transmitted from sentence to sentence for this the discourse of the operator
a very simple example she's with things that we wish additionally get something like that
well
it's quite funny or just
the individual corresponding to she in the complexity
i think on the whole this proposition
so we're going to we use this model of anaphoric expressions of to model the
behaviours discourse connectives to do that's it we're going to add the basic type woman
from the for propositional reference so that matters the different from the g the choose
volume type
because these markers are going to be the arguments of discourse relation
so i consider any sentence describe a propositional market that is given as argument but
can also introduce watermark as for example when there is a number that roles
and so now he's we think is something like that you get this condition e
liable all the proposition law and what happens is based at that
i e is a proposition that use this thing and
and of course because propositions control we need a true a pretty eight that just
a tad bit hes an argument denotes a superposition
is somehow the discourse update it looks like that so what how do you continue
to discourse d with the new sentence s
so what's encountered here but we introduce a new proposition along with played that this
proposition one denotes the true population and that without recognition of s
given this the potential market
for and maybe such a thing
what is important in the we introduce a positional marker you problem that for the
embedded everything that
e interpretation that s
things in prior and we start executing the and the embedded close given prime
we don't like that in practice true because
i think it does not give them so someone can think something that
now the last time
in the this pretty big channel for the because
so what's importantly we have to augment a and b for the two
very first oppositional a quite a that is introduced and passed to the recognition and
then
a propositional a kiwi is introduced and passed to the application of me and then
the main property should he is described as an explanation holding between to do things
to augment that has to be selected
and so you got for instance this term select an appropriate propositional markets in the
context the system
that is really do not have to satisfy the three constrain the this section six
and seven see for a central
a connective so they it needs to satisfy the this we construct a for a
note that connect points and although it's gonna be very similar
the only difference
is going to be the lexicalized discourse relation and the selection function is gonna be
this cell speech section function
that is satisfied only to constrain
so if we consider the sentence print came to the party although jane things will
think what we get it this a big thing
so what happens is we expect that for k and this is okay
then
again
in trying to be a prime
he is think
and we have a conversation between holding between those two times
and in this particular case there is only one
a positional marker related to a in this context so the first argument is going
to be necessarily the case for making the possible
but for this argument
there is actually to a different position long as related to me satisfying the three
that the constraint you have
maybe so anything that was sick and it prior
is applied
which is only a few was so it is but two possibilities are available and
based on the world knowledge this lp algorithm will be able to decide that the
correct one is just a in primes
not all there is less than that with since like you so french state on
because j if you will see
in this case
instead of lp you will have to see so about satisfying different constraint and the
only possibility for those of an argument you think you primes a little or jane
think you will see and in this case it it's and natural interpretation
so we can explain
one
this is correct and this is not
yes
us a lot about this mapping consequences that i mentioned so if we think they
consider a some very elite the actual so before and there exist the true confession
holding between two proposition then the to proposition must be true
if we do that's we get just from where jennifer actually information
this by dollars in mimo
but if we decide to slightly change that sounds for the connective so that they
introduced hedge relations when then selecting the structural argument
we get the interpretation proposed by hands and eyes
so what have been done so we found that the distinction between a parenthetical and
parenthetical you deserve at this is explained in terms of discourse unfolds and that the
distinction between sends one powerful conjunction include specific argument selection constraints of those is thoughtful
and we show that
the average height or discourse connectives can be implemented compositionally using continuation
thank you for your attention and please ask me any questions
so we have to mention that this is very theoretical work so i'm not trying
to actually implement the system that is going to do that because it relies on
like good syntactic analysis and it's very hard to get
and plus if you want to analyze the full discourse you need to have like
to affect syntactic analyses for all the sentences so that the and so it's
not
two days not practically doable but its aim of this work was to explain the
why some connectives a compatible with mismatches wine some not like more linguistic or not
well
so you can is selection function so you can plug any kind of algorithm you
want so probably something related to get the resulting system if you're improve the that's
use
so the competition like the conditional computational aspects are going to be related to what
kind of specific function you implement for this we can do you gotta you could
have a very basic heuristic or input other much texas
yes
so the thing is when
when you like the way it's and this kind of us a discourse on a
light in the literature and i will not have once you infer the explanation relation
you're kind of committing to the actual is a possibility of the other two something
in particular
so
but in this case you are not going to infer the explanation relation or you're
going to attribute this explanation to jane not to use all
actually it's less obvious when you have implicit discourse relation we could you can be
page could be attributed someone else but if we go back to example with for
example like a friend is lucky jane told me for example that he wanted the
last we like we last week it seems very hard to say that she
not an example was lying or so in this case you will kind of committing
i know if you use this kind of children you've prepositional phrases like a according
to j so they behave differently but they are really syntactically parenthetical and so you
can always i whose for any within a connective
yes
but there is there something we're going on when used according to jane because sometimes
it seems like the relation is actually needed to a the like you it the
whole the input the edge between phrase is coping of a discourse connective but sometimes
not it's
a bit more complex
i've been here sorry
and
no well
i know exactly like
what would be according to me difference is like the scoping relations
how it's going to scope of the discourse relation
which is
both because is syntactically complete very different so you expect that the scoping relations are
going to be different and that's probably what