The workshop will take place at the “Salle du conseil de l’IRIT” (IRIT2 building, 2nd floor—1er étage en français). Here are some access instructions.
Note that you will need to go to the reception desk and ask for the Polymnie meeting as access to Irit is not free.
|10:00-10:15||Nicholas Asher, Sylvain Pogodalla||Welcome|
|10:15-11:05||Aleksandre Maskharashvili, Sylvain Pogodalla||Building Semantic Dependencies from TAG with Tree Transducers and ACG|
|11:05-11:55||Jérôme Kirman, Sylvain Salvati||Computational complexity of commutative grammars|
|14:00-14:50||Richard Moot||First-Order Linear Logic As a General Framework for Logic-Based Computational Linguistics|
|14:50-15:40||Jérôme Kirman, Sylvain Salvati||A logical approach to language description|
|16:00-16:50||Christian Retoré||Lexique génératif montagovien, réseaux lexicaux et préférences|
|09:30-10:20||Chloé Braud||Identification automatique des relations discursives “implicites” à partir de données annotées et de données brutes|
|10:20-11:10||Laurence Danlos, Philippe de Groote, Sylvain Pogodalla||A Type-Theoretic Account of Neg-Raising Predicates in TAG|
|11:10-12:00||Laurence Danlos, Aleksandre Maskharashvili, Sylvain Pogodalla||G-TAG and ACG|
|14:00-14:50||Juliette Conrath||Extraction non supervisée de relations sémantiques lexicales|
|14:50-15:40||Julie Hunter, Laurence Danlos||Evidentials in Discourse|
|16:00-16:50||Philippe de Groote||Discourse Context and Conversational Background: some experiments|
Richard Moot: First-Order Linear Logic As a General Framework for Logic-Based Computational Linguistics
Since its introduction in 1958, the Lambek calculus has served as an inspiration for using logic to give a simple and elegant treatment of natural language. However, the problems with the Lambek calculus are well-known: Lambek grammars generate only the context-free languages, and, though the syntax-semantics interface of the Lambek calculus is one of its appealing properties, the string-meaning relations which can be defined using the Lambek calculus are too limited.
A large number of extensions to the Lambek calculus have been proposed to remedy these problems. In this talk, I will focus on the tuple-based extensions – which include the Displacement calculus, lambda-grammars, abstract categorial grammars and hybrid type-logical grammars – and I will show that all of these systems can be seen as fragments of first-order linear logic.
This embedding result has several important consequences: first, it allows us to compare the analyses proposed in different frameworks, to see where they agree but also to clarify their formal limitations. Second, it greatly simplifies the proof theory for each of the systems, allowing us to give simple, alternative proofs for many known results. But most importantly, it also allows us to immediately propose new and improved parsing algorithms for each of these frameworks.
Our general research project is to provide an account of parenthetical uses of reports of saying or attitudes in discourse. In a semantically parenthetical use of a report, the content of the embedded clause conveys the main point of the report. In (1), for example, the embedded clause he is out of town (labeled `β’) conveys the main point because its content offers an explanation of Fred’s absence.
(1) [Fred didn’t come to my party.]α Jane said [he is out of town.]β
If the matrix clause does not contribute directly to the explanation of Fred’s absence in (1), it is arguable that only the content of the β-clause contributes to the second argument of the explanatory relation that holds in this example. In terms of Segmented Discourse Representation Theory (SDRT), for example, the relation explanation in (1) should be taken to hold only between α and β. Similarly, the Penn Discourse Treebank (PDTB) would relate only α and β with implicit because.
The annotation and interpretation of reports is complicated for numerous reasons. First, a semantically parenthetical report need not be syntactically parenthetical, so semantically parenthetical uses cannot be predicted based on syntax alone; such uses are highly sensitive to the discourse context. Second, the idea that the embedded clause of a report can serve as the argument for a discourse relation is in direct tension with the semantics of veridical discourse relations. A veridical discourse relation is one which requires its arguments to be true, but truth is not guaranteed for the embedded clauses of reports. A parenthetical report presents its embedded clause as containing the more discourse-relevant content, but the relevance of this content in no way ensures its truth. A final difficulty posed by reports is that their interpretation can be effected by global discourse structure. While the local discourse context might contain features that suggest a parenthetical interpretation of a given report r, this local bit of discourse might be embedded in a larger chunk of discourse with features that count against treating r as parenthetical.
In this presentation, we will consider a specific case of the interaction between reports and discourse structure, namely the interaction of parenthetical reports and contingency connectives. We show, for English and French, that causal and other contingency connectives, in contrast to many other types of connectives, restrict the possible interpretations of reports in their syntactic scope such that these reports cannot have a semantically parenthetical interpretation. We argue that this result has immediate implications for the semantics of causal relations and for the annotation of implicit connectives. In particular, it shows that the assumption, implicit in some work on NLP, that the semantics of explicit connectives can be translated to implicit connectives is not anodyne.
Laurence Danlos, Philippe de Groote, Sylvain Pogodalla: A Type-Theoretic Account of Neg-Raising Predicates in TAG
Neg-Raising (NR) verbs form a class of verbs with a clausal complement that show the following behavior: when a negation syntactically attaches to the matrix predicate, it can semantically attach to the embedded predicate. This paper presents an account of NR predicates within Tree Adjoining Grammar (TAG). We propose a lexical semantic interpretation that heavily relies on a Montague-like semantics for TAG and on higher-order types.
G-TAG is a formalism dedicated to text generation. It relies on the Tree Adjoining Grammar (TAG) formalism and extends it with several specific notions allowing for the construction of a surface form from a conceptual representation. This conceptual representation is independent from the target language. Our goal is to study G-TAG and its specific notions from the perspective given by Abstract Categorial Grammars (ACG). We use the reversibility property of ACG and the encoding of TAG they offer. We show that the key G-TAG notions of g-derivation tree and lexicalization are naturally expressed in ACG. The construction of surface forms can then rely on the general ACG algorithms.
There are some natural languages that allow the free ordering of words or phrases in their sentences while preserving grammaticality. In order to model such sentences, we consider a term algebra with two operators: one that models the usual concatenation, and another that denotes the “free order combination”, that is the concatenation of its arguments in any possible left-to-right order.
Multiple Context-Free Grammars (MCFG) are an extension of context-free grammars, where non-terminals in a derivation are associated with a tuple of strings (rather than a single string). Each production rule then specifies how the strings on the right-hand side are combined to form the strings of the left-hand side.
We consider an extension of MCFG that, instead of strings, produces terms (or contexts) over the previously described algebra, to represent sentences with free order. We describe several natural subclasses of those so-called commutative grammars, and give an account of the computational complexity associated with their (universal) parsing problems.
We propose a modular, high-level approach to the description of grammars. Using existing results from the fields of logic and lambda-calculus, we construct a formalism close in spirit to abstract categorial grammars and model-theoretic syntax. The resulting tool allows us to describe mildly context-sensitive languages in a concise and effective manner. We illustrate this approach with an incremental description of a grammar for fragment French that covers several phenomena, including wh-movement and agreement.
Aleksandre Maskharashvili, Sylvain Pogodalla: Building Semantic Dependencies from TAG with Tree Transducers and ACG
We present an Abstract Categorial Grammar (ACG) account of (Kallmeyer and Kuhlmann, 2012)’s process of transformation of the derivation trees of Tree Adjoining Grammar (TAG) into dependency trees. We make explicit how the requirement of keeping a direct interpretation of dependency trees into strings results into lexical ambiguity. Since the ACG framework has already been used to provide a logical semantics from TAG derivation trees, we have a unified picture where derivation trees and dependency trees are related but independent equivalent ways to account for the same surface–meaning relation.