Over the last 50 years, model-theoretic semantics has made great
progress in formalising various phenomena of human language,
especially those related to the compositionality of meaning. It has
also left some aspects largely unexamined: in particular, those
concerned with the meaning of content words, which have been the
concern of lexical semantics. Distributional or vector-space models of
meaning, in turn, successfully model many lexical semantic phenomena,
but, despite recent efforts, still do not account for compositionality
or the role of function words.
This workshop aims to foster the integration of formal and
distributional semantics, building on their complementary strengths to
produce better models of meaning in natural language. Its focus is on
giving formal semantics a better handle on lexical semantics, while
still preserving the aspects that have made it useful: for instance,
the tight syntax-semantics interface for phenomena such as
quantification, scope, modification and semantic roles, the notions of
truth and extension, and the modeling of inference, at the level of
both lexical items and propositions. Beyond the goal of better
embedding lexical information in formal semantic representations, we
encourage the investigation of the probabilistic aspects of
distributional models in order to handle cases where truth theory
falters (e.g., truth of generic statements). More generally, the
workshop is open to new ideas about aspects of meaning beyond the
level of truth values, and on how distributional semantics may
contribute to the notion of intension.
A non-exhaustive list of topics of interest includes:
- Inferential properties of language both at the propositional and
lexical level (e.g., how to account for entailment as licensed by
different types of adjectives?).
- Truth theory, reference, and extension (e.g., to what extent is
there a relation between distributional representations built from
corpora and models, which exhaustively enumerate individuals in a
particular set? How could this be used to tie distributional
representations of words with specific objects in real or virtual
- Intension (e.g., can distributional models provide an
alternative or complementary account to the definition of
intension in terms of possible world semantics?).
- Syntax-semantics interface (e.g., composition with
different types of verbal arguments).
- Morphology-semantics interface (e.g., exploring the possibility
to build distributional models of derivational and inflectional
morphology, for instance to capture the semantic difference between
singular and plural forms of nouns).
- Representation and mechanics of logical operators (e.g., how to account for negation, not only in propositions but also with regard to its effects on the lexicon; how to deal with coordination at the various levels at which it applies; how to deal with nominal and VP quantifiers?).
- Semantic representation at the lexical and constituent /
sentential level (e.g., what kind of representations do we need to
retain lexical information at the level of the sentence?).
Organisers / Contacts
For all questions, please contact the TFDS 2013 organisers using the following email address: email@example.com.
14476 Golm, Germany
Center for Mind/Brain Sciences (CIMeC)
University of Trento, Palazzo Fedrigotti, Corso Bettini, 31
38068 Rovereto (TN), Italy
Fax: +39 0464 808654
Phone: +39 0464 80 8613
The University of Texas at Austin
Department of Linguistics
Calhoun Hall 516A
1 University Station B5100
Austin, TX, USA 78712