• seminar

seminar

SEMINAR

Can names be translated, or must they be borrowed?
Andrew Gargett
Saarland University

We might be tempted to answer 'borrowed', if our account presumed that
names were a well-defined class of linguistic objects, largely empty
of content, yet linked to (potentially rich) associated information
about some individual name-bearer (e.g. Anderson 2004). Such an
account would be consistent with the intuitively appealing, and
well-known idea that names are rigid designators (Kripke 1981), which
is to say, labels designating once and for all some unique
object/entity. On the other hand, the cross-linguistic diversity in
forms of person reference (e.g. possessed kin terms rival personal
names as the default form of naming across languages, Stivers et al.
2007), suggests that languages offer a much wider range of strategies
for naming than sometimes assumed, suggesting some potential for
translatability. Indeed, forms of referring in linguistic interaction
are far more dynamic and complex than orthodox accounts of names seem
able to deal with, with a wide variety of alternative forms of entity
reference, and choices between such alternative forms often driven by
context-specific factors (even being conditioned by such
partner-specific information as levels of expertise, e.g. Isaacs &
Clark 1987).

All of this suggests the need for a more general account of person
reference, grounded in actual situations of language use. Blending
ideas from anthropological linguistics (Enfield and Stivers 2007),
cognitive psychology (Valentine et al. 1996, Isaacs & Clark 1987), and
formal linguistics (von Heusinger 2004, McCawley 1996), in particular,
a dynamically oriented model of grammar, Dynamic Syntax (Cann et al.
2005), Gargett (2010) presented a fresh take on this: (i) that
enriching these underspecified forms of person reference is a
potentially unbounded process, drawing on either immediate
(linguistic) context or else longer term (cognitive) context (such as
images, personal recollections, and the like), and (ii) that
enrichments may be stored and reused on subsequent occasions for
referring to the same individual, leading to the idea that names are
"routine" designators. This talk will seek to further motivate this
proposal by way of examining patterns across languages of switching
between alternative forms of naming (e.g. during dialogue).

Date: 2010-08-24 13:15 - 15:00

Location: T340, Gamla hovrätten

Permalink

SEMINAR

Kalervo Järvelin (Tampere): Managing Morphologically Complex Languages in Information Retrieval

English is a morphologically simple, phrase-oriented language. Its treatment in mono- and cross-language IR is fairly simple, save for possible phrase recognition. Over the years, much software has been developed for English IR. However, many other languages have become, or are becoming, essential for IR for example in the Web. All of these languages are not as simple as English but may lack many computational resources available for English. The main challenges on IR posed by many languages relate to (a) lexicons: out-of-vocabulary words used in texts but not listed in dictionaries; (b) morphology: word inflection and compounds. The talk discusses methods for managing these challenges in mono-lingual and cross-language IR in morphologically complex languages both at the indexing stage and the search stage. The methods include stemming, lemmatization, compound splitting, compound translation, inflectional form generation, structured queries, and approximate string matching.

Date: 2010-09-02 10:15 - 12:00

Location: L307, Lennart Torstenssonsgatan 8

Permalink

SEMINAR

Andrew Gargett (Saarland University): Routines, context and figurative language in dialogue

Abstract: This talk presents ongoing work investigating the use of grammars for modelling core dialogue phenomena (clarifications, reformulations, corrections, and the like), in particular, employing the grammar formalism of Dynamic Syntax (DS, Kempson et al. 2001). I will first outline an extension of the DS account of dialogue (e.g. Purver et al. 2006) to explain routinisation via a dual process model. Then I will present recent work on extending this to account for the interaction between the linguistic and conceptual levels, focusing on the use of figurative language to resolve uncertainty in dialogue (e.g. Bavelas et al. 2008). Now, if we presume that the linguistic level reflects an underlying conceptual model, then the emergence of linguistic routinisation over the course of interaction reflects routinised conceptualisation of the task at hand. So it is interesting to consider how interlocutors might respond when difficulty arises with how a task is being conceptualised (difficulty may be environmental, such as restricted visual cues as when on the telephone, or interactional, such as when an interlocutor misunderstands a partner's contribution). I will suggest that, faced with such difficulty, figurative language can be used for re-conceptualisation, as a way of shifting the conceptual ground of the current focus of activity, and consequently, as a strategy for re-grounding understanding of the interaction. The resulting formal model invites reconsideration of what has been termed the "groundedness" of figurative language (Gibbs 2006), in terms of its role in dialogue. Along the way, I connect DS to cognitive linguistics and artificial intelligence, in particular, demonstrating how combining the DS grammar formalism with ATT-Meta, an AI system for modelling the use of figurative language as reasoning-by-simulation (e.g. Barnden 2009), goes some way toward modelling how interlocutors can employ conceptual metaphor to cope with uncertainty.

Date: 2010-08-26 10:15 - 12:00

Location: L307, Lennart Torstenssonsgatan 8

Permalink

SEMINAR

Abstract:

In dialogue we usually neither spell out our arguments, nor present every step necessary to reach a logical conclusion or justify a proposal. These semi–logical arguments, which are fundamental to rhetoric, are what Aristotle calls enthymemes. The few linguistic studies that  have been carried out on the topic indicate that enthymemes also play an important role in spontaneous dialogue. However, there are no precise accounts of what that role actually is.

We would like to provide such an account, and thereby add a rhetorical dimension to dialogue semantics. We hope to show that an analysis taking the enthymeme as its point of departure elucidates some aspects of meaning in spoken language that cannot be explained in terms of, for example, questions, as has been suggested by Ginzburg et al, who introduced the idea that questions under discussion  provide coherence in dialogue and drive it forward (cf. 1).

First, we give some background on enthymemes in general and enthymemes in dialogue illustrated by dialogue examples. Then we move on to discuss how enthymemes can play a role in explaining why we need informational redundancy in dialogue and how the enthymeme can help us choose which apparently redundant utterances we should add to make a dialogue run smoothly (cf. 2 & 3).

Finally, we give a tentative formal account of the enthymeme using an information state update approach, where an agent’s information state is described after each dialogue move (cf. 4). We will also discuss the idea of enthymemes as instantiations of topoi which an agent involved in dialogue has at its disposal. These could be regarded as a rhetorical resource, similar to the way grammatical and lexical competence has been described as resources available to an agent, as envisaged by Cooper & Ranta (cf. 5).

 
(1) Ginzburg, J. The Interactive Stance: Meaning for Conversation. 2010, http://www.dcs.kcl.ac.uk/staff/ginzburg/papers-new.html

(2)  Walker, M. The effect of resource limits and task complexity on collaborative planning in dialogue.  Artificial Intelligence 85. 1996.

(3) Breitholtz, E. & Villing, J. Can Aristotelian Enthymemes Decrease the Cognitive Load of a Dialogue System User? Proceedings of LonDial 2008, the 12th SEMDIAL workshop. 2008.

(4) Larsson, S. & Traum, D. The Information State Approach to Dialogue Management. p. 325-353 Smith & Kuppevelt (eds.): Current and New Directions in Discourse & Dialogue. Kluwer Academic Publishers. 2003.

(5) Cooper, R & Ranta, A. Natural Languages as Collections of Resources. p. 109-120 Language in Flux: Dialogue Coordination, Language Variation, Change and Evolution. Eds. Cooper, R. & Kempson, R. 2008.

Date: 2010-11-04 10:15 - 12:00

Location: L307, Lennart Torstenssonsgatan 8

Permalink

SEMINAR

In psychology, cognitive models, i.e. computational models of the human mind, are used for describing human cognition. In anthropomorphic artificial intelligence, they are used for the purpose of problem-solving in artificial intelligence (AI). The basic idea is this:
(i) start with a search-based AI algorithm 
(ii) use psychological experiments and cognitive models for approximating the set of solution candidates that are accessible to the human mind, the anthropomorphic search space
(iii) run the same AI algorithm restricted to the anthropomorphic search space. 
This yields an algorithm for finding anthropomorphic solutions to the original problem. While the original problem may be unsolvable, the anthropomorphic version of it is solvable whenever the anthropomorphic search space is finite. Transposing to the anthropomorphic version of the problem  is adequate whenever there is a boundary condition on the problem that the solutions produced must be psychologically simple. 
I will show how anthropomorphic methods can be used for solving problems in a wide range of areas including formal verification, number progression analysis, progressive matrix analysis, and natural language generation. I will also discuss potential applications to machine generated music and art.

Date: 2010-06-11 10:15 - 12:00

Location: T219 (Gamla hovrätten)

Permalink

SEMINAR

We distinguish two kinds of information flow in semantic interpretation: that associated with compositional  interpretation (combining the meaning of syntactic constituents) and threading (the passing of information from left to right and top down in a manner related to dynamic semantics).  We shall argue that both kinds are relevant to sentence semantics and also possibly to discourse semantics.  However, for dialogue only a threading or dynamic update flow seems appropriate.  The argument has to do with the fact that dialogues do not have semantic content in the way that sentences or discourses do and that many of the interesting semantic phenomena in dialogue are not best explained by a compositional approach.

Date: 2010-05-28 10:15 - 12:00

Location: T116, Gamla Hovrätten, Olof Wijksgatan 6

Permalink

SEMINAR

Statements about functional dependence can be added to first-order logic with a simple game theoretic semantics. But there is also a model theoretic semantics based on "teams" instead of assignments. An investigation of the logical constants in this model theoretic setting has just begun. I will give some background on dependence logic and some on logical constants, and also say something on what the future might bring.

Date: 2010-04-16 10:15 - 12:00

Location: T219, Gamla Hovrätten (FLoV), Olof Wijksgatan 6

Permalink

SEMINAR

Purver and Ginzburg introduce the
Reprise Content Hypothesis (RCH) and use it to argue for a non-generalized
quantifier approach to certain quantifiers.  Here we will contrast
their approach with an approach which employs a more classical
generalized quantifier analysis and examine what predictions it has
for possible clarifications and reexamine the data which Purver and
Ginzburg present in the light of this.

Date: 2010-03-19 10:15 - 12:00

Location: T116, Gamla Hovrätten (FLoV), Olof Wijksgatan 6

Permalink

SEMINAR

The first in a new series of seminars on Logic and Language Technology.  It will consist of an overview and discussion of logic's role in CLT and a number of short presentations of relevant work that is going on or could be planned for the future.  See http://clt.gu.se/page/llt-seminars

Date: 2010-02-19 10:15 - 12:00

Location: T116, Gamla Hovrätten (FLoV), Olof Wijksgatan 6

Permalink

SEMINAR

I have been working on Urdu Resource Grammar in GF for some time now, I would like to give a talk on my work in the seminar.

Date: 2010-03-18 10:30 - 11:30

Location: room L307, Lennart Torstenssonsgatan 8

Permalink

X
Loading