Language and Communication Technologies Colloquia

The LCT Colloquia started four years ago as an opportunity for students and researchers to get acquainted with the most recent research and to interact with each other.

This year the program features also several colloquia for larger audience aiming to familiarize non-experts with these new cutting-edges key technologies that are already showing important impact on our daily life and are growing rapidly in the international market.

The Colloquia will take place on Thursdays during the months from November 2007 to May 2008 in either Bolzano (FUB), Rovereto (CIMeC), or Trento (FBK) as specified in the program's overview included below. The colloquia on specialised research topics will be given in English, the others in Italian.

A printable PDF Poster with the schedule of the LCT Colloquia for the period of November-December 2007, and January-May 2008 are available for download:

If you are not in the mailing list of the LCT Colloquia and you want to receive a reminder before every seminar, we invite you to subscribe at the following link: https://mail.inf.unibz.it/mailman/listinfo/lct-colloquia.

For more information, please contact us: lct-info@inf.unibz.it


Programme (November 2007 - May 2008)

November   December   January   February   March   April   May  




November ^

Date Location Speaker Affiliation Title Abstract Slides Comments
November 20 (Tuesday)
17:30-18:30
FBK Achille C. Varzi Columbia University Di che cosa parliamo quando parliamo? Abstract  



Abstracts (November)


November 20 - Aula 04, Facoltà di Lettere e Filosofia, Via Santa Croce, 65 I-38100 Trento

Di che cosa parliamo quando parliamo?
Achille C. Varzi, Columbia University


La risposta sembra ovvia: se usiamo certe parole, stiamo parlando delle cose a cui quelle parole si riferiscono. Se diciamo che Mario è simpatico, allora stiamo parlando di Mario, e se diciamo che qualcuno ha telefonato, allora ci stiamo riferendo a una certa persona, ancorché in termini generici. Sfortunatamente non sempre le cose sono così semplici. Tanto per cominciare, spesso il linguaggio di cui ci serviamo è vago (a che cosa ci riferiamo, esattamente, quando parliamo del "centro di Trento", della "Pianura Padana", della "guerra in Iraq"?) e a volte è ambiguo (come quando diciamo che Mario è andato a vedere un film con Benigni, o che tutti i suoi amici sanno suonare uno strumento musicale). Ma anche quando le parole sono precise e gli enunciati non ambigui, non è detto che il nesso tra ciò che diciamo e ciò di cui parliamo sia chiaro e lineare. Per esempio, parliamo comunemente di "mode", "battute di spirito" e "rigori non concessi" cosi come leggiamo di "scandali", "divergenze di opinioni", "leggi di mercato"; ci lamentiamo della "sfortuna" e delle "incompetenze" altrui così come speculiamo sulla "famiglia media" o sulla "probabilità" che domani piova; diciamo di vedere l’ "orizzonte", di conoscere l’ "età" di certe persone, di aver sentito che il "numero dei disoccupati" è in aumento. È vero che cosi facendo ci capiamo abbastanza bene. Ma non è detto che questi nostri modi di esprimerci siano sempre da prendersi alla lettera, ossia che le espressioni tra virgolette corrispondano sempre a *qualcosa*. Anzi, se non vogliamo incorrere in grossolani abbagli (come accadde a Polifemo), la prima cosa da fare è distinguere bene tra come ci esprimiamo e ciòò che intendiamo davvero, tra espressione linguistica e contenuto proposizionale, tra forma grammaticale e struttura profonda. Ebbene, proprio su questa distinzione si reggono buona parte della filosofia del linguaggio e della linguistica contemporanea. Purtroppo è anche una distinzione che non regge, o così intendo sostenere. Non regge nel momento in cui intendiamo usarla per rispondere alla domanda del titolo. E se le cose stanno così, allora il problema si fa serio, sul piano pratico come su quello teorico.




December ^

Date Location Speaker Affiliation Title Abstract Slides Comments
December 13
17:00-18:00
CIMeC Alistair Knott Otago University, New Zealand Sensorimotor cognition and natural language syntax Abstract  
December 20
17:30-18:30
CIMeC Udo Kruschwitz University of Essex Intranet Search with Automatically Acquired Domain Knowledge Abstract  



Abstracts (December)


December 13, 2007, 17:00-18:00 - Aula 7, Palazzo Fedrigotti (CIMeC), Corso Bettini 31, Rovereto

Sensorimotor cognition and natural language syntax
Alistair Knott, Otago University


In this talk, I will propose a hypothesis that links two different disciplines in cognitive science: sensorimotor cognition and theoretical syntax. The hypothesis is that the syntactic structure of a sentence describing a concrete event in the world provides a direct encoding of the sensorimotor processes via which this event is experienced.

To introduce the hypothesis, I will focus on a single event: that of a man grabbing a cup. This event can be studied from two perspectives. Firstly we can study the sensorimotor processes involved in performing or perceiving this cup-grasping action. I will outline a model in which these processes have a rigid sequential structure, comprising (i) an action of attention to the agent; (ii) an action of attention to the patient; and (iii) the execution of a motor action. Secondly, we can study the structure of sentences which describe the event of a man grabbing a cup: for instance, "The man grabbed a cup", and its equivalent in other languages. I will outline a syntactic model of such sentences, using the Minimalist framework of Chomsky (1995). The structure of the cup-grabbing sentence involves a hierarchy of three phrasal units: (i) a unit assigning case to the agent; (ii) a unit assigning case to the patient; and (iii) a unit introducing the verb. I will argue that the isomorphism between the syntactic structure and the sensorimotor sequence is not a coincidence, and will describe several other syntactic structures where a similar isomorphism appears to exist. I will conclude by proposing that syntactic structure can perhaps be characterised in sensorimotor terms, as an encoding of sequential sensorimotor processes.


December 20, 2007, 17:30-18:30 - Room: TBA, Palazzo Fedrigotti (CIMeC), Corso Bettini 31, Rovereto

Intranet Search with Automatically Acquired Domain Knowledge
Udo Kruschwitz, University of Essex


INSERT ABSTRACT




January ^

Date Location Speaker Affiliation Title Abstract Slides Comments
January 17
17:30-18:30
FUB Ian Pratt University of Manchester Logics with Counting Quantifiers Abstract  
January 31
17:30-18:30
Povo Barbara Di Eugenio University of Illinois at Chicago Generation of effective language feedback for intelligent tutoring systems Abstract dieugenio.pdf 



Abstracts (January)


January 17, 2007, 17:30-18:30 - Faculty of Computer Science, Free University of Bozen-Bolzano, Piazza Domenicani 3, Seminar Room (first floor left)

Logics with Counting Quantifiers
Ian Pratt, University of Manchester


Counting quantifiers are expressions in formal languages having English glosses of the forms ``there exist at least/at most/exactly C objects x such that ... x ...'', where C is a natural number. In the context of first-order logic with equality, adding counting quantifiers results in no increase in expressive power, since they can all be defined by means of universal or existential quantifiers in the familiar way. In the context of many *fragments* of first-order logic, however, adding counting quantifiers leads to a considerable increase in expressive power, and it is often of interest to understand the effects which this extension has on various model-theoretic and complexity-theoretic properties of these fragments.

This talk surveys some of the known results in this area. In particular, we consider the following logics: the one-variable fragment with counting quantifiers, the two-variable fragment with counting quantifiers, modal logics with counting modalities and the guarded two-variable fragment with counting quantifiers.


January 31, 2008, 17:30-18:30 - Room: Aula 21, Department of Information Engineering and Computer Science, University of Trento, Via Sommarive 14, Povo (TN)

Generation of effective language feedback for intelligent tutoring systems
Barbara Di Eugenio, University of Illinois at Chicago


In this talk, I will first give a general introduction to the activities we are engaged in as regards language interfaces for educational technology: mostly Intelligent Tutoring Systems (ITSs), but also systems to support peer learning. I will then focus on some recent results on generating language feedback for an ITS that tutors on an abstract sequence learning task. We collected a contrastive corpus where subjects interact with three different tutors, and verified they learn the most with the expert tutor. We then developed a generation framework based on a 3-tier planning model, where planning operators are derived automatically, on the basis of the association rules mined from our tutorial dialogue corpus. We evaluated 5 different versions of the ITS. The version that uses our planning framework is significantly more effective than the other four versions. Importantly, there are no significant differences between this version and the expert tutor it is modelled on. From a cognitive point of view, our work explores some important components of feedback.




February ^

Date Location Speaker Affiliation Title Abstract Slides Comments
February 21
17:30-18:30
FUB Boris van Schooten University of Twente Open issues and future directions in interactive QA Abstract vanschooten.pdf
vanschooten.sxi

February 22 (Friday)
10:00-11:00
11:00-13:00 (discussion)
FUB Luca Dini Celi srl Cross-Language Information Retrieval for Digital Libraries Abstract dini.pdf 



Abstracts (February)


February 21, 2008, 17:30-18:30 - Faculty of Computer Science, Free University of Bozen-Bolzano, Piazza Domenicani 3, Seminar Room (first floor left)

Open issues and future directions in interactive QA
Boris van Schooten, University of Twente


Question answering (QA) systems aim to make searching for information easier, potentially serving as a replacement of traditional search engines. However, most QA systems can only answer isolated questions. In the IMIX project, one of our aims was to incorporate dialogue functionality in existing QA technology, enabling follow-up questions and iterative search refinement.

We soon found that dialogue management for QA is quite different from that of typical natural language dialogue applications (such as form filling), and that interactive QA is in fact a largely unexplored field. One of the major contributions we made was an overall classification of the kinds of follow-up utterances (FU) users will pose, based on corpus data. The classification incorporates speech, text, pictures, and pointing as possible input and output modalities.

The theoretical basis of the latest classification will be discussed. It is not primarily based on traditional "dialogue theory", but rather on how current IR, QA, and reference resolution techniques are or could be equipped to deal with FU. Our corpus analyses also suggest that users can be steered more effectively towards posing more comprehensible FU. The talk will be concluded with a discussion of the issues of interactive QA when applied to real-life applications.


Friday, February 22, 2008, Time: 10:00-11:00 - Faculty of Computer Science, Free University of Bozen-Bolzano, Piazza Domenicani 3, Seminar Room (first floor left)

Cross-Language Information Retrieval for Digital Libraries
Luca Dini, CELI srl


The talk will describe the application of cross language information retrieval techniques to digital libraries. It will be shown how DLs can benefit of CLIR technologies provided that a number of challenges find an adequate answers. Notably, these challenges are:

  1. Semantic ambiguity;
  2. Proper Names Identification;
  3. Lack of domain restrictions;
  4. Coverage;
  5. Metadata vs Document Indexing/Retrieval.
  6. Sustainability by DL organizations.

Some temptative solutions to these challenges will be presented during the talk. After the seminar, Dini will also introduce the CACAO project, a recent cross national project aimed at evaluating the feasibility of the integration of CLIR techniques with federated cross national DL, and there will be a brainstorming on the introduced topics.




March ^

Date Location Speaker Affiliation Title Abstract Slides Comments
March 6
17:30-18:30
CIMeC Alessandro Lenci Università di Pisa Carving Meaning out of Word Spaces
Abstract  
March 13
17:30-18:30
FUB Richard Power The Open University Generating conceptually aligned texts Abstract  



Abstracts (March)


March 6, 2008, 17:30-18:30 - Room: TBA, Palazzo Fedrigotti (CIMeC), Corso Bettini 31, Rovereto

Carving Meaning out of Word Spaces
Alessandro Lenci, Università di Pisa


Representing words in distributional high-dimensional spaces have been proved to be able to model interesting aspects of their semantic phenomenology (e.g. synonymy judgements, priming, etc.). Yet, one of the major shortcomings of these models is that, while they provide a measure of the degree of relatedness among words, the particular type of semantic relation remains underspecified. In this talk, I will report about some experiments to automatically carve word spaces to cluster neighbors of a target word on the grounds of the semantic relation linking them. The goal is to try to discover the different types of semantic dimensions lying underneath the general notion of proximity in word spaces.


March 13, 2008, 17:30-18:30 - Faculty of Computer Science, Free University of Bozen-Bolzano, Piazza Domenicani 3, Seminar Room (first floor left)

Generating conceptually aligned texts
Richard Power, The Open University


A conceptually aligned text is one in which spans are systematically linked to a logical encoding of their meanings. I will describe the model of alignment used in WYSIWYM systems, which support knowledge editing through an automatically generated feedback text, and show that this model has a series of practical advantages through which it can support applications requiring semantic interactivity and inference, as envisaged by the Semantic Web community. I will also give demonstrations of some work in progress that implements these ideas.




April ^

Date Location Speaker Affiliation Title Abstract Slides Comments
April 3
17:30-18:30
FUB Philipp Cimiano Universität Karlsruhe (TH) Learning Qualia Structures from the Web Abstract cimiano.pdf



Abstracts (April)


April 3, 2008, 17:30-18:30 - Faculty of Computer Science, Free University of Bozen-Bolzano, Piazza Domenicani 3, Seminar Room (first floor left)

Learning Qualia Structures from the Web
Philipp Cimiano, Universität Karlsruhe (TH)


Lexical semantic theories, in order to be practically useful, need to be populated with large amounts of lexical entries. WordNet, for example, has been successfully applied for NLP applications partly because it is a rich, broad and domain independent resource with a huge amount of lexical entries. The case frame theory instantiated in the form of FrameNet seems to be following similar footsteps aiming to become a large-coverage lexical resource for natural language processing applications. However, there are up to date no considerable lexical resources for the Generative Lexicon (GL), in particular for the qualia structures which constitute one descriptive dimension of the Generative Lexicon.

In this talk, we will present recent work on learning qualia structures for nouns from the Web. The approach builds on earlier work based on the idea of matching specific lexico-syntactic patterns conveying a certain semantic relation as well as on the usage of the Web as a big corpus to reduce data sparseness. The qualia structures produced by our approach are actually ranked in the sense that the fillers of each qualia role (formal, constitutive, agentive and telic) are ordered according to a certain statistical measure. We assume that this ranking can be useful for NLP applications but also for lexicographers creating lexical entries by hand. In the talk, we will discuss in detail the evaluation and inherent problems of the approach.




May ^

Date Location Speaker Affiliation Title Abstract Slides Comments
May 8
17:00-18:00
CIMeC Caroline Heycock University of Edinburgh Locating finite verbs: A problem for linguists (and children)? Abstract  
May 15
15:30-18:00
FUB LCT-PhD          
May 22
16:00-17:00
CIMeC Michele Prandi Università di Bologna Grammatica delle forme e grammatica dei concetti Abstract  



Abstracts (May)


May 8, 2008, 17:00-18:00 - Room: Aula 9, Palazzo Fedrigotti (CIMeC), Corso Bettini 31, Rovereto

Locating finite verbs: A problem for linguists (and children)?
Caroline Heycock, University of Edinburgh


The Scandinavian language Faroese has been undergoing a syntactic change familiar from the history of other Germanic languages: finite verbs are no longer required to move to a position above negation except in verb-second contexts. The extent to which this change is complete is of particular interest because Faroese is a crucial case for different theories about the "driver" for this movement, frequently argued to be "richness" of agreement in the verbal system . Further, the apparent variability within individual speakers poses familiar questions about optionality and gradience in linguistic systems, and the status of systems in the process of change - questions that we may be in a better position to answer when faced with a contemporary language rather than historical records.

In this talk I will present and discuss new data that allow us to compare the acceptability of verb movement in Faroese and in two other languages: Icelandic - the contemporary language that we believe most closely approximates an earlier stage of Faroese - and Danish - the Scandinavian language with which Faroese has been in most intense contact, and which represents the system which Faroese appears to be moving towards. I will also discuss the potentially confounding distribution of verb second, and present some preliminary data from children's production that suggests the possibility of an early over-generalisation of verb movement, a particularly intriguing finding in the context of the observed direction of diachronic change.


May 22, 2008, 17:30-18:30 - Room: TBA, Palazzo Fedrigotti (CIMeC), Corso Bettini 31, Rovereto

Grammatica delle forme e grammatica dei concetti
Michele Prandi, Università di Bologna


Credo che sia facile, anche per un non linguista, rendersi conto che un'espressione non funziona - non riesce a trasmettere un messaggio, e, entro certi limiti, non riesce a esprimere un significato compiuto - se non è accompagnata da un atto di ragionamento - tecnicamente, di inferenza - del destinatario. Questa caratteristica dell'espressione può essere verificata in due momenti:
a) L'espressione trasmette un messaggio che non coincide necessariamente con il suo significato. In questo caso, l'espressione non adempie alla sua funzione sociale se il destinatario, partendo dal significato e da una costellazione di premesse, non mette in atto un ragionamento che porta come conclusione al messaggio. Se in aula dico Sono le sette e metto via gli appunti, gli studenti capiscono che la lezione è finita. Questo messaggio non è il significato del mio enunciato, ma il punto di arrivo di un loro ragionamento.
b) Negli strati periferici di un significato complesso, la codifica linguistica è tipicamente al di sotto della soglia di coerenza concettuale. Il lavoro della codifica è completato da un atto di inferenza. L'espressione Dopo le piogge la mulattiera è diventata impraticabile codifica una relazione di successione temporale tra due eventi, ma la relazione coerente, e quindi pertinente, è la causa, che il destinatario raggiunge per inferenza. L'ipocodifica è spesso confusa con l'ambiguità. Ma mentre l'ambiguità è un collasso locale della codifica, l'ipocodifica è un fenomeno strutturale dell'espressione. Un'espressione come la ragazza dei fiori non è ambigua: è sottocodificata.

Dove dobbiamo collocare il ragionamento? La risposta comunemente data in letteratura è più o meno questa: l'atto di ragionamento, in quanto atto, ci porta fuori dallo spazio della semantica; lo studio del contenuto delle espressioni linguistiche si dissolve in una pragmatica dell'azione umana non appena si supera il confine della codifica. La mia ipotesi è che ci sono ragionamenti che si esauriscono in una dimensione pragmatica (a), ma anche ragionamenti che presuppongono una vera e propria grammatica dei concetti (b).

Nel primo caso, le premesse del ragionamento sono dati appartenenti a una situazione contingente. Di questi ragionamenti possiamo fonire buoni esempi e principi ispiratori di carattere generale (per esempio le massime di Grice), ma non una grammatica. Così funziona, ad esempio, il rapporto tra il fatto che sono le sette e la fine della lezione. Nel secondo caso, le premesse del ragionamento sono fornite da un sistema di concetti coerenti e condivisi di lunga durata, che formano una vera e propria grammatica. Come la grammatica delle forme, la grammatica dei concetti è regolata da un criterio immanente - la coerenza - e può essere descritta nelle sue strutture. Il riconoscimento di una relazione di causa tra due fenomeni rientra certamente in questo secondo tipo di ragionamento.

Se a una grammatica delle forme affianchiamo una grammatica dei concetti, scopriamo che il significato delle espressioni linguistiche poggia su due basamenti altrettanto stabili: sulla codifica, che rinvia a una grammatica delle forme, e sul ragionamento, nella misura in cui fa affidamento su una grammatica dei concetti. Rientrano invece in una pragmatica dell'azione umana i ragionamenti basati su dati contingenti. Nella mia esposizione, cercherò di chiarire che cosa intendo per grammatica dei concetti.