Events / Kalender

Research Group Formal and Computational Linguistics (ComForT)
Onderzoeksgroep Formele en Computationele Linguïstiek (ComForT)

Also see / zie ook CRISSP Events page
dinsdag 26 april 2016 om 14u (MSI 01.28)
    lezing door Jeroen van Craenenbroeck (CRISSP, KU Leuven)
    Handle your verb clusters with care
    zie abstract
donderdag 31 maart 2016 om 14 u (KU Leuven, Campus Brussel, Zaal Paternoster (6303) van Hermes 3, Stormstraat 2, 1000 Brussel)
    verdediging doctoraal proefschrift door Koen Roelandt
    Most or the Art of Compositionality. Dutch "de/het meeste" at the Syntax-Semantics Interface
dinsdag 22 maart 2016 om 14u (LETT 05.30)
    lezing door Luk Draye (KU Leuven)
    Man. Van onbepaald tot cryptopersoonlijk voornaamwoord van de 1ste en 2de persoon – of omgekeerd?
    zie abstract
vrijdag 4 maart 2016 om 15u (KU Leuven, Justus Lipsiuszaal, Blijde Inkomststraat 21, 3000 Leuven)
    verdediging doctoraal proefschrift door Jeroen Pollentier
    Der Dativcausee in deutschen lassen- und niederländischen laten-Konstruktionen. Eine diachrone und kontrastive Korpusstudie
donderdag 19 november 2015 om 14u (LETT 03.16)
    lezing door Koen Roelandt (CRISSP/KU Leuven Campus Brussel)
    The Internal Logic of Adjectival Lexical Fields

    Aristotle already observed that logical relations apply between expressions containing adjectives. For instance, the relation of contrariety holds between the sentences in (1): they cannot both be true at the same time, but they can both be false at the same time, namely when (2) is true.

    (1a) Socrates is tall.
    (1b) Socrates is short.
    (2) Socrates is neither tall nor short.

    In this talk, I will demonstrate that the four Aristotelian relations (contrariety, contradictoriness, subalternation and subcontrariety) are fundamental to adjectival lexical fields. The relations uncover the internal organisation of adjectival fields and provide insights into how they are generated. More specifically, I will first develop a model using standard logic that computes all logically possible concepts in a given field. A second model, involving natural logic, predicts which of these concepts are actually realised in natural language. These models are then visualised using 3D diagrams that give a straightforward representation of the intricate relations inside a lexical field.
maandag 12 oktober 2015 om 14u (LETT 02.16)
    lezing door Guido Vanden Wyngaerd (CRISSP/KU Leuven Campus Brussel)
    A Puzzle in Gradable Adjectives
    zie abstract
donderdag 8 oktober 2015 om 14u (KU Leuven, Justus Lipsiuszaal, Blijde Inkomststraat 21, 3000 Leuven)
    verdediging doctoraal proefschrift door Liesbeth Augustinus
    On complement raising and cluster formation in Dutch. A Treebank-supported investigation
dinsdag 26 mei 2015 om 14u (LETT 03.30)
    lezing door Vincent Vandeghinste (CCL, KU Leuven)
    Using Lexicalized Parallel Treebanks for Synchronous Tree Substitution Grammar Induction.

    For the SCATE project (Smart Computer-Aided Translation Environment) we are investigating how we can induce a synchronous tree substitution grammar (STSG) from a parallel aligned lexicalized treebank to improve syntax-based statistical machine translation. Such an STSG can serve as a set of transfer rules with weights on each rule. Given a source sentence parse tree, such a grammar allows us to generate target language parse trees, from which the target language sentence can be generated.
    The model is similar but more flexible than the model used in the PaCo-MT system. One difference is the fact that we estimate the probability of an STSG rule by multiplying the probabilities of generating the heads with the probabilities for generating the other dependents, instead of estimating the probability through the relative frequency of a translation pattern with heads and non-heads taken together. A second difference is the fact that we use lexicalized trees, in which each node in the tree contains information about its head-token, head-lemma and head-category. Combining these features with the syntactic category and dependency relations of the trees provides us with five sources of information, which we can unify in order to estimate the probability of unseen cases, based on the probabilities of its features and combination of features.
    We will discuss this model and show how it advances over the model used in the PaCo-MT project through a concrete example.
woensdag 6 mei 2015 om 16u (LETT 06.30)
    lezing door Jeroen Pollentier (KU Leuven)
    Causatiefconstructies in het Nederlands en het Duits

    My dissertation deals with the dative causee in German and Dutch causative constructions with the auxiliary “lassen” resp. “laten”. Up till the 19th century in German causative constructions with “wissen” (to know) and with two-place transitive weakly agentive perception verbs the causee could be put in the dative, e.g. in “Ich lasse dir etwas sehen” (I let you see something). A dative causee is still used in similar constructions in contemporary Dutch.
    It has often been argued that “sehen lassen/laten zien” has been reanalyzed as “zeigen/tonen” (to show). A corpus study has shown, however, that a dative causee could also occur in combination with highly agentive and with one-place intransitive main verbs. In my dissertation I argue that the dative causee is to be interpreted as an interestee (i.e. an experiencer, recipient, beneficiary…) not only in those causative constructions with two-place transitive weakly agentive perception verbs or cognitive verbs, but also in Middle High German “lassen”-constructions with a one-place intransitive main verb. There also exists a second type of dative causee construction, which has only developed later as the result of a grammaticalization process. Here the dative competes with the accusative as the case of the ‘normal’ causee.
dinsdag 28 april 2015 om 11u (LETT 03.16)
    lezing door Liesbeth Augustinus (CCL, KU Leuven)
    Werkwoordsclusters in het Nederlands. Een treebank-gebaseerd onderzoek.

    In deze presentatie staat een bekend fenomeen uit de Nederlandse syntaxis centraal: de constructie van werkwoordsclusters, zoals in (1):

    (1) … dat de minister hierop wel zal1 willen2 reageren3. (CGN, fng000177_39)

    In (1) vormen alle werkwoorden een cluster in de tweede zinspool, die voorafgegaan wordt door de niet-werkwoordelijke argumenten. Het doel van dit onderzoek is om na te gaan hoe corpus-gebaseerd onderzoek tot nieuwe inzichten kan leiden voor de syntactische analyse van werkwoordsclusters. In het eerste deel van de presentatie zal een definitie van Nederlandse werkwoordsclusters gegeven worden die aansluit bij observaties uit twee syntactisch geannoteerde corpora of treebanks (CGN en LASSY). Daarbij zal ook de relatie tussen clustervorming en daarmee verwante fenomenen aan bod komen, zoals Infinitivus Pro Participio (2), de doorbreking van de cluster door niet-werkwoordelijke elementen (3) en woordvolgordevariatie (4).

    (2) Met wat meer geluk hadden we hier staan/*gestaan juichen. (LASSY, WR-P-P-H-0000000020.p.14.s.8)
    (3) De dokters zeggen wel dat het gaat1 goed komen2. (CGN, fva400643_87)
    (4) Er zijn toch zo’n paar boeken die ge moet1 gelezen3 hebben2 in uw leven. (CGN, fva400503_11)

    In het tweede deel van de presentatie zal aangetoond worden hoe de empirische observaties bijdragen tot een accuratere syntactische analyse geformuleerd in het Head-driven Phrase Structure Grammar (HPSG) framework. Een bijkomend voordeel van deze analyse is dat ze naast werkwoordsclusters ook niet-clusterende constructies kan behandelen, zoals de derde constructie (5) en adpositiestranding (6).

    (5) ik heb ’r geprobeerd te bellen maar d’r werd niet opgenomen (CGN, fna000583_351)
    (6) … dat zij daar nog wel van hield. (CGN, fna000741_12)

    Net zoals in constructies met werkwoordsclusters zijn in dergelijke constructies de argumenten (resp.” ’r” en “daar”) niet adjacent aan hun selector (resp. “bellen” en “van”).
maandag 23 maart 2015 om 14u (LETT 03.30)
    lezing door Tanja Temmerman en Will Harwood (CRISSP / KU Leuven)
    Pushing the Boundaries: Idioms in Dutch dialects and English.

    This talk asks the question of what the building blocks of idioms are in Dutch and its dialects, and in English.
    Idioms are commonly constructed from the verb and its arguments, e.g. kick the bucket (verb + object), or the shit hit the fan (subject + verb + object). If these items are altered or replaced, the idiomatic interpretation is lost. Indeed, it has often been claimed that idioms can only be comprised of the verb and its argument.
    We show however, that idioms can be dependent upon material other than the verb and its arguments. In particular, it is shown that idioms exist which are dependent on passive voice, aspect and modality. Furthermore, we show that there is cross-linguistic variation with regards to what constitutes the building blocks of idioms. In English, idioms exist that are dependent on passive voice or progressive aspect, but not perfect aspect or modality. Dutch dialects, however, not only exhibit idioms dependent on passive voice and progressive aspect, but also idioms dependent on perfect aspect and modality. We argue that this variation stems from the difference in size of the traditional VP domain within the syntax of these languages.
dinsdag 17 februari 2015 om 10u (LETT 03.16)
    lezing door Marijke De Belder (CRISSP, KU Leuven Campus Brussel)
    Morphological irregularities as a gateway to understanding variation: vowel lengthening in the Dutch nominal domain
    zie abstract
woensdag 3 december 2014 om 10u (MSI 01.28)
    lezing door Leen Sevens (KU Leuven CCL)
    Able-To-Include: Automatic Translation from Pictographs to Text and Vice Versa
    zie abstract
donderdag 6 november 2014 om 10u (MSI 02.08)
    lezing door Dany Jaspers (CRISSP, KU Leuven Campus Brussel)
    Beyond Logic and Colour

    On the basis of a parallellism between logical operators and colour terms described in earlier work (Jaspers 2012), it will be explored in this talk how the opposition pattern found there extends to a specific set of other closed lexical fields, including those of number terms and demonstrative pronouns. On the basis of the system that keeps popping up, the question arises whether certain fixed “morphogenetic” laws might govern the growth of closed lexical fields. (Jaspers, D. 2012. Logic and Colour, Logica Universalis, 6 :227-248)
donderdag 5 juni 2014 om 14u (LETT 02.16)
    lezing door Hans Smessaert & Lorenz Demey (KU Leuven, ComForT & CLAW)
    Inleiding in de Logische Geometrie

    In een eerste deel van de presentatie (i) lichten we de algemene doelstellingen van het framework toe, (ii) definiëren we de Aristotelische relaties/geometrie, (iii) introduceren we het bitstring-formaat, (iv) beschrijven we verschillende types Aristotelische diagrammen, nl. vierkanten en hexagons, en (v) bespreken we de band met de basisprincipes van de formele semantiek.
    In een tweede deel beargumenteren we dat de Aristotelische geometrie hybride is tussen twee andere sets van logische relaties, nl. de oppositiegeometrie en de implicatiegeometrie. De kracht en populariteit van de Aristotelische geometrie schuilt dan in het feit dat die systematisch de meest informatieve relaties uit de twee andere geometrieën kiest.
    In een derde deel bestuderen we de rhombische dodecahedron (RDH) van opposities. Dat is een 3D figuur die ons in staat stelt om de complexe netwerken van Aristotelische relaties tussen 14/16 formules/concepten te visualiseren. We stellen een systematische strategie voor om de interne structuur van de RDH exhaustief te beschrijven. We bieden op die manier een antwoord op de vraag: welke en hoeveel kleinere Aristotelische diagrammen zijn er in de RDH-structuur ingebed?
vrijdag 16 mei 2014 om 14u (LETT 02.16)
    lezing door Johan Rooryck (Universiteit Leiden)
    Core knowledge systems and the language faculty

    In this talk I will discuss Core knowledge systems and their relation to the language faculty. Core knowledge theory holds that humans are born with a small number of language-independent task-specific systems of core knowledge (SCK). These innate cognitive skills can be related to neuronal substrates that represent an initial capacity for building distinct mental representations of conceptual objects and agents, number and geometry, as well as social interaction (Spelke & Kinzler 2007). Properties of SCK have been primarily determined by studies based on direction and duration of gaze. It is now assumed that SCK are universal, domain-specific, and characterized by signature limits. In combination with the compositional nature of the autonomous faculty of language that expands their limits, SCK may serve as a base for further knowledge through experience.
    I will show how recent insights into the nature of SCK raise interesting questions for the faculty of language in its narrow (FLN) and broad (FLB) sense (Hauser, Chomsky & Fitch 2002). If time permits, I will present the NWO-sponsored ( 2m) Horizon project 'Knowledge and culture', in which specific research topics from the humanities (music cognition, poetry, language, visual arts) are studied from the perspective of SCK.
maandag 24 februari 2014 om 14u (HUBrussel, Campus Hermes 3, Stormstraat, lokaal 4303.1)
    lezing door Marijke De Belder (CRISSP, KU Leuven Campus Brussel)
    The root and nothing but the root: primary compounds in Dutch

    The root hypothesis (Halle and Marantz 1993, Borer 2005a,b, 2013, De Belder 2011) states that a lexical projection consists of nothing but an acategorial stem which receives its category from the functional projection it merges with. For example, if kiss merges with a determiner, it will be interpreted nominally, if it merges with tense it will be interpreted verbally.
    There is no obvious reason why a compound’s non-head should necessarily project any functional structure. If the root hypothesis is valid, one therefore expects the existence of compounds of which the non-head is nothing but a bare root. I show that the non-head of a specific subtype of Dutch primary (i.e. non-synthetic, non-phrasal) compounds is indeed demonstrably a bare root. It is fully acategorial. It does not contain categorial heads or functional projections.
    Dutch has a type of compounding (henceforth root compounds) of which the non-head (i.e. the left-hand part) may be associated with just any category, including N, V, A, cardinals, interjections, conjunctions and adverbs, see (1) for some examples. It is directly adjacent to the head. In what follows I claim that root compounds do not allow for just any type of category as their non-head. They rather systematically select for one and the same syntactic category: a bare root, which cannot merge with functional or categorial heads.
    • slaap-pil (sleep-pil/‘sleeping pill’)
    • drie-luik (three-panel/‘triptych’)
    • ja-woord (yes-word/‘marriage vows’)
    • achter-deur (back-door/‘backdoor’)
    • of-poort (or-gate/‘or-gate’)
    As such, the structure of these compounds contrasts sharply with those compounds of which the root may be considered nominal(ized). In the nominalized type the non-head is followed by a so-called linking element (LE), which is most presumably a piece of nominal inflection (kat-en-staart ‘cat-LE-tail’).
donderdag 30 januari 2014 om 14 uur (KU Leuven, Promotiezaal, Naamsestraat 22, 3000 Leuven)
    verdediging doctoraal proefschrift door Adrienn Jánosi
    Long Split Focus Constructions in Hungarian with a View on Speaker Variation
woensdag 4 december 2013 om 14u
    lezing door Gosse Bouma (RU Groningen)
    Om-omission in Dutch verbal complements

    The complementizer om, which heads to-infinitival clauses in Dutch, is optional if the clause it introduces is a complement. We investigate which linguistic features influence the distribution of om in this construction, using data collected from an automatically parsed corpus of Dutch.
    A large part of the variation is accounted for by the governing verb. In addition to this, features that account for syntactic or processing complexity play a significant role, as well as features that characterize typical complement clauses for a given governor and typical purpose or goal modifier clauses. The fact that lexical variation plays a dominant role in our model motivates our choice for a mixed effects logistic regression model, where verbs are used as random effects.
dinsdag 5 november 2013 om 15u
    lezing door Ineke Schuurman (KU Leuven/U Utrecht)
    CLARIN behind the scenes (or: those parts of a digital humanities research infrastructure end users will not be confronted with...)

    CLARIN is a large European programme, focussing on making all kinds of resources and technology related to language available to students and researchers in human and social sciences: linguistics, literature, sociolology, psychology, history, political science, ... all fields in which language/text in whatever format (spoken, wrtten, signed, ...) is of importance. In the end (CLARIN being under construction), end users need to know only one address for all tools and resources, and nothing about the processes used. But how do the people behind the CLARIN infrastructure take care of that?
    In this talk I will focus on the status/use of (de facto) standards and on semantic interoperability, both with respect to metadata and to content. Among the items I will address:
    • will only resources/tools making use of standards be supported? And what are the consequences for the end user ?
    • what about conflicting standards ?
    • what about lacking standards ?
    • what about conflicting needs of end users ?
    • what about legacy data ?
    • What about resources in no longer used variants of a language ?