Lectures on Cognitive Semantics and Natural Language Processing

Universidad Autónoma de Barcelona, May 20 and 21, 2010
Facultad de Filosofía y Letras, Edificio B
Bellaterra, Spain
Google map


These lectures will cover two topics in cognitive linguistics: (1) context-dependency and embodiment in event interpretation, and (2) cross-linguistic differences in conceptualization patterns. These lectures are organized by the Spanish FrameNet Project.


1. Thursday May 20, 12:00 - 13:30
Nancy Chang, Sony Computer Science Laboratory, Paris
"Reframing aspectual composition"

Video and Powerpoint


Languages employ a wide range of devices that shape event interpretation--that is, that affect whether an event is seen as static or dynamic, whether it unfolds gradually over time or represents a discrete change of state, whether it involves multiple iterations (at a particular level of granularity), etc. These devices include explicit grammatical markers of tense and aspect, as well as the aspectual constraints considered inherent to specific verbs or classes of verbs (cf. Aktionsart). But the crosslinguistic evidence suggests that event interpretation depends on properties of the sentence (and the event) as a whole: the constraints imposed by various lexical, phrasal and clausal constructions interact with contextual factors to produce subtle differences in interpretation and acceptability. A full account of these phenomena thus requires a principled means of combining temporal and aspectual constraints from disparate constructions, as well as a richer model of event structure that not only identifies relevant features but also captures their dynamic interactions in context.

In this talk I address these requirements by exploiting two key ideas: (1) frame-specific aspectual constructions; and (2) simulation-based event representation. I describe a framework that combines these ideas to provide a natural and elegant account of a wide variety of crosslinguistic aspectual phenomena, including both traditional aspectual distinctions and subtler context-dependent ones (e.g., resource production and consumption, permanent and reversible changes of state, goal achievement, iteration on different time scales). In particular, I define both basic event frames (corresponding to states, continuous processes and discrete transitions) and more complex event frames, and then illustrate various constructions that can evoke and constrain these frames (including verbs, auxiliaries, argument structure constructions and temporal modifiers). I also show how these schematic event frames serve as a principled, limited interface between linguistic knowledge and the situational context: constructions need only specify their compositional constraints in terms of these schematic structures, allowing the dynamic simulation process to account for the potentially unbounded range of context-dependent inferences and coercions present in more challenging aspectual phenomena. The resulting account demonstrates how a unified approach to integrating specifically linguistic knowledge with richer embodied representations can elegantly capture context-dependent syntactic and semantic generalizations.

2. Friday May 21, 12:00-13:30
Michael Ellsworth, International Computer Science Institute
"Conceptualization of motion and emotion in English and Spanish"

Video and PowerPoint


As shown by the groundbreaking research of Talmy and Slobin, there are differences in conceptualization patterns in different languages which have a deep influence on the frequency and availability of particular means of expression. Spanish, in particular, has been shown to conform to the Verb-framed language type (otherwise known as Path-in-verb), contrasting with the Satellite-framed (otherwise known as Path-not-in-verb) language type of languages such as English, but there are many other patterns of difference that have received less attention. In particular, the constructions--especially lexical constructions--used to describe emotion in Spanish are more likely to profile a change of state than corresponding English language descriptions. This type of linguistic difference has been described as a lexicalization pattern in the dimension of aspect (Talmy 1985), but this purely featural description fails to relate the preference of languages like Spanish for state-change conceptualization to the propensity for Verb-frame conceptualization patterns. A frame-based understanding of these kinds of data is, by contrast, shown to be arbitrarily extensible to lower level generalizations and allows us to relate multiple cooccurrent aspects of frames to each other. In addition, evidence from the emotional domain, unlike that of the motion domain, points to a need to distinguish at which levels particular preferences are operative: English and Spanish are similar in having many basic verbal roots pertaining to emotion with causative conceptualization (En. piss [off], Sp. enoj-a/enfad-a). Spanish often lexicalizes causative roots as inchoative structures, accompanied by corresponding syntax, eg. enojarse. In English, by contrast, a static conceptualization of emotions is preferred, a preference that shows up at the lexical level, with many stative participles (pissed off), at the phrasal level, with the use of the progressive (you're pissing me off), and at the pragmatic level (The company pissed/pisses me off).


Collin Baker, International Computer Science Institute (ICSI)
Marc Ortega, Universidad Autónoma de Barcelona (UAB)
Carlos Subirats, UAB and ICSI

Image in logo by Wikimedia Commons / CC