Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision 		Previous revision
esslli:background [2008/08/13 16:58]
schtepf created 		— (current)
Line 1: Line 1:
-====== Lexical Semantics Workshop (ESSLLI 2008)  ====== 
- 
-===== Background and motivation ===== 
- 
-[[http://wordspace.collocations.de/lib/exe/fetch.php/esslli:lexical_semantics_workshop_-_motivational_poster.pdf?id=esslli%3Astart&cache=cache|{{ esslli:motivational_poster_thumb.jpg|Motivational Poster}}]] 
- 
-Corpus-based distributional models (such as LSA or HAL) 
-have been claimed to capture interesting aspects of word meaning 
-and provide an explanation for the rapid acquisition of semantic 
-knowledge by human language learners.  
-However, although these models have been proposed as 
-plausible simulations of human semantic space organization, 
-careful and extensive empirical tests of such claims are still lacking. 
- 
-Systematic evaluations typically focus on large-scale 
-quantitative tasks, often more oriented towards engineering 
-applications (see, e.g., the recent SEMEVAL evaluation campaign) than 
-towards the challenges posed by linguistic theory, 
-philosophy and cognitive science.  This has resulted in a great divide  
-between corpus-driven computational approaches to semantics on the one 
-hand and theory-driven symbolic approaches on the other -- a situation that 
-is characteristic of the linguistic and of most 
-of the cognitive tradition. 
-Moreover, whereas human lexical semantic competence is obviously 
-multi-faceted -- ranging from free association to taxonomic judgments to 
-relational effects -- tests of distributional models tend to focus on a 
-single aspect (most typically the detection of semantic similarity), 
-and few if any models have been tuned to tackle 
-different facets of semantics in an integrated manner. 
- 
-Our workshop purports to fill these gaps by inviting research teams and 
-individual scholars to test their computational models on a variety of small but 
-carefully designed tasks that aim to bring out linguistically and 
-cognitively interesting aspects of semantics 
-([[#Tasks and data sets|see below]] for details).  To this effect, annotated 
-datasets are available to the participants, who are 
-encouraged to explore them and highlight interesting aspects of their 
-models' performance, conduct quantitative and qualitative error analysis, etc. 
- 
-The focus is NOT on competition, but on understanding how different 
-models highlight different semantic aspects, how far we are from 
-an integrated model, and which aspects of semantics are beyond the 
-reach of purely distributional approaches.  
-In fact, we believe that at the current state of the art in 
-computational and distributional semantics, our goal should not be 
-to develop the best-performing model for a specific application, but rather 
-to enlarge our understanding of the limits and potentialities  
-of different approaches when confronted with cognitively realistic tasks.  
- 
-In addition to these practical tasks, theoretical and experimental papers 
-discussing the relation between distributional and symbolic approaches to meaning 
-are also invited.  We are particularly interested in papers that analyze our 
-task data sets from a theoretical perspective or that discuss simulation results 
-and their implications for semantic and cognitive theory. 
- 
-Through collaborative preparatory work on the Word Space wiki ([[http://wordspace.collocations.de/|wordspace.collocations.de]]) and thanks to the ESSLLI multiple-day workshop format, we hope that this initiative will foster 
-collaboration among the nascent community of researchers interested in computational semantics from a theoretical rather than engineering-oriented point of view. 
  

You are here: start » esslli » background