Memory-based models of language learning and processing

Antal van den Bosch (Centre for Language Studies, Radboud University Nijmegen, the Netherlands)

Walter Daelemans (CLiPS, University of Antwerp, Belgium)


Memory-based language processing (MBLP) is an approach to language processing based on exemplar storage during learning and analogical reasoning during processing. From a cognitive perspective, the approach is attractive as a model for human language processing because it does not make any assumptions about the way abstractions are shaped, nor any a priori distinction between regular and exceptional exemplars. Schema-like behavior and the emergence of categories can be explained in MBLP as implicit by-products of analogical reasoning over exemplars in memory.

MBLP has been implemented as a computational model and toolkit in the form of the TiMBL software package (http://ilk.uvt.nl/timbl), and as such offers an interesting alternative to the popular mixed-effect regression modelling approach. The tutorial briefly introduces the software and a review of work in which the software has been used in the computational modeling of linguistic processes, including comparisons with alternative approaches such as Skousen's analogical modeling, and mixed-effects models. Although MBLP offers direct and explicit modes of learning (and forgetting), its application to modeling child language acquisition has been infrequent. We review recent work on modeling individual syntactic variation in CHILDES data.


Handouts

The presentation of the tutorial is available as a 6-slides-per-page handout file (PDF).


Tutorial Structure:

13.00 - 13.50:  Introduction, memory-based learning algorithms and metrics, TiMBL demonstration (case study: acquisition and processing of a morphological process; German plurals).

Literature: 

TiMBL tutorial and reference guide    

14.00 - 14.50: Case studies: acquisition of word stress, acquisition of the dative alternation.


15.00 - 15.50: Comparison alternative approaches, psychological validity (memory research), pointers to other relevant research, discussion.