Proceedings of the EACL Workshop on Computational Models of Language Acquisition and Loss

Résumé En

The past decades have seen a massive expansion in the application of statistical and machine learning methods to speech and natural language processing. This work has yielded impressive results which have generally been viewed as engineering achievements. Recently researchers have begun to investigate the relevance of computational learning methods for research on human language acquisition and loss. The human ability to acquire and process language has long attracted interest and generated much debate due to the apparent ease with which such a complex and dynamic system is learnt and used on the face of ambiguity, noise and uncertainty. On the other hand, changes in language abilities during aging and eventual losses related to conditions such as Alzheimer's disease and dementia have also attracted considerable investigative efforts. Parallels between the acquisition and loss have been raised, and a better understanding of the mechanisms involved in both, and of how the algorithms used to access concepts are affected in pathological cases can lead to earlier diagnosis and more targeted treatments. The use of computational modeling is a relatively recent trend boosted by advances in machine learning techniques, and the availability of resources like corpora of child and child-directed sentences, and data from psycholinguistic tasks by normal and pathological groups. Many of the existing computational models attempt to study language tasks under cognitively plausible criteria (such as memory and processing limitations that humans face), and to explain the developmental stages observed in the acquisition and evolution of the language abilities. This was the third edition of this workshop that was first held at ACL 2007 in Prague and then in EACL 2009 in Athens. The workshop was targeted at anyone interested in the relevance of computational techniques for understanding first, second and bilingual language acquisition and change or loss in normal and pathological conditions. We invited submissions on relevant topics, including: * Computational learning theory and analysis of language learning * Computational models of first, second and bilingual language acquisition or of the evolution of language * Computational models and analysis of factors that influence language acquisition and loss in different age groups and cultures * Data resources and tools for investigating computational models of human language processes * Empirical and theoretical comparisons of the environment and its impact on acquisition * Investigations and comparisons of supervised, unsupervised and weakly-supervised methods for learning Submissions included works on specific languages like English, Portuguese and Hebrew, and also crosslinguistic studies. Besides paper presentations the technical program included resources and systems demonstrations, and two invited talks by Mark Steedman, from University of Edinburgh (UK) and Alessandro Lenci, from University of Pisa (Italy).