tailieunhanh - Báo cáo khoa học: "Modelling Early Language Acquisition Skills: Towards a General Statistical Learning Mechanism"

This paper reports the on-going research of a thesis project investigating a computational model of early language acquisition. The model discovers word-like units from crossmodal input data and builds continuously evolving internal representations within a cognitive model of memory. Current cognitive theories suggest that young infants employ general statistical mechanisms that exploit the statistical regularities within their environment to acquire language skills. | Modelling Early Language Acquisition Skills Towards a General Statistical Learning Mechanism Guillaume Aimetti University of Sheffield Sheffield UK Abstract This paper reports the on-going research of a thesis project investigating a computational model of early language acquisition. The model discovers word-like units from cross-modal input data and builds continuously evolving internal representations within a cognitive model of memory. Current cognitive theories suggest that young infants employ general statistical mechanisms that exploit the statistical regularities within their environment to acquire language skills. The discovery of lexical units is modelled on this behaviour as the system detects repeating patterns from the speech signal and associates them to discrete abstract semantic tags. In its current state the algorithm is a novel approach for segmenting speech directly from the acoustic signal in an unsupervised manner therefore liberating it from a pre-defined lexicon. By the end of the project it is planned to have an architecture that is capable of acquiring language and communicative skills in an online manner and carry out robust speech recognition. Preliminary results already show that this method is capable of segmenting and building accurate internal representations of important lexical units as emergent properties from cross-modal data. 1 Introduction Conventional Automatic Speech Recognition ASR systems can achieve very accurate recognition results particularly when used in their optimal acoustic environment on examples within their stored vocabularies. However when taken out of their comfort zone accuracy significantly deteriorates and does not come anywhere near human speech processing abilities for even the simplest of tasks. This project investigates novel computational language acquisition techniques that attempt to model current cognitive theories in order to achieve a more robust speech recognition system. .