Constructing A Connectionist Acquisition Model For Deep Knowledge In L2Vocabulary

In the study of L2vocabulary acquisition, researchers have found that sometimesone cannot use the words accurately and skillfully in real context even if they havememorized a large quantity of words. Hence, the quantity of words remembered is notdirectly related to the quality of language usage. In recent years, researchers havefocused on the dimensions of vocabulary knowledge. They think that vocabularyknowledge is multidimensional rather than unidimensional. That is, vocabularyknowledge covers not only a word’s static aspects such as sound, form, meaning(breadth) but also its dynamic aspects (depth), including word frequency, familyrelationship, syntactic functions, collocation etc. On such basis, various frameworksof vocabulary knowledge have been built. Though they define vocabulary dimensionsdifferently, they all include the dimensions of depth and breadth in vocabulary, whichhave been identified as the most basic and core elements in vocabulary knowledge. Ithas also been pointed out that deep knowledge plays a vital role in vocabularyacquisition, and the two dimensions of depth and breadth should be closely linked inacquisition rather than isolated.Connectionism believes that vocabulary knowledge is not stored and retrievedindependently in human brains. Instead, it is stored as units of network and these unitsare interconnected with weights, which are strengthened by activation of the unitscontinually. Therefore it is believed that a connectionist network corresponds to thebiological understandings of the human cerebral neural system.Connectionism is a theory based on neurophysiology, neuropsychology andcognitive psychology. Its purpose is to simulate human brain’s neural activities bybuilding neural networks in order to explore human cognitive process. Classicsymbolic theory views that human brain processes information in a discrete, orderly,non-selforganized and local way, while connectionism network holds that humanbrain processes information in a parallel and distributed way, which showssub-symbolic, successive, parallel and distributed features (Li ping,2003:164-175). In addition, a connectionist neural network also has functions of self-study,self-adaptation and self-organization, which enables the network to adjust the weightsbetween neural units by adopting a series of training strategies until the outputmatches the desired input. Such features of a connectionist network are close to theneural structure of human brains and the real human psychological conditions. Hencelinguists have great interest in building connectionist models to study the languageproduction, comprehension and application and try to make reasonable interpretationson them.Up till now, the acquiring processes of vocabulary, grammar, reading andparagraph comprehension have been successfully simulated. Yet researchers abroadmainly take first language as their study object and a few models on L2vocabularyacquisition are set on the static level, that is, only the processes of vocabulary storageand access have been depicted. There are so few models on vocabulary production.What’s more, there has been no model covering deep knowledge in L2vocabularyacquisition.Deep knowledge acquisition is one of the central issues in vocabulary research(Henrisken,2008). Vocabulary acquisition is the essential and vital part in L2study.This thesis will construct a connectionist acquisition model for deep knowledge in L2vocabulary based on connectionism to explore how a learner acquires an L2word andthe influence of L1on such process. It is anticipated that this research will offer acognitive model for the learners in L2vocabulary acquisition and a theoreticalenlightenment for L2vocabulary teaching.