中国教育类核心期刊 CSSCI来源期刊 RCCSE中国权威学术期刊

杂志文章

您的位置:首页 >文章目录 >2019年第8期>

    基于深度学习技术的注意力转移模式的挖掘 ——以二语学习者的眼动数据为例

    The Mining of Attention Transfer Pattern Based on Deep Learning Technology : A Case of Eye Movement Data of Second Language Learners

【作      者】:

严薇薇, 旷小芳, 肖云霞, 郑梦雪, 刘 俊, 杨 娟


【关 键 词 】:

二语习得; 注意力控制; 注意力模式; 注意力转移; 深度学习; 卷积神经网络; 眼动追踪


【栏      目】:

理论探讨


【中文摘要】:

注意力在二语习得领域被认为是将输入转化为吸收的充分必要条件,是影响二语学习的主要认知因素。目前大多数研究集中在注意力分布上,很少有研究涉及二语学习者的注意力转移模式,然而注意力转移具有的时序特征更能准确地反映二语学习者的思维过程。关于注意力模式的发现方法主要有描述统计方法和基于白盒的回归/预测技术,前者能获得具有统计学意义的结论,后者能建构较复杂的因果关系模型,但是均无法直接从高纬度特征空间中获取有意义的指征,因而无法建立较高准确率且可解释的模型。基于此,本研究使用深度学习技术以及可视化技术挖掘二语(英语)学习者处理在线任务时的注意力转移模式。正反例的关键特征热度图显示,低龄二语学习者的线性注意力控制模式与其在线任务表现紧密关联,可直观解释其线性注意力控制能力对在线任务完成度的影响。该模型同时具备较高回归/预测准确率。此结论对我国低龄儿童英语学习的认知干预研究有着重要意义。


【英文摘要】:

Attention in second language acquisition is claimed to be the necessary and sufficient condition for conversing input to intake, and it is one of the main cognitive factors affecting L2 learning. Currently, most studies focus on attention allocation, and few work on the attention transfer pattern of L2 learners. However, the temporal features of attention transfer can reflect the thinking process of L2 learners more accurately. At present, methods used for discovering attention patterns include descriptive statistical methods and white-box-based regression/prediction techniques. Although the former can obtain statistically significant conclusions and the latter can construct causal relationship models, neither method can directly obtain meaningful indicators from high-latitude feature space. Therefore, it is impossible to establish an explainable model with high accuracy. Based on this, this study uses deep learning technology and visualization techniques to explore the attention transfer pattern of second language(English) learners when they are dealing with online tasks. The heat maps of the positive and negative examples show that the linear attention-control patterns of young second language learners is closely related to their online task performance, which can intuitively explain the influence of their linear attention control on online task completion. The model also has high regression/prediction accuracy.This conclusion is of great significance to the cognitive intervention research on English learning of young children in China.

网站首页 | 联系我们 | 意见建议| 杂志订阅| 管理部登录

版权所有 © 电化教育研究 CopyRight ©e-EDUCATION RESEARCH 2011-2018 电话:0931-7971823 地址:西北师范大学《电化教育研究》杂志社