摘要
很多学习者在面对海量资源时,难以快速筛选出对自己有用的内容。鉴于此,该研究探索了英语教学与信息技术的融合,提出在混合学习情景下,利用协同推荐技术将学习者与优质视听资源匹配,为学习者动态定制个性化学习内容。在大样本数据MovieLen 100k上进行定量分析,发现用户普遍只对少量资源熟悉,需要推荐机制帮助扩大认知领域,且该方法可以提升推荐效果;在学生社团中进行教学实践,发现不同背景的学生对视听资源的偏好存在差异,且该方法对视听学习产生了积极影响。该研究旨在为英语教育资源服务的信息化和智能化提供一定参考。
Blended learning builds up seamless language environment with the help of mobile devices.In the blended learning environment,learners can study at anytime and anywhere.However,the learners are overwhelmed by information and suffer from information overload.At the same time,blended learning separates the knowledge into pieces and causes difficulties in knowledge construction.It is urgent to help learners identify the information that best fits their interests,raising the prominence of recommender systems.Collaborative filtering is the most successful and widely used technique for building recommender systems.Collaborative filtering is classified as neighborhood-based approaches and latent factor approaches.Latent factor approaches are always based on matrix factorization.The original matrix factorization models were designed to model users’explicit feedback,which we refer as ratings,such that user-item relationships(ratings)can be captured by their latent factors’dot product.When explicit feedback is not available,implicit feedback can be utilized,like browsing,likes and purchasing.The rating data is in a matrix with one dimension representing users and the other representing items of interest.The key idea is to factorize the user-item matrix into user factors and item factors that represent user tastes and item characteristics in a common latent space.This paper discussed the integration of English visual-aural-oral resources and information technology,and utilized a collaborative recommendation model to fulfill personalized learning by matching the learners and resources.It first develop a mobile intelligent education platform to display multimedia resources and collect the data.Then it uses matrix factorization to predict unobserved ratings by factorizing the user-item matrix into user factors and item factors.Constructing an additional regularization term based on the geometric intuition that distances between similar items should be minor in the latent space,so that item-wise similarity is incorporated into matrix factorization through manifold regularization.Several groups of experiments were presented to assess the effectiveness of the proposed framework.In particular,the dataset used in the experiment is Movie Lens 100 K,which contains 943 users,1682 movies and 100 thousand ratings.It conducts a statistical analysis of the users’behavior characteristics by counting the number of ratings for each user and depicting the distribution of ratings on all users.It can be observed that most users rated only very few movies and minority users contributes to most ratings.Using mean absolute error as measurement criteria of prediction accuracy,the experimental results show that the model,which reveals more about item latent features and measures the similarity accurately,outperforms some state-of-the-art models such as Basic MF and Bias MF.The users are only familiar with a few resources and the recommender system can help them learn more.The paper also conducted investigations of the cognitive abilities of 30 undergraduates in an English group and analyzed the major factors influencing their interest in learning resources.The investigation shows that the method had a positive effect on English visualaudio-oral class.The experimental results suggest that the proposed framework meets students’needs and improves their learning efficiency.
作者
王嘉琦
顾晓梅
王永祥
WANG Jia-qi;GU Xiao-mei;WANG Yong-xiang(School of Foreign Languages and Cultures,Nanjing Normal University,Nanjing,Jiangsu 210046,China)
出处
《外语电化教学》
CSSCI
北大核心
2020年第3期54-60,9,共8页
Technology Enhanced Foreign Language Education
基金
2019年度江苏省高校哲学社会科学研究一般项目“大数据背景下英文学术文献的智慧推荐策略研究”(项目编号:2019SJA0237)
2018年度南京师范大学教育教学改革项目“大学英语自主学习课程的个性化教学模式研究与实践”的阶段性研究成果。
关键词
协同推荐
混合学习
英语视听说
智慧教育
Collaborative Recommendation
Blended Learning
Visual-Aural-Oral
Smart Education
