There are both pedagogical and theoretical grounds for asking second language writers to plan before they start writing.The question then arises whether pre-task planning(PTP)improves written output.To address this qu...There are both pedagogical and theoretical grounds for asking second language writers to plan before they start writing.The question then arises whether pre-task planning(PTP)improves written output.To address this question,this article reviewed 32 studies by comparing the effect of PTP either with no planning or with unpressured online planning(OLP).These studies also investigated the moderating effect of variables relating to the writer participants,the nature of the planning,and the writing tasks.The main findings are:(1)There is no clear evidence that PTP leads to better overall writing quality when this is measured using rating rubrics,(2)PTP generally results in more fluent writing,(3)its impact on syntactical and lexical complexity is inconsistent and negligible,(4)OLP does sometimes result in increased linguistic accuracy,and(5)there is insufficient evidence to reach clear conclusions about the role that moderating variables have on the impact of PTP,but the results suggest that collaborative(as opposed to individual planning)can lead to increased accuracy and that PTP tends to lead to more complex language when the writing task is a complex one.The article concludes with a set of principles to ensure better quality research and three general proposals for the kind of future research needed.展开更多
This study aims at investigating the effects of the pre-task planning mode on listening comprehension of Chinese EFL （English as a foreign language） learners as well as students＇ opinions towards the pre-task plann...This study aims at investigating the effects of the pre-task planning mode on listening comprehension of Chinese EFL （English as a foreign language） learners as well as students＇ opinions towards the pre-task planning mode Three step~., namely, concept mapping, strategic planning, as well as rehearsal, are employed in the pre-task planning mode in the present study. There are altogether 40 second-year English majors participated in the present study, and they are randomly assigned into the experimental group and the control group. The instruments employed in the present study to achieve the research purposes include the listening comprehension test, the teacher＇s log, the students＇ opinion questionnaires, the students＇ open-ended questionnaires, and the semi-structured interviews. Research findings revealed that there were significant main effects on the listening comprehen,;ion of Chinese EFL learners after the training by the pre-task planning mode. Moreover, the students held positive attitudes towards the pre-task planning mode展开更多
This paper studies the effect of different lengths of pre-task planning time on the frequency and accuracy of self-repairs for Chinese intermediate and advanced English learners.The findings reveal that the increased ...This paper studies the effect of different lengths of pre-task planning time on the frequency and accuracy of self-repairs for Chinese intermediate and advanced English learners.The findings reveal that the increased pre-task planning time strongly improves fluency and accuracy of self-repairs for both lexical and syntactic errors in the advanced group,but enhanced fluency and accuracy are not witnessed in the intermediate group as planning time increases.The differences are mainly due to the fact that some intermediate learners are not well equipped with the appropriate way of preparation for oral presentation tasks since some write down their ideas in Chinese.Thus errors will increasingly emerge in the transfer.Besides,they usually give priority to accuracy while neglecting fluency.展开更多
多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问...多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.展开更多
对话状态追踪(DST)是任务型对话系统中一个重要的模块,但现有的基于开放词表的DST模型没有充分利用槽位的相关信息以及数据集本身的结构信息。针对上述问题,提出基于槽位相关信息提取的DST模型SCELDST(SCE and LOW for Dialogue State T...对话状态追踪(DST)是任务型对话系统中一个重要的模块,但现有的基于开放词表的DST模型没有充分利用槽位的相关信息以及数据集本身的结构信息。针对上述问题,提出基于槽位相关信息提取的DST模型SCELDST(SCE and LOW for Dialogue State Tracking)。首先,构建槽位相关信息提取器(SCE),利用注意力机制学习槽位之间的相关信息;然后,在训练过程中应用学习最优样本权重(LOW)策略,在未大幅增加训练时间的前提下,加强模型对数据集信息的利用;最后,优化模型细节,搭建完整的SCEL-DST模型。实验结果表明,SCE和LOW对SCEL-DST模型性能的提升至关重要,该模型在两个实验数据集上均取得了更高的联合目标准确率,其中在MultiWOZ 2.3(Wizard-of-OZ 2.3)数据集上与相同条件下的TripPy(Triple coPy)相比提升了1.6个百分点,在WOZ 2.0(Wizard-of-OZ 2.0)数据集上与AG-DST(Amendable Generation for Dialogue State Tracking)相比提升了2.0个百分点。展开更多
文摘There are both pedagogical and theoretical grounds for asking second language writers to plan before they start writing.The question then arises whether pre-task planning(PTP)improves written output.To address this question,this article reviewed 32 studies by comparing the effect of PTP either with no planning or with unpressured online planning(OLP).These studies also investigated the moderating effect of variables relating to the writer participants,the nature of the planning,and the writing tasks.The main findings are:(1)There is no clear evidence that PTP leads to better overall writing quality when this is measured using rating rubrics,(2)PTP generally results in more fluent writing,(3)its impact on syntactical and lexical complexity is inconsistent and negligible,(4)OLP does sometimes result in increased linguistic accuracy,and(5)there is insufficient evidence to reach clear conclusions about the role that moderating variables have on the impact of PTP,but the results suggest that collaborative(as opposed to individual planning)can lead to increased accuracy and that PTP tends to lead to more complex language when the writing task is a complex one.The article concludes with a set of principles to ensure better quality research and three general proposals for the kind of future research needed.
文摘This study aims at investigating the effects of the pre-task planning mode on listening comprehension of Chinese EFL （English as a foreign language） learners as well as students＇ opinions towards the pre-task planning mode Three step~., namely, concept mapping, strategic planning, as well as rehearsal, are employed in the pre-task planning mode in the present study. There are altogether 40 second-year English majors participated in the present study, and they are randomly assigned into the experimental group and the control group. The instruments employed in the present study to achieve the research purposes include the listening comprehension test, the teacher＇s log, the students＇ opinion questionnaires, the students＇ open-ended questionnaires, and the semi-structured interviews. Research findings revealed that there were significant main effects on the listening comprehen,;ion of Chinese EFL learners after the training by the pre-task planning mode. Moreover, the students held positive attitudes towards the pre-task planning mode
文摘This paper studies the effect of different lengths of pre-task planning time on the frequency and accuracy of self-repairs for Chinese intermediate and advanced English learners.The findings reveal that the increased pre-task planning time strongly improves fluency and accuracy of self-repairs for both lexical and syntactic errors in the advanced group,but enhanced fluency and accuracy are not witnessed in the intermediate group as planning time increases.The differences are mainly due to the fact that some intermediate learners are not well equipped with the appropriate way of preparation for oral presentation tasks since some write down their ideas in Chinese.Thus errors will increasingly emerge in the transfer.Besides,they usually give priority to accuracy while neglecting fluency.
文摘多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.
文摘对话状态追踪(DST)是任务型对话系统中一个重要的模块,但现有的基于开放词表的DST模型没有充分利用槽位的相关信息以及数据集本身的结构信息。针对上述问题,提出基于槽位相关信息提取的DST模型SCELDST(SCE and LOW for Dialogue State Tracking)。首先,构建槽位相关信息提取器(SCE),利用注意力机制学习槽位之间的相关信息;然后,在训练过程中应用学习最优样本权重(LOW)策略,在未大幅增加训练时间的前提下,加强模型对数据集信息的利用;最后,优化模型细节,搭建完整的SCEL-DST模型。实验结果表明,SCE和LOW对SCEL-DST模型性能的提升至关重要,该模型在两个实验数据集上均取得了更高的联合目标准确率,其中在MultiWOZ 2.3(Wizard-of-OZ 2.3)数据集上与相同条件下的TripPy(Triple coPy)相比提升了1.6个百分点,在WOZ 2.0(Wizard-of-OZ 2.0)数据集上与AG-DST(Amendable Generation for Dialogue State Tracking)相比提升了2.0个百分点。
