An experiment on the role of graphical elements in architecture visualization

The evolution and maintenance of large-scale software systems requires first an understanding of its architecture before delving into lower-level details. Tools facilitating the architecture comprehension tasks by visualization provide different sets of configurable, graphical elements to present information to their users. We conducted a controlled experiment that exemplifies the critical role of such graphical elements when aiming at understanding the architecture. In our setting, a different configuration of graphical elements had significant influence on program comprehension tasks. In particular, a 63% gain in effectiveness in architectural analysis tasks was achieved simply by changing the configuration of the graphical elements of the same tool. Based on the results, we claim that significant effort should be spent on the configuration of architecture visualization tools and that configurability should be a requirement for such tools.

社交网络智慧搜索研究进展与发展趋势

在以社交网络为代表的Web 2.0时代,搜索需求已经远超出Web 1.0搜索引擎的能力范畴,数据具有模式多态、快速产生、动态交互、碎片化、变化演绎等特点,这些新特点给搜索引擎技术带来了革命性和颠覆性的挑战。归纳了在线社交网络智慧搜索研究进展和技术要点,分析了研究发展过程将面临3个主要问题：在线社交网络中智慧与知识的发掘与推演、用户真实搜索意图的理解与表示以及满足用户真实意图的智慧解答在线响应,并以解决此3个问题为目标讨论了相关技术发展。     

基于UIO序列的类重要性度量

程序理解是测试和维护大规模面向对象程序的关键,选择程序的关键类优先开展分析是理解程序结构的一个有效的方法.为支持自动识别软件系统中的关键类,本文提出了一种基于UIO序列的类重要性度量方法.首先将软件系统抽象为一个以类为转换的有限自动机模型,随后求解该自动机的UIO序列,将该序列集合转化为状态转换树.通过递归计算状态转换树的节点复杂度求得类重要度.并在考虑异常传播的基础上改进了算法.最后通过实验验证了算法的有效性.     

MRC-PBM: 一种中文电子病历嵌套命名 实体识别方法

中文电子病历实体包含大量的医学领域词汇并具有明显的嵌套特征。嵌套实体识别时往往存在目标实体定位不完整、不准确的问题。针对这一问题,提出了一种基于机器阅读理解的中文电子病历嵌套命名实体识别模型MRC-PBM (machine reading comprehension-position information biaffine and MLP)。该模型将命名实体识别(named entity recognition, NER)转化为机器阅读理解任务,将中文电子病历文本和预定义的查询语句串联作为输入,使用基于医学的预训练模型MC＿BERT获取词向量,然后通过双向长短期记忆网络模型(BiLSTM)和多粒度扩张卷积模型分别获取双向的特征信息以及单词之间的信息,得到相应的特征向量,最后使用Hybrid-PBM预测器进行实体预测。在嵌套和平面NER数据集上进行实验。实验表明,该模型在糖尿病语料和公开医学数据集上优于其他主流神经网络模型,F1值比基线模型提高了1.21%～5.80%。     

Juror comprehension and public policy: Perceived problems and proposed solutions.

Laypersons, the media, and many legal scholars tend to attribute problems in the jury system to the dispositions of individual jurors and to recommend reforms in jury selection procedures and relaxation of the unanimity rule. Social scientists view problems as a consequence of the structure of the jurors' task and recommend reforms in trial procedures. After years of apathy, the legal system has proposed, and in some jurisdictions implemented, a variety of reforms, most of which are based on the social science perspective that the problem is not due to bad jurors but to unnecessary procedural obstacles to high-quality decision making. These reforms are described in the final section of the article. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

An Improved Algorithm for Identifying Objects in Code

The identification of abstractions within existing software systems is an important problem to be solved to facilitate program comprehension and the construction of a set of reusable artifacts. In particular, of interest is the identification of object-like features in procedural programs. Existing techniques and algorithms achieve some level of success but do not, in general, always precisely identify a coherent set of objects. The identified objects tend to contain spurious methods that are only tenuously related to the object and require a great deal of human effort and understanding to unravel. This paper presents an improved algorithm that overcomes these drawbacks and enables the precise identification of objects with less human intervention and understanding by exploiting simple statistical techniques. The algorithm is applied to several sample programs and the results are compared with existing algorithms. Finally, the application of the algorithm to a real medium-size system is described and discussed. The algorithm was developed as part of the RE² project in which the identification of object-like features in existing systems is the basis for a re-engineering process aimed at populating a repository of reusable assets.     

Embedding reading comprehension training in content-area instruction.

This study evaluated the effectiveness of comprehension training embedded in a program that taught science content to 2nd graders. The program included instruction about the structure of compare-contrast expository text, emphasizing clue words, generic questions, graphic organizers, and the close analysis of well-structured text exemplars. This program was compared with a program that focused on the science content but included no compare-contrast training as well as with a no-instruction control. Regular classroom teachers (14 from 4 schools), randomly assigned to treatment, provided the instruction; 215 students (7-8 years old) participated. The study replicated acquisition and transfer effects found in an earlier study, that is, transfer to compare-contrast text with content related and unrelated to the instructional content (with no loss in the amount of science content acquired). The program also led to better performance on written and oral response measures and on 1 of the 2 measures involving authentic (less well-structured) compare-contrast text. These findings support and extend previous findings that explicit instruction in comprehension is effective as early as the primary-grade level. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Text skimming: The process and effectiveness of foraging through text under time pressure.

Is Skim reading effective? How do readers allocate their attention selectively? The authors report 3 experiments that use expository texts and allow readers only enough time to read half of each document. Experiment 1 found that, relative to reading half the text, skimming improved memory for important ideas from a text but did not improve memory of less important details or of inferences made from information within the text. Experiment 2 found no advantage of skimming over reading the first or second half of every paragraph. Two final experiments using a hierarchical, Website-like layout of documents showed that the advantage of skimming found in Experiment 1 was dependent on the linkages between pages and, thus, the ease with which participants could navigate through the text. Data on page-by-page reading times and eye-tracking analyses from Experiment 2 indicated that Skim readers spent more time reading text that was earlier in the paragraph, toward the top of the page and in an earlier page of the document. These findings were interpreted as evidence in support of a “satisficing” account of skimming process. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

“The contributions of working memory and executive functioning to problem representation and solution generation in algebraic word problems”: Correction.

Reports an error in "The contributions of working memory and executive functioning to problem representation and solution generation in algebraic word problems" by Kerry Lee, Ee Lynn Ng and Swee Fong Ng (Journal of Educational Psychology, 2009[May], Vol 101[2], 373-387). In this article, the URL published for the supplemental material was incorrect. The correct URL is http://dx.doi.org/10.1037/a0013843.supp. (The following abstract of the original article appeared in record 2009-04640-010.) Solving algebraic word problems involves multiple cognitive phases. The authors used a multitask approach to examine the extent to which working memory and executive functioning are associated with generating problem models and producing solutions. They tested 255 11-year-olds on working memory (Counting Recall, Letter Memory, and Keep Track), ability to inhibit inappropriate responses (inhibition: numeric Stroop, Stop Signal), mental flexibility (switching: Number–Letter and Plus–Minus), English literacy, and algebraic problem-solving skills (problem representation, solution generation, and other subcomponents). Working memory explained about a quarter of the variance in both representation and solution formation. Literacy explained an additional 20% of the variance in representation formation. Ability to discern quantitative relationships explained an additional 10%. The findings go beyond a demonstration of an association between working memory and problem-solving accuracy. They show that success in word problems is particularly reliant on ability to decode and assign mathematical operators to quantitative relationships, 2 phases of problem solving that also draw heavily on working memory resources. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于自适应知识选择的机器阅读理解

目前针对知识增强机器阅读理解的研究主要集中在如何把外部知识融入现有的机器阅读理解模型,却忽略了对外部知识的来源进行选择。该文首先基于注意力机制对外部知识进行编码,然后对不同来源的外部知识编码进行打分,最后自适应地选择出对回答问题最有帮助的知识。与基线模型相比,该文提出的基于自适应知识选择的机器阅读理解模型在准确率上提高了1.2个百分点。