流动注射分析专家系统的自然语言理解

本文提出自然语言理解技术作为流动注射分析专家系统的人机交互模式,并对自然语言理解程序的三个组成部分—词法分析、语法分析和语义分析的具体实现方法进行了阐述。同时说明了如何用PROLOG语言设计流动注射分析专家系统的自然语言理解模块。     

Texting insincerely: The role of the period in text messaging

Text messaging is one of the most frequently used computer-mediated communication (CMC) methods. The rapid pace of texting mimics face-to-face communication, leading to the question of whether the critical non-verbal aspects of conversation, such as tone, are expressed in CMC. Much of the research in this domain has involved large corpus analyses, focusing on the contents of texts, but not how receivers comprehend texts. We ask whether punctuation – specifically, the period – may serve as a cue for pragmatic and social information. Participants read short exchanges in which the response either did or did not include a sentence-final period. When the exchanges appeared as text messages, the responses that ended with a period were rated as less sincere than those that did not end with a period. No such difference was found for handwritten notes. We conclude that punctuation is one cue used by senders, and understood by receivers, to convey pragmatic and social information.     

How attitude strength biases information processing and evaluation on the web

In this study we examined the influence of attitude strength on the processing and evaluation of sources and information on the Web. Seventy-nine German university students participated in an eye-tracking study in which they read information from eight pre-selected websites from different sources on the controversial topic of organic foods. Results showed that participants who felt strongly about the topic (i.e., those with strong prior attitude strength) scrutinized website logos of attitude-inconsistent websites shorter and judged the credibility of attitude-inconsistent websites lower. They also included more attitude-consistent information in an essay task than participants with weaker prior attitudes. Participants who felt less strongly about the topic (i.e., those with weaker prior attitudes) fixated longer on text from attitude-inconsistent websites than from attitude-consistent websites. By contrast, for participants with strong prior attitudes the time spent on the texts from attitude-consistent websites and attitude-inconsistent websites did not differ significantly. The results show that prior attitudes can bias evaluation and processing of information in different ways. Even though participants were not fully biased during initial information processing, they were so when evaluating the information and presenting it in an essay task.     

Recording concerns in source code using annotations

A concern can be characterized as a developer׳s intent behind a piece of code, often not explicitly captured in it. We discuss a technique of recording concerns using source code annotations (concern annotations). Using two studies and two controlled experiments, we seek to answer the following 3 research questions: (1) Do programmers׳ mental models overlap? (2) How do developers use shared concern annotations when they are available? (3) Does using annotations created by others improve program comprehension and maintenance correctness, time and confidence? The first study shows that developers׳ mental models, recorded using concern annotations, overlap and thus can be shared. The second study shows that shared concern annotations can be used during program comprehension for the following purposes: hypotheses confirmation, feature location, obtaining new knowledge, finding relationships and maintenance notes. The first controlled experiment with students showed that the presence of annotations significantly reduced program comprehension and maintenance time by 34%. The second controlled experiment was a differentiated replication of the first one, focused on industrial developers. It showed a 33% significant improvement in correctness. We conclude that concern annotations are a viable way to share developers׳ thoughts.     

机器阅读理解中观点型问题的求解策略研究

针对机器阅读理解中观点型问题的求解,提出一个端到端深度学习模型,使用Bi-GRU对文章和问题进行上下文语义编码,然后运用基于拼接、双线性、点乘和差集4种函数的注意力加上Query2Context和Context2Query两个方向注意力的融合算法获取文章和问题的综合语义信息,之后运用多层注意力转移推理机制不断聚焦,进一步获取更加准确的综合语义,最终将其与候选答案进行比较,选出正确答案。该模型在AIchallager2018观点型阅读理解中文测试数据集上准确率达到76.79%,性能超过基线系统。此外,该文尝试文章以句子序列作为输入表示进行答案求解,准确率达到78.48%,获得较好试验效果。     

Preference for newspaper size

The past few years has seen a change in the size of newspapers, with publishers moving to a smaller size format. Five ‘standard’ newspaper sizes are used in different countries: Broadsheet, Rhensch, Tabloid, Tall Tabloid and Berliner. These papers vary in both width and height of pages and hence there are implications for human reading comfort, which may be dependent on reading location such as on a lounge chair or on a train. Experiments were carried out to determine preferences for the different sizes and to relate these preferences to the geometric characteristics of the newspapers. For both comfortable and cramped/uncomfortable reading conditions, the rank order of preference for paper types was, from least to most-preferred, Broadsheet, Rhensch, Berliner, Tall Tabloid and Tabloid. Preferences were much stronger when determined in cramped/uncomfortable reading conditions, where most comparisons were significantly different. There was good correlation between participant ratings on several scales and preference, where most factors were related to comfort of holding and controlling the paper.     

非监督、多级嘴唇区域分割方法

该文介绍了一个非监督、多级嘴唇区域分割检测方法。首先提出利用fisher变换增强嘴唇区域,然后利用嘴唇在人脸区域的分布面积比,提出利用统计阈值完成嘴唇的初步分割,最后利用嘴唇对称性和轮廓光滑性的先验知识,提出了基于局部阈值调整完成嘴唇的精细分割。实验证明该方法在不同光照条件下、对不同人、各种表情都能自动地、鲁棒地、精确地检测出嘴唇区域,并较准确提取初步的嘴唇几何参数。利用这些几何参数作为轮廓定位的初始化条件,可以大大提高变形模板和ASM模型等嘴唇轮廓定位算法的速度和准确度。     

Reading electronic books as a support for vocabulary,story comprehension and word reading in kindergarten and first grade

The effect of reading an electronic storybook (e-book) on Israeli children’s language and literacy was examined in kindergarten children (N = 40; age 5:2–6:3) compared to first graders (N = 50; age 6:3–7:4). The children in each age group were randomly assigned to two groups: an intervention group which read the e-book five times and a control group which was afforded the regular school program. Pre- and post-tests included vocabulary and word reading measures. Post-tests included story comprehension and production. Children who read the e-book exhibited significant progress in word meaning and word reading compared to the control group. Kindergarten children progressed in word reading more significantly than first graders across treatment groups. This could be explained by the ceiling effect of the first graders’ word reading level which did not leave much room for progress in this skill compared to the kindergarten children. No interaction was found between age and treatment groups. Kindergarten children exhibited a good level of story comprehension, similar to first graders, although their story production was lower. Implications for future research and education are discussed.     

基于受限自然语言和模块组合的代码自动生成

针对界面层自然语言理解的应用,该文设计一个基于受限自然语言和模块组合的代码自动生成系统。该系统允许用户输入,格式比较随意的中文需求语句,,把用户的输入转换为标准的C程序,再返回给用户。提出一种IMM分词算法,改进原有MM法在局部的最大匹配,减少错误切分句子的数量,提高分词的精度。     

The Package Blueprint: Visually analyzing and quantifying packages dependencies

Large object-oriented applications are structured over many packages. Packages are important but complex structural entities that are difficult to understand since they act as containers of classes, which can have many dependencies with other classes spread over multiple packages. However to be able to take decisions (e.g. refactoring and/or assessment decisions), maintainers face the challenges of managing (sorting, grouping) the massive amount of dependencies between classes spread over multiple packages. To help maintainers, there is a need for at the same time understanding, and quantifying, dependencies between classes as well as understanding how packages as containers of such classes depend on each other.In this paper, we present a visualization, named Package Blueprint, that reveals in detail package internal structure, as well as the dependencies between an observed package and its neighbors, at both package and class levels. Package blueprint aims at assisting maintainers in understanding package structure and dependencies, in particular when they focus on few packages and want to take refactoring decisions and/or to assess the structure of those packages. A package blueprint is a space filling matrix-based visualization, using two placement strategies that are enclosure and adjacency. Package blueprint is structured around the notion of surfaces that group classes and their dependencies by their packages (i.e., enclosure placement); whilst surfaces are placed next to their parent node which is the package under-analysis (i.e., adjacency placement). We present two views: one stressing how an observed package depends upon the rest of the system and another stressing how the system depends upon that package.To evaluate the contribution of package blueprint for understanding packages we performed an exploratory user study comparing package blueprint with an advanced IDE. The results show that users of package blueprint are faster in analyzing and assessing package structure. The results are proved statically significant and they show that package blueprint considerably improves the experience of standard browser users.