Purpose–The purpose of this paper is to investigate temporal tweet patterns and their effectiveness in predicting the financial performance of a movie.Specifically,how tweet patterns are formed prior to and after a m...Purpose–The purpose of this paper is to investigate temporal tweet patterns and their effectiveness in predicting the financial performance of a movie.Specifically,how tweet patterns are formed prior to and after a movie’s release and their usefulness in predicting a movie’s success is explored.Design/methodology/approach–Volume was measured and sentiment analysis was performed on a sampleof Tweetsposted fourdays beforeand afterthe releaseof 86 movies.Thetemporal patternof tweeting for financially successful movies was compared with those that were financial disappointments.Using temporal tweet patterns,a number of machine learning models were developed and their predictive performance was compared.Findings–Results show that the temporal patterns of tweet volume,length and sentiment differ between“hits”and“busts”in the days surrounding their releases.Compared with“busts”the tweet pattern for“hits”reveal higher volume,shorter length,and more favourable sentiment.Discriminant patterns in social media features occur days in advance of a movie’s release and can be used to develop models for predicting a movie’s success.Originality/value–Analysis of temporal tweet patterns and their usefulness in predicting box office returns is the main contribution of this research.Results of this research could lead to development of analyticaltools allowingmotionpicture studiosto accurately predictand possiblyinfluencethe opening night box-office receipts prior to the release of the movie.Also,the specific temporal tweet patterns presented by this work may be applied to problems in other areas of research.展开更多
随着自然语言处理(NLP,natural language processing)技术的快速发展,语言模型在文本分类和情感分析中的应用不断增加。然而,语言模型容易遭到盗版再分发,对模型所有者的知识产权造成严重威胁。因此,研究者着手设计保护机制来识别语言...随着自然语言处理(NLP,natural language processing)技术的快速发展,语言模型在文本分类和情感分析中的应用不断增加。然而,语言模型容易遭到盗版再分发,对模型所有者的知识产权造成严重威胁。因此,研究者着手设计保护机制来识别语言模型的版权信息。现有的适用于文本分类任务的语言模型水印无法与所有者身份相关联,且鲁棒性不足以及无法再生成触发集。为了解决这些问题,提出一种新的适用于文本分类任务模型的黑盒水印方案,可以远程快速验证模型所有权。将模型所有者的版权消息和密钥通过密钥相关的哈希运算消息认证码(HMAC,hash-based message authentication code)得到版权消息摘要,由HMAC得到的消息摘要可以防止被伪造,具有很强的安全性。从原始训练集各个类别中随机挑选一定的文本数据,将摘要与文本数据结合构建触发集,并在训练过程中对语言模型嵌入水印。为了评估水印的性能,在IMDB电影评论、CNEWS中文新闻文本分类数据集上对3种常见的语言模型嵌入水印。实验结果表明,在不影响原始模型测试精度的情况下,所提出的水印验证方案的准确率可以达到100%。即使在模型微调和剪枝等常见攻击下,也能表现出较强的鲁棒性,并且具有抗伪造攻击的能力。同时,水印的嵌入不会影响模型的收敛时间,具有较高的嵌入效率。展开更多
The design of this paper is to present the first installment of a complete and final theory of rational human intelligence. The theory is mathematical in the strictest possible sense. The mathematics involved is stric...The design of this paper is to present the first installment of a complete and final theory of rational human intelligence. The theory is mathematical in the strictest possible sense. The mathematics involved is strictly digital—not quantitative in the manner that what is usually thought of as mathematics is quantitative. It is anticipated at this time that the exclusively digital nature of rational human intelligence exhibits four flavors of digitality, apparently no more, and that each flavor will require a lengthy study in its own right. (For more information,please refer to the PDF.)展开更多
Logic flaws within web applications will allow malicious operations to be triggered towards back-end database. Existing approaches to identifying logic flaws of database accesses are strongly tied to structured query ...Logic flaws within web applications will allow malicious operations to be triggered towards back-end database. Existing approaches to identifying logic flaws of database accesses are strongly tied to structured query language （SQL） statement construction and cannot be applied to the new generation of web applications that use not only structured query language （NoSQL） databases as the storage tier. In this paper, we present Lom, a black-box approach for discovering many categories of logic flaws within MongoDB- based web applications. Our approach introduces a MongoDB operation model to support new features of MongoDB and models the application logic as a mealy finite state machine. During the testing phase, test inputs which emulate state violation attacks are constructed for identifying logic flaws at each application state. We apply Lom to several MongoDB-based web applications and demonstrate its effectiveness.展开更多
文摘Purpose–The purpose of this paper is to investigate temporal tweet patterns and their effectiveness in predicting the financial performance of a movie.Specifically,how tweet patterns are formed prior to and after a movie’s release and their usefulness in predicting a movie’s success is explored.Design/methodology/approach–Volume was measured and sentiment analysis was performed on a sampleof Tweetsposted fourdays beforeand afterthe releaseof 86 movies.Thetemporal patternof tweeting for financially successful movies was compared with those that were financial disappointments.Using temporal tweet patterns,a number of machine learning models were developed and their predictive performance was compared.Findings–Results show that the temporal patterns of tweet volume,length and sentiment differ between“hits”and“busts”in the days surrounding their releases.Compared with“busts”the tweet pattern for“hits”reveal higher volume,shorter length,and more favourable sentiment.Discriminant patterns in social media features occur days in advance of a movie’s release and can be used to develop models for predicting a movie’s success.Originality/value–Analysis of temporal tweet patterns and their usefulness in predicting box office returns is the main contribution of this research.Results of this research could lead to development of analyticaltools allowingmotionpicture studiosto accurately predictand possiblyinfluencethe opening night box-office receipts prior to the release of the movie.Also,the specific temporal tweet patterns presented by this work may be applied to problems in other areas of research.
文摘The design of this paper is to present the first installment of a complete and final theory of rational human intelligence. The theory is mathematical in the strictest possible sense. The mathematics involved is strictly digital—not quantitative in the manner that what is usually thought of as mathematics is quantitative. It is anticipated at this time that the exclusively digital nature of rational human intelligence exhibits four flavors of digitality, apparently no more, and that each flavor will require a lengthy study in its own right. (For more information,please refer to the PDF.)
基金supported by China Scholarship Council,Tianjin Science and Technology Committee(No.12JCZDJC20800)Science and Technology Planning Project of Tianjin(No.13ZCZDGX01098)+2 种基金NSF TRUST(The Team for Research in Ubiquitous Secure Technology)Science and Technology Center(No.CCF-0424422)National High Technology Research and Development Program of Chia(863Program)(No.2013BAH01B05)National Natural Science Foundation of China(No.61402264)
文摘Logic flaws within web applications will allow malicious operations to be triggered towards back-end database. Existing approaches to identifying logic flaws of database accesses are strongly tied to structured query language （SQL） statement construction and cannot be applied to the new generation of web applications that use not only structured query language （NoSQL） databases as the storage tier. In this paper, we present Lom, a black-box approach for discovering many categories of logic flaws within MongoDB- based web applications. Our approach introduces a MongoDB operation model to support new features of MongoDB and models the application logic as a mealy finite state machine. During the testing phase, test inputs which emulate state violation attacks are constructed for identifying logic flaws at each application state. We apply Lom to several MongoDB-based web applications and demonstrate its effectiveness.
