DC: Small: Cross-Language Bayesian Models for Web-Scale Text Analysis Using MapReduce

网络提供了前所未有的机会，让我们可以接触到无数人对各种各样问题的看法。要将这些尚未被驯服的杂音转化为有意义的见解，需要处理语言的多样性和网络的规模。目前大多数的研究都集中在专门的任务上，比如跟踪消费者的意见，而且几乎所有的研究都将网络视为单一的单一语言，忽略了所代表的语言的多样性以及所讨论的主题和问题之间丰富的相互作用。该项目通过关注两个关键挑战来推动技术的发展。首先，高度可扩展的MapReduce算法在贝叶斯框架内进行语言建模，利用变分推理在web规模的数据集上实现高度并行化。第二，新颖的贝叶斯模型，学习跨语言文本的一致解释和广泛的感兴趣的响应变量（例如，对一个问题的看法，相对于一个事件的情感强度，以及注意力的焦点）。本项目中开发的技术将在Web页面和博客的大型抓取上进行演示。这些技术的潜在应用包括帮助小学生了解不同国家的人对某些问题的看法可能非常不同，帮助政治家了解选民对拟议立法的反应，或帮助情报分析人员了解敌对国家的民意如何演变。欲了解更多信息，请参阅项目网页：http://www.umiacs.umd.edu/~jimmylin/cloud-computing

The Web promises unprecedented access to the perspectives of anenormous number of people on a wide range of issues. Turning thatstill untamed cacophony into meaningful insights requires dealing withthe linguistic diversity and scale of the Web. Most current researchfocuses on specialized tasks such as tracking consumer opinions, andvirtually all current research treats the Web as both monolithic andmonolingual, ignoring the variety of languages represented and therich interplay between topics and issues under discussion.This project moves the state of the art forward by focusing on two keychallenges. First, highly-scalable MapReduce algorithms forlinguistic modeling within a Bayesian framework, making use ofvariational inference to achieve a high degree of parallelization onWeb-scale datasets. Second, novel Bayesian models that learnconsistent interpretations of text across languages and a wide rangeof response variables of interest (for example, views on an issue,strength of emotion relative to an event, and focus of attention).The techniques developed in this project will be demonstrated on largecrawls of Web pages and blogs. Potential applications for thesetechnologies include helping a schoolchild learn that people indifferent countries may view some issues very differently, helping apolitician understand how constituents are reacting to proposedlegislation, or helping an intelligence analyst understand how publicopinion is evolving in a hostile country.For further information see the project Web page:http://www.umiacs.umd.edu/~jimmylin/cloud-computing