Collaborative Research: Discovering and Exploiting Latent Communities in Social Media

尽管社交网络平台和应用程序在当今的日常生活中非常流行，但现有的社交媒体研究将“社交”和“媒体”分开使用，未能解决智能管理和利用社交媒体的重要需求，例如找到用户想要的信息，将信息置于赋予其意义的社交环境中，以及为错综复杂的关系网络提供秩序和结构。这个跨学科的项目将结合社会智能语言处理和语言动机的社会网络分析，提供社会媒体的整体视图。具体而言，该项目将：(a)发现社会语言学社区，并确定构成社区成员的人口和社会学因素；(b)发现不同层次的跨社区语言差异，开发新的计算工具，用于辩证法和社会语言学分析以及预测用户兴趣和趋势；(c)推荐跨社区的内容和社会关系，这将帮助人们通过新的信息、观点和社会关系拓宽他们的视野。智力优势：该项目将导致(a)新的建模形式，将语言信息与社交网络元数据联合起来；(b)从原始文本进行社会语言学调查的新计算方法；(c)灵活的语言变化模型，模拟时间动态，超越简单的词袋方法，进入高阶现象，如多词表达、句法和联合正字法变化。更广泛的影响：该项目将推动统计机器学习、社会科学和语言技术等基础研究的进步。它还将在所有这些领域带来创新和实际应用，比如能够智能地推断社区结构、社交媒体中的语言模式和惯例的软件。这些发现将有利于广泛的需求，如个性化信息服务、情报和安全行动，这些都需要对社会文化事件和趋势进行精确和及时的了解。该项目还将为本科生提供研究机会，并通过暑期课程向高中生提供拓展服务。

Despite the prevalence of social network platforms and apps in nowadays daily life, existing research on social media takes the terms "social" and "media" separately, and fails to address important needs for intelligently managing and utilizing social media, such as finding the information that users want, situating information in a social context that gives it meaning, and providing order and structure to an intricate and intertwined network of relationships. This interdisciplinary project will provide a holistic view of social media by combining socially intelligent language processing with linguistically motivated social network analysis. Specifically, the project will: (a) discover sociolinguistic communities and identify the demographic and sociological factors that underlie community membership; (b) discover cross-community linguistic variation at various levels and develop new computational tools for dialectometric and sociolinguistic analysis and for prediction of user interests and trends; and (c) recommend content and social connections across community boundaries, which will help people to broaden their perspectives with new information, opinions, and social relationships.Intellectual merit: The project will lead to (a) new modeling formalisms that jointly incorporate linguistic information with social network metadata; (b) a new computational methodology for sociolinguistic investigation from raw text; and (c) flexible models of linguistic variation that model temporal dynamics and move beyond simplistic bag-of-words approaches to higher-order phenomena such as multi-word expressions, syntax, and joint orthographic variation. Broader impacts: The project will lead to advancements in basic research in statistical machine learning, social sciences, and language technology. It will also bring innovations and practical applications in all these areas, such as software that reasons intelligently about community structures and linguistic patterns and conventions in social media. The findings will benefit a wide range of needs, such as personalized information service and intelligence and security operations, which require precise and timely understanding of social-cultural events and trends. The project will also provide undergraduate research opportunities and outreach to high school students through summer programs.