III: Medium: Collaborative Research: Connecting the Ephemeral and Archival Information Networks

这个合作研究项目(IIS-1160894，W.Bruce Croft，马萨诸塞大学阿默斯特分校和IIS-1160862，Jamie Callan，卡内基-梅隆大学)解决了作为社会互动一部分产生的短暂信息的复杂问题，这些信息在时间规模、数量和质量方面与网络上的档案信息不同。这个项目调查的假设是，由于所提供的上下文，使用它们之间的联系来增强对短暂或档案信息的搜索。它开发了新的检索模型和功能，用于对一系列搜索任务中的功能进行排序，从而可以利用集成的临时/档案网络。一些搜索任务是基于以前的TREC博客、微博和网络活动。它还研究了两个新的任务，会话检索和聚合社交搜索。会话检索以“对话”或“事件”的形式针对信息单元，而不是简单地检索社交帖子或网页。聚合的社交搜索基于潜在的查询意图以不同的粒度对信息进行排名，例如句子、帖子、对话或帖子。探索短暂信息和档案信息之间联系的研究需要包含这两种类型信息的数据集。该项目的一个重要部分是使用链接到Web数据的临时微博、博客和论坛数据来扩展存档ClueWeb12数据集。该扩展作为ClueWeb12++数据集分发给研究社区。这个(http://ciir.cs.umass.edu/research/ephemeral/)项目是第一个解决搜索的全部可能性的项目，它利用了通过将两个信息“世界”结合在一起而创建的所有联系和上下文。它还开发和分发一个独特的新数据集，该数据集支持开发新一代工具以获取广泛的信息。合作机构、马萨诸塞大学阿默斯特分校和卡内基-梅隆大学的学生将参与教育活动，并从研究经验中受益。

This collaborative research project (IIS-1160894, W. Bruce Croft, University of Massachusetts Amherst and IIS-1160862, Jamie Callan, Carnegie-Mellon University) addresses the complex issues of ephemeral information that is generated as part of social interactions is different in terms of time scale, quantity, and quality to archival information found on the web. This project investigates the hypothesis that, because of the context provided, searching either ephemeral or archival information is enhanced using the connections between them. It develops new retrieval models and features for ranking functions in a range of search tasks that can exploit an integrated ephemeral/archival network. Some search tasks are based on previous TREC blog, microblog, and web activities. It also investigates two new tasks, conversation retrieval and aggregated social search. Conversation retrieval targets information units in the form of "conversations" or "events" instead of simply retrieving social postings or web pages. Aggregated social search ranks information in different granularities, such as sentence, posting, conversation, or thread, based on the underlying query intent. Research that explores the connections between ephemeral and archival information requires a dataset that contains both types of information. A crucial part of this project extends the archival ClueWeb12 dataset with ephemeral microblog, blog, and discussion forum data that links to the web data. This extension is distributed to the research community as the ClueWeb12++ dataset. This project (http://ciir.cs.umass.edu/research/ephemeral/) is the first to address the full possibilities of search that exploits all the connections and contexts created by bringing together the two "worlds" of information. It also develops and distributes a unique new dataset that supports the development of a new generation of tools to access a broad range of information. Students at collaborating institutions, University of Massachusetts Amherst and Carnegie-Mellon University will be involved in educational activities and benefit from research experience.