面向协作生成服务的社交搜索研究

The rapid development of Web 2.0 brings forward User Generated Service (UGS), a new concept that aims to meet non-skillful users' needs of personalized service generation. It stimulates the requirement on the network tool to support UGS based on the socially collaboration among online social network which is named as Collaborative User Generated Service (Co-UGS). The core functionalities of network tool for Co-UGS are collaboration task propagation and collaboration groups identification both of which belong to social search. The research on social computing mechanism for social collaboration relationship mining is fundamental for establishment and optimization of social collaboration tools orienting to Co-UGS. However, there is only few efforts on social search and it computing mechanisms for Co-UGS. The objective of the porject is to explore the mechanisms of social search as well as its underlying social computing for Co-UGS, achieving the understand of paradigm on how collaboration group in Co-UGS system can achieve high-quality collaboration output. By decompose social computing mechanisms for social search of Co-UGS into several parts, the main issues of the projects are set as 1)behavior analysis and modeling for multi-user collaboration of Co-UGS; 2)collaboration group identification for Co-UGS; 3)estimation with prediction on collaboration as well as its organization mechanism. The main goal of the project is to effectively mine and organize the collaboration based on online social network users, and to bring in theoratical guidance for the establishment of collaborative innovation platforms as well as their mechanisms on ubiquitous services for non-skillful users of online social networks.

互联网2.0将用户放到服务生成愿景的中心，催生了用户生成服务（User Generated Service, UGS）的出现，也对支持社会化协作生成服务(Collaborative UGS Co-UGS)的网络工具提出了需求。负责协作任务发布以及对协作关系组织的社交搜索，是Co-UGS协作工具的核心功能，而面向社会协作关系挖掘的社会计算机理研究是建立和完善社会化协作工具的基石，而与Co-UGS相关的计算机理研究还存在空白。申请人团队以Co-UGS的社交搜索为研究对象，从复杂网络科学理论及分析方法出发，探索面向Co-UGS高质量产出的协作关系模式，并研究其社会计算机理，研究多用户网络协作行为及建模、面向协作关系的团队甄别算法、协作预测评估与任务组织机制，实现对高效协作关系的挖掘和组织，为服务创新的在线社会协作平台的设计以及运行机制的探索提供理论指导。

Web2.0时代，涌现出诸多用户生成内容场景，也提供了“无组织”的普通用户协作生成成果的服务(Collaborative User Generated Services, Co-UGS)的服务机会。针对社交网络上更有效地组织“草根协作”以产生可观群体智慧价值的需求，课题对社交网络协作依托社交计算机理开展研究。主要研究内容如下：..1）研究了社交网络用户社交圈层结构模型、内容主题和人格、不同主题的社交传播力等行为特征及建模方法。提出了吸引子平衡指数与强弱链平衡指数并推导电信社交数据上自我中心网络临界规模，支持了顿巴数假说；基于用户移动行为社交相关性在时空维度的影响方式的发现，提出考虑空时社交影响的移动签到行为预测模型；基于用户上网偏好的社交强度相关性的验证，提出基于矩阵分解的社交亲近程度规则化项约束的偏好预测模型。.2）基于社交关系和用户属性，研究可支持潜在协作群体甄别或任务/内容推荐的相似性模型方法。提出了一种既考虑与用户偏好相关的小范围信息，又考虑了全局的用户行为表现的相似度模型，以改善冷启动的性能；对用户的在线评分效果进行了分析，从而对偏好的显式量化不一致的现象提供解释；提出了基于语义轨迹来进行特征模式表达的方案，并通过语义轨迹相似性来挖掘行为背后的社交网络关系；提出了一种随机路径采样的方法RSSim，用来解决大规模异质网络中对象间的相似度搜索问题。.3）研究面向协作的预测和分派组织机制。基于回音壁效应homophily提出了一种优化的推荐算法；提出一种基于团队意识的社会化协作分发机制和方法；提出基于新的关系分类视角的改进的度量空间向量嵌入模型TransHR；提出了一种新的近似子图匹配查询算法ASMQ；提出了基于时间序列同时考虑社交影响因素的远期离网预测模型。..项目对广义协作关系中对影响用户行为的关键因素开展了多案例分析，成果在国内、外学术期刊和相关领域的重要国际会议上发表论文13篇（其中SCI 论文6 篇），申请相关领域专利3 项。成果可为扁平在线社会协作平台、运行机制和优化策略方面的应用实践提供理论指导。