悬赏模式下文本类众包方案过滤研究

发布时间：2018-03-03 00:01

本文选题：众包　切入点：悬赏模式　出处：《合肥工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：众包是开放式创新环境下产生的一种网络新模式,随着互联网的快速发展,众包模式受到了企业的青睐,涌现出以亚马逊土耳其机器人、猪八戒网、任务中国等为代表的诸多众包网络平台。悬赏模式是最主要的众包模式。悬赏模式下,一个众包任务能收集到多个众包方案,发包方从中选择合适的方案。目前众包方案的选择往往以人工选择的方式进行,这种方式效率低、成本高、主观性强。如何快速从大量众包方案中选择合适方案是众包平台亟待解决的问题。本文从文本信息过滤的角度研究悬赏模式下的文本类众包方案选择,构建一个众包方案的双层过滤模型,目的在于实现众包网络平台的方案自动化初筛选。过滤模型包括三个功能模块:众包文本数据预处理、根据Rocchio算法建立的双层过滤器和过滤处理模块。其中,众包数据的预处理分四步进行:首先,收集整理众包平台上发包方发布的任务和对应接包方提交的方案文本数据;其次,用中文分词工具NLPIR对众包文本分词处理,得到具有词频统计的特征词集;再次,根据文档词频法提取特征词,建立特征矩阵;最后,采用向量空间模型描述众包文本内容。为了验证过滤模型的可行性,用该过滤模型对众包网络平台上的实际众包方案过滤筛选,并将过滤结果与人工选择的结果比较。结果表明,本文的提出过滤方案具有一定的有效性和适用性,能够为实现众包方案自动化初筛选提供了一种可行思路和理论基础,有一定的实际意义和现实价值。
[Abstract]:Crowdsourcing is a new network model under the open innovation environment. With the rapid development of the Internet, crowdsourcing model has been favored by enterprises. Task China and other representative crowdsourcing network platforms. The reward model is the most important crowdsourcing model. Under the reward mode, a crowdsourcing mission can collect multiple crowdsourcing schemes. At present, the selection of crowdsourcing schemes is often carried out by manual selection, which is inefficient and costly. How to quickly choose the appropriate scheme from a large number of crowdsourcing schemes is an urgent problem to be solved in crowdsourcing platform. From the perspective of text information filtering, this paper studies the selection of text crowdsourcing schemes in reward mode. A two-layer filtering model of crowdsourcing scheme is constructed, which aims to automate the initial screening of crowdsourcing network platform. The filtering model includes three functional modules: crowdsourcing text data preprocessing, The preprocessing of crowdsourcing data is divided into four steps: first, collecting and arranging the tasks issued by the provider and the corresponding scheme text data submitted by the receiver on the crowdsourcing platform; secondly, The Chinese word segmentation tool NLPIR is used to deal with crowdsourcing text participle, and the feature word set with word frequency statistics is obtained. Thirdly, the feature words are extracted according to the document word frequency method, and the feature matrix is established. The vector space model is used to describe the content of crowdsourcing text. In order to verify the feasibility of the filtering model, the filtering model is used to filter the actual crowdsourcing schemes on the crowdsourcing network platform. The comparison between the filtering results and the results of manual selection shows that the proposed filtering scheme is effective and applicable, and can provide a feasible train of thought and theoretical basis for the realization of automated initial screening of crowdsourcing schemes. Have certain practical meaning and realistic value.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F713.36

【相似文献】