SBIR Phase I: A Radically Efficient Search and Visual Mapping Tool for the Social Sciences

这个SBIR第一阶段项目旨在开发一个交互式搜索和可视化工具的工作原型，重新定义对学术研究成果感兴趣的社会科学家和商业从业者如何识别这些发现。用户不必阅读大量的、无结构的、通常不相关的文本结果(如传统学术搜索引擎产生的结果)，而是(A)通过结构直观、可点击的可视地图，即时查看和轻松导航研究成果，以及(B)通过语义智能索引和可视协调，准确地识别与研究成果最相关的论文。这种提高效率的工具使研究结果更容易理解和探索，同时大幅减少对此类发现的搜索时间(减半甚至更少)，这最终将提高美国大学的研究工作效率。此外，各种调查显示，市场营销或管理等应用领域的从业者高度重视社会科学的学术研究，但往往发现此类研究难以理解。拟议的解决方案将社会科学成果提炼成易于理解的格式，从而促进学术界和企业之间的知识转移，并提高学术研究对社会的价值。最后，通过向学者和商业从业者推销基于订阅的服务，拟议的工具有可能在长期内产生巨大的商业价值(年收入高达5000万美元)。拟议的工具从根本上改变了社会科学的现有搜索范式，改变了学术论文研究成果的索引方式和可视化呈现方式。它结合了后端的创新(即使用自然语言理解(NLU)自动从学术研究论文中提取概念和因果关系，并根据一组特定学科的同义词对这些概念进行语义分类)和前端的创新(即使用聚合因果映射以交互地图的形式表示学术文献，这些地图可以可视化地探索和缩小范围，以便准确定位相关论文)。主要的研究目标是测试(1)使用自然语言理解准确地识别和提取大量社会科学研究论文(约1,000篇已发表论文)中描述的研究的潜在概念/变量和因果结构，以及(2)以学术和非学术用户都可以直观地理解和导航的因果地图的形式自动呈现所提取的信息的可行性。如果一组测试用户使用所提出的工具来成功地识别考察特定概念和关系的研究论文，并且这样做的时间大约是使用传统学术搜索引擎来完成相同任务所需时间的一半，那么就达到了这一研究目标。

This SBIR Phase I project aims to develop a working prototype for an interactive search and visualization tool that redefines how social scientists and business practitioners interested in scholarly research findings identify such findings. Instead of reading through numerous, unstructured, and often irrelevant text results (such as those produced by traditional academic search engines), users get to (a) instantly view and easily navigate research findings via intuitively-structured, clickable visual maps, and (b) accurately identify those papers most relevant to them, thanks to semantically intelligent indexing and visual reconciliation. This efficiency-enhancing tool makes research findings substantially easier to understand and explore, while drastically cutting down search times for such findings (to half or even less) which will ultimately enhance the efficiency of research endeavors at U.S. universities. Moreover, various inquiries have shown that practitioners in applied fields such as marketing or management place high value on academic research in the social sciences, yet often find such research difficult to understand. The proposed solution distills social science findings into an easily digestible format, hence facilitating the knowledge transfer between academia and businesses, and enhancing the value of academic research to society. Finally, by being marketed as a subscription-based service to both academics and business practitioners, the proposed tool has the potential to generate substantial commercial value in the long term (up to $50 million in annual revenue).The proposed tool fundamentally alters the existing search paradigm in the social sciences, by changing both the way in which research findings from academic papers are indexed, and how such findings are visually presented. It combines an innovation on the back-end (i.e., using Natural Language Understanding (NLU) to automatically extract concepts and causal relationships from academic research papers, and semantically categorize those concepts against a set of discipline-specific thesauri) with an innovation on the front-end (i.e., using aggregate causal mapping to represent the academic literature in the form of interactive maps that can be visually explored and narrowed down in order to precisely locate relevant papers). The main research objective is to test the feasibility of (1) using NLU for accurately identifying and extracting the underlying concepts/variables and causal structure of the studies described in a large set of social science research papers (approximately 1,000 published papers), and of (2) automatically rendering the extracted information in the form of causal maps that both academic and non-academic users can intuitively understand and navigate. This research objective has been reached if a group of test users employ the proposed tool to successfully identify research papers examining particular concepts and relationships, and do so in about half the time needed when using a traditional academic search engine for the same task.