SBIR Phase I: A Semantic Data-Driven Human Capital Recommendation System

该SBIR第一阶段项目建议将语义匹配和知识获取循环的组合应用于人力资本选择过程，以生成按顺序排列的合格候选人短列表。现有的工作匹配方法的问题在于，它们主要依赖于关键词搜索，缺乏全面的学习，包括对推荐质量的反馈。机会是建立技术来创建一个候选人推荐系统，包括对其表现相关性的反馈。研究目标是确定整合语义匹配和机器学习是否为创建商业上可行的预测性候选推荐系统提供了可行的基础。本研究中开发的自然语言处理技术和本体知识库，有可能促进使用自然文本来表达技能和经验的可用性的知识和理解。预期的结果是创建一个人力资本本体结构和规则，从等级顺序成功的反馈循环中学习，以改善语义匹配。虽然该项目评估了本体和机器学习对提高候选人推荐质量的重要性，但结果也增强了在更广泛的劳动力科学研究中结合本体和机器学习的技术理解。该产品的更广泛/商业影响是，通过向雇主提供更准确的候选名单，显著提高被选中申请人的成功率，并提高雇主以更低成本高效管理人力资本选择的能力。在美国，企业每年花费超过60亿美元用于在线招聘，确定最合格的候选人将增加投资回报，最终产生更大的组织能力和竞争优势。通过实施基于技能、资格和从历史选择中进行机器学习的自动化求职者排名过程的技术，企业将能够降低与识别优质求职者相关的成本，同时应对人才短缺日益严峻的挑战，特别是在科学、技术、工程和数学（STEM）专业。反馈回路提供了知识获取的整合；因此，随着时间的推移，改进结果并减少识别合格候选人所需的努力。这种对人才数据的洞察将帮助企业做出更明智的招聘、雇佣和其他资源管理决策。这一创新有可能扩展到不同来源的人才信息，为许多人才管理功能提供大数据分析解决方案，并为人力资本管理提供前所未有的见解。

This SBIR Phase I project proposes to apply the combination of semantic matchmaking and a knowledge acquisition loop to human capital selection processes to generate a rank-ordered short-list of qualified candidates. The problem with available job matching methods is they rely primarily on keyword searches and lack comprehensive learning that includes feedback on quality of recommendations. The opportunity is to build technology to create a candidate recommendation system incorporating feedback on relevance of its performance. The research objective is to determine if integrating semantic matching and machine learning provides a feasible basis for creating a commercially viable predictive candidate recommendation system. The natural language processing techniques, and the ontology knowledge base developed during this research, have the potential to advance knowledge and understanding of using natural text to express the availability of skills and experience. The anticipated result is the creation of a human capital ontology structure and rules that learn from feedback loops of rank-order successes to improve semantic matching. While this project assesses the significance of ontology and machine learning to improve the quality of candidate recommendations, the results also enhance technological understanding of combining ontology and machine learning within broader work-force science research.The broader/commercial impact of this product is to significantly increase the success of selected applicants and improve employers' ability to manage human capital selection efficiently and at less cost by presenting them with a more accurate short-list. Identifying the most qualified candidates should increase the return on investment in over $6 billion spent annually by businesses in the United States for online recruiting efforts ultimately resulting in greater organizational competence and competitive advantage. By implementing technology that automates the process of rank-ordering job candidates based on skills, qualifications, and machine-learning from historical selections, businesses will be able to reduce costs associated with identifying quality job candidates while dealing with the growing challenge of talent shortages, especially in Science, Technology, Engineering, and Math (STEM) professions. The feedback loop provides for knowledge acquisition to be integrated; thus improving results over time and reducing effort needed to identify qualified candidates. Such insights into talent data will help businesses make more informed recruiting, hiring, and other resource management decisions. This innovation has the potential to extend to talent information in disparate sources offering Big Data analytics solutions for many talent management functions and providing unprecedented insights for human capital management.