Semantic search using deep networks****

使用深度网络进行语义搜索****

• 批准号: 531051-2018
• 负责人: Milios, Evangelos
• 金额: $ 1.82万
• 依托单位: Dalhousie University
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=659787
• 关键词: Semantic; search; using; deep; networks

Hire Ground is a leading service provider of Applicant Tracking Software (ATS) supporting its clients in their human resource hiring and retention efforts. A key component of ATS is the ability to compare candidates regarding their suitability for a given job description, the most time-consuming and labour intensive task of the hiring process because it depends on the number of candidates, which can often be in the hundreds. Matching resumes with job descriptions computationally is a major challenge due to language variety and nuance used in resumes. **We propose to adapt novel semantic methods developed in our lab in the last several years for automatically ranking resumes of applicants to a given job ad on their suitability for the position. The relevance of expertise will be estimated based on the semantic distance between the job description and each resume based on a term embedding representation of resume and job description. We will evaluate different variants of deep networks, such as Convolutional and Recurrent Neural Networks, and different variants of word embeddings. A key component of the proposed methodology is the human-in-the-loop: the human expert (human resources specialist of a company) will provide feedback to allow the training of the deep network models with minimal training data. The research will advance the state of the art in semantic search and help the industry partner achieve a competitive advantage in their market. ****

Hire Ground是一家领先的申请人跟踪软件（ATS）服务提供商，为客户的人力资源招聘和保留工作提供支持。“自动测试服务”的一个关键组成部分是比较候选人是否适合某一职务说明的能力，这是招聘过程中最耗时和最耗力的任务，因为它取决于候选人的人数，而候选人的人数往往可能有数百人。由于简历中使用的语言多样性和细微差别，通过计算将简历与职位描述匹配是一个重大挑战。 ** 我们建议采用我们实验室在过去几年中开发的新语义方法，用于自动对给定招聘广告的申请人简历进行排名，以确定他们是否适合该职位。专业知识的相关性将基于职位描述和每个简历之间的语义距离来估计，该语义距离基于简历和职位描述的术语嵌入表示。我们将评估深度网络的不同变体，例如卷积和递归神经网络，以及单词嵌入的不同变体。所提出的方法的一个关键组成部分是人在回路：人类专家（公司的人力资源专家）将提供反馈，以允许用最少的训练数据训练深度网络模型。这项研究将推进语义搜索的最新发展，并帮助行业合作伙伴在其市场中获得竞争优势。****