AUTHOR=Ren Zhuo-Ming , Zheng Yue , Du Wen-Li , Pan Xiao TITLE=A Joint Model for Extracting Latent Aspects and Their Ratings From Online Employee Reviews JOURNAL=Frontiers in Physics VOLUME=Volume 10 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/physics/articles/10.3389/fphy.2022.822351 DOI=10.3389/fphy.2022.822351 ISSN=2296-424X ABSTRACT=The personal description of the company associated with job satisfaction, company culture, and opinions of senior leadership becomes available on the workplace community website like the Glassdoor.com which allows people to evaluate and review the companies they have worked for or are working for. However, it is almost impossible to read all of the different and possibly even contradictory reviews and make an accurate overall rating. Therefore, extracting aspects or sentiments from online reviews and the corresponding ratings is an important challenge. We collect online anonymous employees' reviews of the Fortune Global Top100 companies from the Glassdoor.com, totaling more than 200,000. The data of numerical evaluation is the overall rating ranging from 1 to 5 stars, and the company's five aspects: Work/Life Balance, Culture \& Values, Senior Management, Career Opportunities and Salary \& Benefits. The data of reviewed context includes pros(positive comments about the company), cons(negative comments), advices(suggestions for the company). Here, based on the data of numerical evaluation and reviewed context, we are proposing a joint rules-based model which firstly learning aspect extraction aspects from online reviews and calculates the relative weights of different aspects by the Boot-strapping semi-supervised learning, and then uses latent rating regression to update the corresponding ratings according to the relative weights of different aspects. Our experimental evaluation has shown better results as compared with an un-supervised learning of the latent dirichlet allocation.