Employment discrimination analysis of Library and Information Science based on entity recognition

Abstract

Abstract The main purpose of this research is to provide an aggregated description of current employment discrimination in Library and Information Science (LIS) job market in Mainland China utilizing entity recognition approach. Specifically, a recruitment corpus with ontology-driven rules is firstly built. Then, the Bi-LSTM-CRF model on an annotated subset of the corpus is trained and verified by the rest of the corpus. Further, a quantitative statistics of the discrimination (via entities annotations) and an aggregation of the prediction of annual demand for jobs were conducted. Finally, we evaluate our approach by collecting 5297 LIS job advertisements in the public sector from 2015 to 2019 and conclude the result that average F1 of the entity recognition on 520 posts with 5411 entities is up to 91.06%. We statistically find that there exists serious institutional and employer discrimination ranging from political status (22.5%), age (15.4%), household registration (14.0%), to educational background (13.8%), etc. To our best knowledge, this is the first study investigating employment discrimination in the field of LIS in mainland China.