CapStyleBERT: Incorporating Capitalization and Style Information into BERT for Enhanced resumes parsing
Maiia Y. Bocharova, Eugene V. Malakhov · 2024
The growing prevalence of online recruitment has led to a significant accumulation of resumes within recruitment databases. These resumes are typically crafted in diverse formats, incorporating various font sizes, colors, and layouts, all aimed at capturing the recruiters' interest. However, the quality of extraction outcomes significantly impacts all subsequent tasks, including automated vacancy-profile matching, applicant ranking etc. Rule-based methods, grammar and deep-learning based methods have been introduced to parse and structure resumes accurately, however, these methods tend to overlook the rich information present in capitalization and style. In the paper a new Style and Capitalization aware modification of the BERT model is introduced and the effectiveness of the proposed method is proven.