從線上員工評論萃取雇主品牌：萃取方法開發及其應用

Abstract

雇主品牌在當前高度競爭的人力市場中至關重要，是企業建立人力資本的有效策略。然而，過往研究對於雇主品牌資訊的取得主要仰賴問卷方法，因而受到資料規模、即時性不足、員工沉默等限制；僅有少數研究對線上員工評論進行有限運用，而未能深入捕捉評論文本中蘊含的雇主品牌洞察。 本研究提出一個基於線上員工評論的雇主品牌萃取方法，該方法結合了預訓練語言模型、注意力機制及多種分類架構，充分利用雇主品牌與員工評論間的細緻語意線索，並考量不同雇主品牌面向間的交互影響。此外，本研究進一步將萃取結果應用於企業層級的員工流失率及員工滿意度分析。實驗結果顯示，我們的雇主品牌萃取方法具有良好表現，亦能有效揭示雇主品牌與兩項關鍵人力資本指標間的關聯。 本研究不僅提供了一種新穎的雇主品牌萃取方式，該方法更可作為企業雇主品牌管理與人力資源策略的重要依據，總體而言拓展了雇主品牌的應用場景，並具備高度實務價值。

Employer brand plays a critical role in today’s highly competitive labor market and serves as an effective strategy for organizations to build human capital. However, previous research has primarily relied on survey-based methods to obtain employer brand information, which are limited in terms of data scale, timeliness, and employee silence. Only a few studies have made limited use of online employee reviews and have not adequately captured the in-depth employer brand insights embedded in their textual content. This study proposes an employer brand extraction method based on online employee reviews. The method integrates a pretrained language model, an attention mechanism, and diverse classification architectures, fully leveraging the nuanced semantic cues between employer brand dimensions and employee reviews, while also accounting for interdependencies among the dimensions. Furthermore, the extracted results are applied to company-level analyses of turnover rate and employee satisfaction. Experimental results demonstrate that our employer brand extraction method yields promising performance and effectively reveals the relationships between employer brand and the two key human capital indicators. This study not only introduces a novel approach to employer brand extraction but also provides important guidance for corporate employer brand management and human resource strategies. Overall, this research broadens the application scope of employer brand and offers substantial practical value.